Conference Proceedings

Caption 3: Caption 6: Caption 4: Caption 7: Caption 5: Caption 3 is the Original RGB Images were taken with a smartphone camera. Caption 4 is a prediction from the Random Forest model on the same image in caption 3. Caption 5 is the XG Boost prediction on caption 3. Caption 6 is a prediction on Caption 3 using the Naive Bayes model. Caption 7 is a prediction made on the SVM model trained using the data collected. The water-stressed plants are indicated by a green colour and the non-water stressed plants are indicated by a red colour for the Random Forest, XG Boost and Naive Bayes image predictions. Caption 5 are predictions from the XG Boost model that shows clear lines within its predictions. Both Captions 5 and 4 have clear predictions with distinct edges between the plant and background. Caption 6 and Caption 7 both have blurred predictions which is expected considering the accuracies from Table 2 are at 0.40 and 0.33 for the Naïve Bayes and Linear SVM respectfully. For the Linear SVM, the black represents the non-water- stressed plants, and the white represents water-stressed plants the image had very little distinguishable features which is indicative of poor model performance. Conference Proceedings 45

See the pairs of Capsicum annuum image samples for Caption 12: each of the following categories: Caption 8: Caption 9: Caption 8 is the original RGB Images taken with a Caption 10: UAV. Caption 9 is the visualisation of Random Forest Caption 11: model prediction made on the image in Caption 8. Additionally, Caption 10 shows XG Boost model 46 prediction visualisation based on Caption 8 while Caption 11 shows the Naive Bayes model prediction visualisation of Caption 8. Lastly, Caption 12 shows the Linear SVM model prediction visualisation of the original image in Caption 8. The crops are indicated by the green colour and the weeds are indicated by the red colour in the model prediction visualizations for the XG Boost (Caption 10), Random Forest (Caption 9), and Naive Bayes (Caption 11) models. For the Linear SVM model visualization, Caption 12, the grey represents the crops, and the black represents the weeds. Discussion Machine Learning Models Used 1. Support vector machine The Support vector machine (SVM) is a supervised classical machine learning algorithm best known for its strengths in solving classification problems but is also often widely used in regression analysis. The term “supervised” indicates that the algorithm requires labelled input and output data for training. The SVM determines the most suitable hyperplane for separating a set of data points referred to as support vectors. In simple terms, the hyperplane or decision boundary represents a subspace of one less dimension than that of the space within which it exists. The hyperplane acts as a boundary between sets of support vectors and optimally categorises the data points into distinct classes. The ideal position of the hyperplane is determined in such a manner that maximises the size of the margin, that is, the perpendicular distance between the hyperplane and the data points nearest to it from Conference Proceedings

either class . SVMs are intrinsically linear but can be The equation represents the maximum margin, that is applied as non-linear depending on the complexity of the perpendicular width of a hyperplane. the problem. 2. Random forest classifier In nature, support vectors will rarely be perfectly The random forest model is an application of ensemble classified using the linear approach. Therefore, SVMs learning which has many uses in predictive modelling, can be specially applied using a nonlinear approach for classification, and regression problems. Ensemble making classifications. For nonlinearly separable data, learning improves predictive accuracy by combining a soft margin is used to tolerate a specific number of several classifier models. The random forest classifier misclassifications. The ideal hyperplane in such a case comprises several decision tree models and mirrors the will maximise the size of the margin while attempting concept of “bagging” to decide on the result. “Bagging” to minimise the number of misclassifications on either or bootstrap aggregation involves training the models side. Additionally, experimentation with different independently on random subsets of data taken from types of kernels and tolerance levels has the potential the original data set. to help improve the performance of the SVM. Figure 2: Two decision tree models being combined to Of course, SVMs may not be suitable for all problems. form the random forest models. For example, SVMs are extremely sensitive to outliers. Robust variants of SVMs have been proposed to better It is important to note that these samples of data may detect outliers through the introduction of bounded not be unique since each sample is taken from the loss. In addition to this, the training complexity of a full data set and thus every row has an equal chance normal SVM is highly dependent on the size of the of appearing in each sample. This step is referred to dataset thus making SVMs more suitable for smaller as row sampling with bootstrap. The results of all the datasets. classifiers are then combined in the aggregation step where the majority class is chosen as the final output. The performance of an SVM also suffers when the Overall, this technique decreases the likelihood of dataset used is highly imbalanced and this can account overfitting by decreasing the variance of the decision for the 0.33 and 0.37 accuracy shown by the Linear tree models. Assessing feature importance from the SVM models created. Implementing class weights dataset at the pre-processing stage may also help avoid then becomes necessary since hyperplanes tend to the problem of overfitting. However, bootstrapping be skewed toward the minority class in imbalanced may introduce correlation between the models thus datasets. When using the radial base function kernel, producing less accurate results. noisy datasets can also be an issue. For this paper, a linear SVM was implemented through a stochastic gradient descent classifier with hinge loss. Overall, the performance for the water-stress data and weed detection data was poor at 0.33 and 0.37 accuracy respectively. Figure 1: Support vector machine (SVM) visualisation Conference Proceedings 47

The “feature randomness” technique is an alternative The equation for Gini Index is as follows: that allows each tree to use a random subset of the available features. This reduces the correlation between where Pj represents trees as well as offers a buffer to protect other trees the proportion of samples that belong to class c for a from possible errors. The performance of this model particular node. with respect to both water-stress and weed detection was 0.70 and 0.74 respectively. This may have been Entropy is defined as : due to the size of the dataset and the ability of the model to compensate for biasness. Information Gain is defined as : 3. Decision tree 4. Naive Bayes classifier Naive Bayes, named after Bayes’ theorem, is yet Decision tree is a supervised machine learning model another supervised learning classification algorithm. It widely used for classification problems. The name of is described as naive because it assumes that attributes this model is inspired by the tree-like structure that is are all independent of each other and have an equal formed through the process of predicting the class label. effect on the outcome. The Bayes theorem produces The process starts at a point called the root node which the probability of event A, given that another event B recursively branches into multiple leaf nodes until it has already occurred. In the context of the naive Bayes reaches the terminal node. The root node represents classifier, event B represents a vector of n features and the entire training set, that is, all of the inputs. The event A is the class variable. Hence, the goal of the branches represent a different decision that is made as algorithm is to find the probability of the class variable a result of comparing the attribute of the inputs to a given a combination of features. In this process, the specific attribute in question. The dataset then becomes probability of the class variable given each of the n split by the different decisions as it progresses through features must be calculated separately. Subsequently, this repeated process until class labels can be assigned these probabilities are combined to produce the final to the final subsets of the training data at the terminal classification probability. nodes. Terminal nodes can no longer be split. The performance of a decision tree is highly dependent on the set of attributes involved in the process as well as how the data is strategically split into subsequent nodes. XG boost is an implementation of gradient-boosted decision trees designed for speed and performance. As such the performance from this model had the highest accuracy. The main goal of each data split is such that each resulting dataset down the tree structure increases in homogeneity until they are sufficiently homogenous and can no longer be split based on the available attributes. Homogeneity is measured by Gini index and information gain, for example. The Gini index provides the probability of incorrectly classifying a feature when it is selected randomly; therefore, small values are considered favourable. Information gain uses the concept of entropy, the degree of uncertainty or randomness in a dataset, to determine how much information a feature provides about a class. This determines which attribute is used for comparison at the root node and how the data is split through other attribute comparisons until reaching the last split. 48 Conference Proceedings

Three popular versions of the naive Bayes classifier 0.92 accuracy are sufficient for identifying problematic are the Bernoulli naive Bayes, Gaussian naive Bayes areas on the farm that may require cleaning or and the multinomial naive Bayes. The family of weedicide. XG Boost is the most accurate of the naive Bayes algorithms are simple and well known models applied to the data. for their speed compared to other more complex algorithms. Additionally, they are suitable for large, Conclusion high-dimensional datasets since the number of features Overall, the project has achieved its goal of setting a scales linearly with time complexity. Although the baseline for the performance of our future models. The assumptions of the naive Bayes classifier may not next steps for the project are to collect additional data always hold in the majority of real-world applications, in the dry season, improve our established models, it outperforms many other algorithms when the compare the data labelling techniques we have done assumptions are met. thus far to improve the performance of the models and finally implement useful metrics for our target Challenges audience as part of our application. Due to the COVID-19 pandemic data collection and labelling is a slower process. The restricted movements The current models require image augmentation to have not only affected the data collectors and labellers improve their robustness of the models. Additionally, but also the farmers’ abilities to maintain their crops. labelling techniques previously established through Additionally, rainfall is an unregulated factor as such model testing can be improved. The community the data collected was not evenly distributed but rather members have established a need for understanding skewed towards over-watered data. and identifying plant deficiencies. The application of the models developed for the FAAIR project will ● The datasets which are currently feeding remain farmer-focused and accessible to all. into the AI algorithms are extremely small. This is due to the limited time for image Acknowledgements processing, labelling and filtering thus far. Resources and funding were generously sponsored by It also requires data collection through The Cropper Foundation, The National Geographic the field agent application. Additionally, Society, NVIDIA and Microsoft Azure. Special thanks data could not be collected during rainfall. to Mr Fazard Mohammed who participated in the water-stress experiment and to Mr Khalil Ali and Mr ● There were major delays in data collection Farouk Khan for allowing the team to collect images due to difficulties in obtaining the drone. from their farms. Therefore, due to the time constraints, planned deliverables were delayed and References results were compromised as a result. Dore, Mohammed H.I. 2005. “Climate Change and ● There is limited memory and disk space. To compensate for this, we scaled down the images Changes in Global Precipitation Patterns: to 12.5 times smaller than the original size. What Do We Know?” Environment ● The data is skewed due to rainfall conditions. The data was collected during peak rainfall International. Pergamon, May 25, times thus skewing the data by creating a class imbalance which favours water-stressed plants. 2005. https://www.sciencedirect.com/ ● If these concerns are addressed, it stands to science/article/pii/S0160412005000553. reason that there will be a significant increase in the performance of the models built. FAO. 2022a. “Digitalization Is the Way of The purpose of the models is to identify issues with the New Life and the New Economy.” FAO. excessive water or weed growth. Thus far a 0.76 and Accessed March 1, 2022. https://www.fao.org/ news/story/en/item/1400816/icode/. FAO. 2022b. Small island developing states: FAO convenes a global dialogue to update on the latest initiatives. Small Island Developing States: FAO convenes a global dialogue to update on the latest initiatives. https:// www.fao.org/asiapacific/news/detail-events/ en/c/1470970/ Retrieved February 2022. Conference Proceedings 49

Fontes de Meira, Luciana, and Willard Phillips. 2019. OECD. 2022. Technology and Digital in Agriculture.” “An Economic Analysis of Flooding in the OECD. Accessed March 23, 2022. https:// Caribbean.” STUDIESAND PERSPECTIVES, www.oecd.org/agriculture/topics/technology- 78, January 2019, 1–54. and-digital-agriculture/. Fintech News. 2021. [email protected], Moccia. Stawarz, Steve. 2021. “Artificial Intelligence “Artificial Intelligence in Agriculture: Using and Farming.” The Conservation Modern Day AI to Solve Traditional Farming Foundation, August 30, 2021. https://www. Problems.” August 2, 2021. https://www. theconservationfoundation.org/artificial- fintechnews.org/artificial-intelligence-in- intelligence-and-farming/ agriculture-using-modern-day-ai-to-solve- traditional-farming-problems/. Tuquero, Joe, Raymond Chargualaf, and Mari Marutani. 2018. “Growing Bok Choy (Brassica Ganpat, W., Badrie, N., Walter, S., Roberts, L., Nandlal, Rapa Chinensis Group) Varieties for Guam.” J., & Smith, N. 2014. Compliance with good UNIVERSITY OF GU AM COLLEGE OF agricultural practices (gaps) by state-registered NATURAL & APPLIED SCIENCES PPP, and non-registered vegetable farmers in no. 07:16. https://doi.org/www.uog.edu%2F_ Trinidad, West Indies. Food Security, 6(1), 61– resources%2Ffiles%2Fwptrc%2FBokChoy. 69. https://doi.org/10.1007/s12571-013-0322- pdf&clen=1845754&chunk=true. 4 Panda, D. P. and Rosenfeld A. 1978. “Image Segmentation by Pixel Classification in Hickey, Gordon M., and Nigel Unwin. 2020. (Gray Level, Edge Value) Space,” in IEEE “Addressing the Triple Burden of Malnutrition Transactions on Computers, vol. C-27, no. 9, in the Time of COVID-19 and Climate Change pp. 875-879, DOI: 10.1109/TC.1978.1675208. in Small Island Developing States: What Role for Improved Local Food Production?” Food Security 12, no. 4: 831–835. https://doi. org/10.1007/s12571-020-01066-3. 50 Conference Proceedings

A survey of pet shop operations in Trinidad, Republic of Trinidad and Tobago R. A. Macfarlane1 and G. W. Garcia2 1PhD Candidate, Department of Food Production, Faculty of Agriculture and Science, University of the West Indies, St Augustine Campus 2Retired Professor of Livestock Science, Department of Food Production, Faculty of Agriculture and Science, University of the West Indies, St. Augustine Campus A study was conducted on 44 pet shops in Trinidad between May 2019 and February 2020 comparing their existing modus operandi to international best practices regarding disease management, animal welfare and industry safety. The SPSS 17 statistical package was used to compare means among the three pet shop sizes and among the four geographic locations. Results showed that 66.3% of the answers to the relevant questions indicated that pet shops were vulnerable to disease pathogens; 32.3% of answers to the relevant questions indicated that animal welfare standards were not being met and 56.5% of answers to the relevant questions indicated that health and safety standards were not met. There was a lack of proper disease prevention protocols and record keeping. It was recommended that there was a need for pet shop legislation in Trinidad and Tobago to address standards and best practices by instituting licensing and oversight for pet shop establishments. Key words: Disease, animal welfare, safety Introduction regarding all relevant aspects of animal specimens Pet shops are establishments which carry on the retail along the supply chain including the final sale to the aspect of live animals and related items for the pet trade. customer (Vale of Glamorgan Council 2022; Pearson The animals traded include both domesticated and BTEC 2020). wild animals and also both native and exotic species. Pathogen introduction is addressed by ensuring In order to properly manage pet shop establishments appropriate transportation conditions (London many jurisdictions have established legislation, Borough of Croydon 2001; BCP 2019 and animal guidelines and codes of practice for various aspects of quarantine (DEFRA 2018; Animal Welfare Victoria routine operations. 2022; Northern Territory Government 2016; State of Licences to operate pet shops usually specify the Queensland 2008). There is also legislation to ensure conditions for operations including the number and that the relevant authorities were to be notified within a taxa of animals to be sold and the oversight authority given time frame of any disease outbreak (Ministry of (DEFRA 2018; London Borough of Croydon 2001; the Environment, Forestry and Climate Change, India Legizlazzjoni, Malta 2014). Ensuring that transactions 2018). are legal, micro-chipping or closed ring systems are Pathogen transmission is addressed by mandatory included so that animal specimens can be distinguished maintenance of enclosure cleanliness (London from other conspecifics (Ministry of the Environment, Borough of Croydon 2001; Legizlazzjoni Malta 2014); Forestry and Climate Change, India 2018; Brussels the implementation of trays between the top and Animal Welfare Council 2019). This is enhanced by bottom enclosures in tiered systems (Pearson BTEC the mandatory creation and maintenance of records 2020; Animal Welfare Victoria 2022; Burton 2008); Conference Proceedings 51

mandatory requirements for programme against pests East location was from Morvant Junction, Eastern (Vale of Glamorgan Council 2022; Burton 2008); and Main Road eastwards to Matelot and southwards to proper food preparation and storage (DEFRA 2018); the Churchill Roosevelt Highway; west was west of London Borough of Croydon 2001; London Borough Morvant Junction; Central was south of the Churchill of Barking and Dagenham 2018). Roosevelt to Claxton Bay; south was south of Claxton Bay. The survey was carried out by interview with the Animal welfare is addressed with guidelines for pet shop operator / owner or any available personnel. regular inspections (DEFRA 2018; London Borough of Barking and Dagenham 2018); forbiddance of purchase The survey was conducted during the period May or sale of un-weaned mammals (London Borough of 2019 to February 2020. A total of 44 pet shops were Croydon 2001; BCP 2019); and prevention of the sale surveyed and the surroundings were inspected to assess of puppies, cats, ferrets or rabbits aged under 8 weeks size, animal welfare and disease prevention efforts. (DEFRA 2018; Legizlazzjoni Malta 2014; Arizona Appdedix 1 is a map of the pet shop locations. State Legislature 2022). The questionnaire was designed with 85 questions, Jurisdictions thus ensure that their license have 20 had direct consequences regarding animal welfare, guidelines to minimize issues under three broad 20 to disease management, and 11 to environmental aspects: animal welfare, disease management and health and safety. Disease management was further industry safety. The objective of this paper is to analyzed from the perspectives of the potential for determine if pet shop operations in Trinidad are on disease introduction, disease transmission and disease par with international best practices regarding those spread. three aspects. The outcome of this study will be an appropriate guide towards future pet shop legislation Results for Trinidad and Tobago. Methodology There were 14 pet shops which did not have any animal specimens (mammal, bird or reptile) offered Step 1: A sample survey of pet shops in Trinidad was for sale at the time of the visit but stocked related conducted using a questionnaire. The sample size products. Of the 30 of those which had animals, six was based on surveying at least 75% of the estimated were small establishments of which one was located in population of 50 pet shops establishments with a 95% the west, two were located in the central area and three confidence level and 5% margin of error. The sampling in the south; 16 were medium sized establishments, design was descriptive to highlight differences six located in the east, one in the west, three in central of operations based on establishments sizes, and and six in south; eight were large establishments, three geographic locations. The SPSS 17 statistical package located in the east, one in the west, and four in south. was used to compare means among three pet shop sizes and among four geographic locations; for post-hoc, The following are responses by the 30 pet shops (which the Scheffe test was used when equality of variances had animals at the time of the visit) regarding animal was not violated, and the Games-Howell test was used welfare, disease management and industry safety when equality of variances was violated. (i) 68.0% of answers to the 20 questions Step 2: The results from the field survey were compared regarding disease management (408 of the to the international best practices. The locations of the 600 responses) confirmed that international establishments were documented with handheld GPS best practices were not met systems. GPS Geoplaner - GeoConverter Routeplanner online was used to generate a map of Trinidad showing a) 66.1% of the answers to relevant the spatial locations of the pet shops sampled. questions indicated that international best practices to prevent disease Pet shops were categorised into three sizes and four pathogen introduction were not met geographic locations. Small pet shops were entities with no more than 150 sq ft of ground space; medium b) 64.5% of the of the answers to relevant pet shops were entities between 150 - 450 sq ft ground questions indicated that international space; large pet shops exceeded 450 sq ft ground space. best practices to prevent disease pathogen transmission were not met c) 73.3% of the of the answers to relevant questions indicated that international 52 Conference Proceedings

best practices to prevent disease While the standards for animal welfare used in other pathogen spread were not met jurisdictions do not perfectly reflect the Trinidad (ii) 32.0% of the answers to relevant questions circumstances, the issues have not been officially indicated that international best practices to considered in a meaningful manner in Trinidad. The prevent animal welfare violation were not sale of un-weaned animals was found to be the norm met rather than the exception. This occurred because the (iii) 50.6 % of the answers to relevant questions customer wanted to develop a bond with the animal as indicated that international best practices it grew and developed. That dependency on the owner for industry health and safety were not met. for food eventually led to the animals ‘seeing’ the owners in a manner similar to its parent. This happened From the responses to questions regarding animal mainly when psitticine birds were purchased. welfare, disease management and industry safety, medium sized pet shops had the total highest percentage The separation of animals from parents too early was of responses which indicated that best practices were found to have resulted in increased aggression and not done (57.6%), followed by large establishments stereotypic behaviour of the offspring (Ahola et al. (21.9%) and small establishments (20.4%). Significant 2017). The need to bond with the parent is a natural differences among these groups are shown in Table 1 occurrence for social development purposes. Young in Appendix 2. animals learn from older conspecifics what should be included in their diets because trial and error learning at Geographically the highest percentage of responses such a young age can be lethal (Galef Jr and Geraldaut which indicated that best practices were not done 2001). Even though addressing issues of animal regarding animal welfare, disease management and welfare at pet shops in Trinidad cannot be all based on industry safety was from establishments in the south established practices elsewhere, some may be relevant (43.9%), followed by east (29.3%), central (15.1%) for future animal welfare guidelines for Trinidad and and west (11.56%). Tobago. Discussion It is not surprising that medium sized establishments are of most concern because there are much more of The fact that pet shops were located in all geographic those establishments than small or large ones. For areas in Trinidad meant a pet shop related disease similar reasons, pet shops in south Trinidad are of outbreak could originate from any geographical more concern than those in the east, west or central area. More importantly, effective disease monitoring geographic locations. Knowledge of pet shop locations, mechanisms were not in place anywhere. Also, some their physical conformation and modus operandi are of the various types of taxa kept had the capacity to be important factors for authorities who may wish to reservoirs or vectors for certain disease pathogens and establish a priority list for addressing pet shop issues. they were not examined by veterinarians prior to or upon arrival at the pet shops. That disregard for proper The shortage of trained personnel in pet shop disease management and health and safety practices environments where there is no guiding legislation increased the risks of such pathogens being introduced further disadvantages the industry regarding animal to pet shops. That should have been major cause for welfare, disease management and safety. Trinidad is concern as the responses to the relevant questions definitely in such disadvantaged position as only eight showed that that there was 66.1% vulnerability to of the establishments surveyed had personnel with disease pathogen introduction, 64.5% to transmission some sort of training. Therefore, Trinidad pet shops and 73.3% to spread. Moreover, some establishments have limited personnel capable of identifying potential could have been seen as variety stores which sold circumstances which put animals and humans at high pets. The absence of legislation, guidelines or codes risk for injury and disease and who can provide proper of practice to provide and uphold operating conditions advice to customers. at pet shops in Trinidad meant that the personal preferences of the operators was the determining factor The size of pet shops was shown to be a major in the variety and quantum of items sold; much different factor regarding operations. Table 1 shows that there from jurisdictions which have relevant licences. were significant differences among pet shop sizes regarding their vulnerability to disease and violation Conference Proceedings 53

of environmental health and safety issues and Table Recommendations 2 shows significant differences among pet shop sizes regarding products offered for sale. These findings are Legislation, codes of practice and written guidelines critical as they provide basic facts which should be are needed to be established for governance of pet considered for future pet shop legislation in Trinidad. shop operations in Trinidad to ensure that pet shops are licensed and there is a named oversight authority The survey outcome could have been affected by for monitoring and enforcing the law. The legislation the fact that questions were asked regarding some should be geared towards the elimination (or at aspects of pet shop operations which were clearly least reduction) of the existing level of vulnerability being carried out in an illegal manner. Even though to diseases, improve animal welfare and improve some answers to the questionnaire were questionable, industry safety. Further work should be done to seeking verification to those responses would not have identify any threats which can exploit the existing or led to any clarification. Moreso, it would have led to future vulnerabilities of the industry and implement a premature end of the interview process. Therefore, mitigation measures. information provided regarding staff vaccination, veterinary arrangements and emergency protocol Ethical approval (which was claimed to have existed) had to be accepted as it was given. No ethical approval was requested nor granted for this study. Conclusions References The absence of a requirement for a pet shop licence in Trinidad and Tobago has left these establishments Ahola, M.K., K. Vapalahti, K. and H. Lohi. 2017. “Early without the requisite guidelines for operations. This Weaning Increases Aggression and Stereotypic has resulted in individual operators managing their Behaviour in Cats”, Scientific Reports. 7: establishments according to personal preferences 10412 | DOI:10.1038/s41598-017-11173-5., leading to varying standards of management and September, 2017. Available in: https://www. vulnerability to pathogen introduction, transmission nature.com/articles/s41598-017-11173-5. and spread, and animal welfare and health and safety concerns. Overall, pet shop establishments are AnimalWelfareVictoria. 2022. “Code of Practice for the vulnerable to disease introduction, transmission and Operations of Pet Shops.” Available in: http:// spread. Their operations also cause concerns regarding agriculture.vic.gov.au/livestock-and-animals/ animal welfare violations and environmental health animal-welfare-victoria/domestic-animals-act/ and safety. codes-of-practice/code-of-practice-for-the- operation-of-pet-shops. A pet shop associated disease outbreak in Trinidad would be difficult to detect in its early stages and Arizona State Legislature. 2022. “Arizona Revised difficult to contain because of a lack of records which Statutes Title 44. Trade and Commerce. are needed to facilitate contact tracing. Chapter 11. Regulations Concerning The administering of medical care to animals Particular Businesses. Article 17. Pet by untrained personnel can lead to the misuse Dealers.” Available in: http://www.azleg.gov/ of pharmaceutical products and pathological viewdocument/?docName=https://www.azleg. misdiagnosis. Medium sized pet shop establishments gov/ars/44/01799.htm and those located in the south of Trinidad were the ones which were of most concern. Some existing pet BCP Council. 2019. “Animal Activity Licensing.” shop guidelines in other jurisdictions may be useful Available in: https://www.bcpcouncil.gov. to the development of future pet shop legislation for uk/Business/Licences-and-permits/List-of- Trinidad and Tobago. licences-and-permits/Animal-activity-licence. aspx. Brussels Animal Welfare Council (2019). Animal Welfare Report 2017-2018. Available in: https://leefmilieu.brussels/sites/default/files/ user_files/doc_20190524_en_activity_report_ aw_council_2017_2018.pdf 54 Conference Proceedings

Burton, R. 2008. “Animal Welfare Code of Practice: Ministry of the Environment, Forestry and Climate Animals in Pet Shops”. New South Wales Change, India. 2018. “Prevention of Cruelty to Department of Primary Industries. August Animals (Pet Shop) Rules, 2018.” The Gazette 1 2008. Available in: https://www.dpi.nsw. of India. Available in: http://www.egazette.nic. gov.au/__data/assets/pdf_file/0012/1310421/ in/WriteReadData/2018/189243.pdf. Animal-welfare-code-of-practice-animals-in- pet-shops.pdf. Northern Territory Government. 2016. “Northern Territories Animal Welfare Advisory DEFRA. 2018. “The Animal Welfare (Licensing Committee Guidelines for Care and Welfare of Activities Involving Animals) (England) of Animals in Retail Pet Shops.” Available in: Regulations. 2018. Guidance notes for https://industry.nt.gov.au/__data/assets/pdf_ conditions for selling animals as pets November file/0003/258843/guidelines-retail-pet-shops. 2018.” Available in: https://www.gov.uk/ pdf. animal-boarding-establishment-licence-wales- scotland. Pearson BTEC. 2020. “Pearson BTEC Level 3 National Extended Certificate in Animal Galef Jr., G. B. and L. Geraldaut, 2001. “Social Management Specification.” Available influences on foraging in vertebrates: causal in: https://qualifications.pearson.com/ mechanisms and adaptive functions.” Animal content/dam/pdf/BTEC-Nationals/Animal- Behaviour 2001, 61: 3–15. Available in: https:// Management/2016/specification-and-sample- www.sciencedirect.com/science/article/abs/ assessments/9781446936429_BTEC_Nat_ pii/S0003347200915574 ExtCert_ANMGT_Spec_Iss2C2.pdf. Legizlazzjoni Malta. 2014. “PROTECTION OF State of Queensland. 2008. “Queensland ANIMALS OFFERED IN PET SHOPS Code of Practice for Pet Shops.” (MINIMUM STANDARDS REGULATIONS). Available in: https://www.daf.qld.gov. LEGAL NOTICE 244 of 2013.” Available in: au/__data/assets/pdf_file/0017/192410/ https://legislation.mt/eli/sl/439.16/20130814/ QldCodeOfPracticeForPetShops.pdf. eng. Vale of Glamorgan Council. 2022. “Vale of Glamorgan London Borough of Barking and Dagenham. 2018. Council License Conditions and Guidance “Licence Conditions for Pet Shops.” Available for Pet Shops.” Available in: https://www. in: https://www.lbbd.gov.uk/sites/default/files/ valeofglamorgan.gov.uk/Documents/Working/ attachments/Pet-animals-licence-conditions. Business%20Support/Licensing/Licence- pdf. Conditions-and-Guidance-for-Pet-Shops.pdf. London Borough of Croydon. 2001. “Rules and Management for Pet Shops.” Available in: https://cupdf.com/document/managament-pet- shops. Conference Proceedings 55

Appendix 1 Figure 1: Map of Trinidad showing spatial existence of the pet shops surveyed 56 Conference Proceedings

Appendix 2 Table 1: Differences among pet shop sizes regarding disease vulnerability and environmental health and safety concerns Aspect of pet shop Impact of Percentage of P value for differences between operation concern establishments in each establishments category Food and water un- Disease 0.026, between medium and spoilt transmission Small Medium Large large Disease 0.026 between medium and large Food and water transmission .50 .81 .25 receptacles capable of 0.003 between medium and large being cleaned Disease .33 .81 ,25 Regular inspections transmission 0.027 between small and large done to detect .50 .81 .13 abnormalities 0.024 between medium and large Disease .67 .88 1.00 Suspected infected transmission .25 0.029 between medium and large animals also 0.049 between small and quarantined Disease spread .67 .81 medium Dogs vaccinated for < 0 .001 between small and distemper, parvovirus Disease .83 .88 .38 medium and hepatitis introduction 1.00 .69 .88 0.001 between small and large Keeping of records of .00 .81 .88 medical visits Environmental Personnel had first aid health and safety knowledge Environmental Personnel knowledge health and safety of psittacosis Table 2: Differences in items offered for sale among pet shops of different sizes Item offered for sale Percentage of establishments P value for differences between in each size category establishments Domestic companion Animals Small Medium Large 0.048 between medium and large .43 .32 .75 Agricultural .14 .32 .75 0.028 between medium and large chemicals .14 .60 0.036 between small and large Plants .67 0.017 between small and large Conference Proceedings 57

An evaluation of the Poultry Surveillance Unit, National Animal Disease Centre, Animal Production and Health Division, Ministry of Agriculture, Land and Fisheries R. A. Macfarlane1 and G. W. Garcia2 1PhD Candidate, Department of Food Production, Faculty of Agriculture and Science, University of the West Indies, St Augustine Campus 2Retired Professor of Livestock Science, Department of Food Production, Faculty of Agriculture and Science, University of the West Indies, St. Augustine Campus An evaluation of the Poultry Surveillance Unit (PSU) was conducted to determine its surveillance effectiveness, its challenges and the likelihood of the unit achieving early detection of an avian influenza outbreak. The evaluation consisted of a face-to-face interview with the resident Veterinary Officer; a completed questionnaire; an examination of the previous year’s annual report and an evaluation of the Trinidad and Tobago Avian Influenza Contingency Plan. The unit was found to be plagued by human resource and financial challenges and the contingency Trinidad and Tobago Avian Influenza Contingency Plan had elements of impracticality. It was concluded that the PSU’s surveillance efforts are ineffective and it was unlikely to achieve early detection of an avian influenza disease outbreak. It is recommended that the PSU should be removed from the public service structure and made a state authority; the Veterinary Diagnostic Laboratory should be made functional, training should be made available to PSU’s personnel; and the Trinidad and Tobago Avian Influenza Contingency Plan should be revised. Key words: Surveillance, effective, early detection Introduction several forms of active surveillance including vector surveillance (Verhagan et al. 2017; Hietala et al. 2005), Disease surveillance is the monitoring for pathogen sentinel surveillance (Zhao et al. 2016) and (Coman introduction, transmission and spread. The process et al. 2014), serological determination (Luciano et establishes patterns of progression, enabling al. 2011; Saadat et al. 2014), pathogen determination predictions and prevents or minimises the damage (Ricke et al. 2019) and participatory surveillance (Azhur which can result from disease outbreak. A disease et al. 2010). Disease surveillance is also conducted in surveillance system has four basic components: data passive modes which is after the infection or outbreak collection, data analysis, information dissemination has occurred. Passive surveillance techniques include and response (Nsubuga et al. 2018). These components clinical investigation (Wilking et al. 2009; Slavec et al. must be put in place for routine use and not only 2012), syndromic surveillance (Veldhuis et al. 2011), when a disease incident is reported. Objectives of a mortality investigation (Wilking et al. 2009) and disease surveillance system include demonstrating parameter monitoring (Lutz et al. 2019). freedom from disease, early detection of disease, measuring the level of disease, and discovering new Several methodical tools have been used in disease disease cases (Cameron 2012). Such an undertaking surveillance. Some used in predicting outcomes include requires technical and administrative support staff risk assessment (Martinez et al. 2007; Martinez et al. at different levels, financial resources, transport and 2011), data mining (Ortiz-Pelaez and Pfeiffer 2008), communications resources and laboratory facilities. regression modelling (Ludwig et al. 2010; Malladi et al. 2011), decision modelling (Souza-Monterio Active surveillance is a proactive approach and it et al. 2012; Goutard et al. 2012; Liu et al. 2007) and provides the most accurate and timely information to simulation modelling (Sanson and Thornton 1997). disease surveillance (Nsubuga et al. 2018). There are 58 Conference Proceedings

There are also spatial and environmental approaches influenza pathogen and to identify short comings with to disease surveillance including environmental the Trinidad and Tobago Avian Influenza Contingency modelling (Menzies et al. 2011; Li et al. 2009), spatial Plan (M.A.L.F. 2007). epidemiology (Linthicum et al. 2007; Dórea et al. 2013; Veldhuis et al. 2011), and cluster analysis (Odoi et al. Methodology 2009). For quantitative risk assessment approaches A questionnaire was used to obtain information the general technique is basic statistics (Carpenter and regarding the PSU’s surveillance approaches and tools. Greiner 2007; EFSA 2012). An interview was conducted with the veterinary officer with the day-to-day responsibilities at the PSU’s head The poultry industry in Trinidad has contributed office to determine the organizational challenges. An significantly to the agricultural economy. The analysis of the Trinidad and Tobago Avian Influenza operational structure of the poultry sector in Trinidad Contingency Plan was done to determine its practical facilitated 3,000 pluck shops (Phillips 2011). The feasibility and an analysis of the PSU’s Annual Poultry Surveillance Unit (PSU), National Animal Report for 2018-2019 was examined to determine Disease Centre (NADC), Animal Production and how the organization’s objectives were affected by its Health Division, Ministry of Agriculture, Land and challenges. Fisheries is the unit which has the responsibility of conducting disease surveillance on poultry farms in Results Trinidad. Information regarding surveillance modalities and tools used by PSU (from questionnaire): The two most likely avian influenza disease release pathways are associated with the illegal bird trade 1. Serological surveillance was the only form of and migratory waterfowl. Various means of active active surveillance used as part of the PSU’s and passive surveillance techniques are required to surveillance activities and none had been ensure early detection of any avian influenza pathogen conducted within the preceding 12 months due circulation. In that regard the PSU should have had to financial constraints. Clinical investigation the appropriate organisational structure, the required and mortality investigation were the only types authority, trained personnel, funding and equipment of passive surveillance done (Table 1) for effective surveillance and appropriate action. 2. Simulation modelling was the only surveillance The objectives of this study were to determine if the tool employed. This was done on no more than PSU’s avian influenza surveillance was effective, if the two occasions (Table 2). PSU was likely to achieve early detection of the avian Table 1: Completed questionnaire regarding surveillance modes in the 24 months prior to interview Name of surveillance approach Mode of surveillance Response Remarks Vector surveillance Active No Never done Sentinel surveillance Active No Never done Serological determination Active No None in the last 2 years Participatory surveillance Active No Never Clinical investigation Passive Yes Done Syndromic surveillance Passive Yes Never Mortality investigation Passive No Done Parameter monitoring Passive Yes Not applicable Table 2: Questionnaire results regarding PSU’s surveillance tools in the 24 months prior to interview Surveillance tool Response Spatial epidemiology (GIS) No Risk assessment No Data mining No Basic statistics No Regression modelling No Digital surveillance No Simulation modelling Yes Cluster analysis No Environmental modelling No Conference Proceedings 59

Information provided by the veterinary officer, PSU 10. There was a need for training at all levels in the regarding the unit’s avian influenza surveillance organisation. protocol: 1. The PSU is part of the public service structure Information from the Trinidad and Tobago Avian and funding used to operate the surveillance Influenza Contingency Plan (subsequent to an system is obtained from annual recurrent outbreak) revealed: expenditure namely the public sector investment projects (PSIPs) allocation 1. A multisector avian influenza task force was identified to become enabled following 2. There were some constraints: shortage of an declaration of an outbreak Administrative Officer I (AOI); late payments of travelling allowances; an increase in workload 2. An operations centre headed by the Senior from high staff attrition which resulted in low Veterinary Officer (SVO) would be established morale and decreased performance and a lack for communication with the Chief Veterinary of training opportunities at all levels Officer (CVO), the media, private contractors, members of the task force and law enforcement 3. The Veterinary Diagnostic Laboratory was personnel non-functional 3. The SVO will have the responsibility for 4. Active wild bird surveillance was reported to coordination and monitoring of activities have been done on high-risk migratory species during an outbreak. These proposed SVO and at wetlands sites within 5 km proximity to activities included coordinating with field poultry flocks staff, coordinating with the members of the Task Force, confirmation of positive cases, 5. The PSU’s poultry farm surveillance designating and identifying farms as an infected programme for avian influenza virus involved premise, suspected premise or dangerous serological testing using ELISA and AGID tests contact premise and investigating pathogen as well as rapid field test kits. The sampling release and exposure pathways protocol was 15 – 20 farms per month, sample sizes of 15 - 30 sera samples per pen and 2 – 3 4. Funds will be required for the procurement of ml blood per bird on broiler farms, layer farms maps, meals, personal protective equipment for and duck farms field staff, disinfectants and other consumables. 6. Passive surveillance was done on live Review of the PSU Annual Report from 2018-2019 bird markets, back yard poultry flocks and revealed that: pigeon lofts when resources were available and on dead wild birds using FluDetect®. 1. The PSU conducted surveillance on layer Active surveillance at live bird markets was chicken farms, broiler chicken farms, duck inconsistent and had been greatly reduced farms, poultry hatcheries, smaller backyard chicken farms, guinea fowl flocks, geese, 7. No surveillance had ever taken place at bird turkeys and racing pigeons singing competitions, pet shops or animal rehabilitation and rescue centers 2. Some of the large integrated private producers were left to conduct their own surveillance 8. Sampling intervals for migratory birds were done during the periods October – December 3. Other routine activities included inspection and March – May each year. Sampling intervals and landing of imported hatching eggs, table for resident wild birds were during the periods eggs and day old poultry at ports of entry January – February and June – September each and at hatcheries; inspection and export year. The PSU’s target for both groups together certification of avian species for export and was a minimum number of 298 samples per verification; extension services on poultry year husbandry to farmers, students and the general public; participation in poultry and agriculture 9. Screening and diagnosis from suspect farms related workshops, seminars, programmes and were done on a needs basis; priority action was exhibitions determined by the relevant veterinary officer for the district who also identified high-risk 4. There was some integration with other poultry farms systems as relevant information was received from UWI School of Veterinary Medicine 60 Conference Proceedings

Laboratory; stakeholders who received PSU’s delay was more than ample time for an infection to findings and recommendations were various rapidly spread under favourable circumstances. poultry associations, UWI School of Veterinary Medicine, Ministry of Health, Inter-American The Trinidad and Tobago Avian Influenza Contingency Institute for Cooperation on Agriculture, Food Plan did not provide any clear legal or administrative and Agriculture Organization and Carib Vet. mechanism which gave the Senior Veterinary Officer as a civilian authority to instruct law enforcement Discussion arms of the state during an outbreak. Without a proper Memorandum of Understanding (MoU) stating the With regards to surveillance modalities there are at least circumstances that civilian can instruct some particular four types of active surveillance which are applicable law enforcement or military personnel, coordination to the Trinidad poultry context but only serological among the relevant agencies with the police and the surveillance was ever employed. Also, there are four army will be ineffective during an outbreak. Again, applicable passive surveillance modalities which could such circumstances will only favour the pathogen by have been applied but only clinical investigation and allowing it more time to spread. Also, the Trinidad mortality investigation were ever done. Similarly, and Tobago Avian Influenza Contingency Plan made there are a total of nine surveillance tools which are no mention of the Forestry Division which has the applicable to Trinidad but only simulation modelling authority to enter into pet shops in Trinidad. Not was ever employed. Failure to recognise those other including this division, denied the PSU access one techniques and tools as options which could have of the two major possible sources of avian influenza enhanced surveillance was due to the fact that staff introduction, pet shops. required relative training. There was no guarantee that the large producers who The PSU’s organisational existence as part of the public were allowed to conduct self-surveillance regarding service made it conceptually ineffective. The unit was their farms would monitor properly or immediately unable to conduct surveillance activities at pet shops alert the PSU of a positive sample or even of a suspected because it had no authority to do so. It was unable solve case. Even if such an event were eventually reported, its own human recourses issues of hiring persons to fill there may be so much time elapsed that the pathogen the vacancies for Animal Health Assistant l posts as would have been given adequate time to spread. protocol dictated that the line ministry must coordinate with the Service Commissions Department to have the The inconsistent and reduced active surveillance at live posts filled, a process which takes an excessive amount bird markets (pluck shops) greatly reduced the ability of time to resolve. and opportunity to achieve early detection of an avian influenza outbreak. Those establishments have been Since prevention of an outbreak was very much shown to be important in the epidemiology of avian dependent on identification of circulating pathogens, influenza (van Kerkhove et al. 2011: Abdelwhab and no funds to carry out the relevant active surveillance for Hafez 2011; Amonsin et al. 2008; Soares-Magalhães at least 1 year left the country very much vulnerable to et al. 2010; Song et al. 2008; Shortridge and Stuart- the introduction of the avian influenza pathogen. Also, Harris 1982; Senne et al. 1993; Fournie et al. 2011). in the event that there was an outbreak, containment Considering that there are 3,000 such establishments efforts as planned in the Trinidad and Tobago Avian in Trinidad and there was a problem regarding their Influenza Contingency Plan could not be carried out waste disposal (Phillips 2011), they remain a potential successfully as there would be no funds to purchase avian influenza release and exposure interface. personal protective equipment, maps, disinfectants and meals for field staff as detailed in the Trinidad and The PSU’s concept of priority farms is those which are Tobago Avian Influenza Contingency Plan. located within 5 km from wetland areas. Even though that is in keeping with international best practices The lack of necessary training across the board hampered (ceteris paribus), each poultry farm has its own risk the development of the technical staff who were the factors for the introduction of pathogens, development ones to apply appropriate surveillance tools. Also, the of disease and spread of pathogens within the farm non-existent of an equipped, functioning veterinary and to other farms. While the location of the farm is diagnostic laboratory resulted in confirmatory tests important, the farm can have structural and operational having to be sent abroad, thus delaying any relevant biosecurity issues which should be considered in risk action by at least 2 weeks waiting for results. Such a determination. Based on the PSU’s protocol some real Conference Proceedings 61

high-risk farms not located within the 5 km radius pubmed.ncbi.nlm.nih.gov/20521726/ of a wetland may have been overlooked for priority Coman, A., D.N. Maftei, R.M. Chereches, E. Zavrotchi, surveillance. P. Bria, C. Dragnea, P.P. McKenzie, M.A. Valentine, and G.C. Gray. 2014. “Avian Conclusions Influenza Surveillance in the Danube Delta Using Sentinel Geese and Ducks”. Hindawi The PSU’s surveillance for avian influenza is Publishing Corporation Influenza Research and ineffective; it is plagued by funding issues and human Treatment. Volume 2014, Article ID 965749. resources issues such as staff training; it is unlikely to Available in: https://downloads.hindawi.com/ achieve early detection of an avian influenza outbreak archive/2014/965749.pdf and its attempts of bio-exclusion and biocontainment Cameron, A.R. (2012). “The consequences of risk- of the pathogen are likely to be ineffective. based surveillance: Developing output-based standards for surveillance to demonstrate Recommendations freedom from disease”. Preventive Veterinary The PSU should be removed from the public service Medicine Volume 105, Issue 4, 280-28 structure and made a state authority so that human Carpenter T.E. and M. Greiner. 2007. “An analysis of resource issues (such as training and recruitment) can be internally addressed. The Veterinary Diagnostic an early-warning system to reduce abortions Laboratory should be made functional and the Trinidad in dairy cattle in Denmark incorporating and Tobago Avian Influenza Contingency Plan should both financial and epidemiologic aspects”. be revised to provide the Senior Veterinary Officer with Preventive Veterinary Medicine 2007; 78: authority to give instructions to police and soldiers and 1–11. Available in: https://pubmed.ncbi.nlm. to include the Forestry Division in the task force. nih.gov/17169450/ Dórea, F.C., B.J. McEwen, W.B. McNab, J. Sanchez and C.W. Revie. 2013. “Syndromic surveillance References using veterinary laboratory data: algorithm combination and customization of alerts”. Abdelwhab, E.M. and Hafez, H.M. 2011. “An PLoS One. 8(12):e82183. doi: 10.1371/journal. overview of the epidemic of highly pathogenic pone.0082183. PMID: 24349216; PMCID: H5N1 avian influenza virus in Egypt: PMC3859592. Available in: https://pubmed. epidemiology and control challenges”. ncbi.nlm.nih.gov/24349216/ Epidemiol Infect. 139(5):647-57. doi: 10.1017/ EFSA. 2012. “Assessment of different monitoring S0950268810003122. Available in: https:// strategies for early detection of FMD incursion pubmed.ncbi.nlm.nih.gov/21281550/ in a free wild boar population area: a simulation modelling approach”. EFSA Journal 10: 2656 Amonsin. A., C. Choatrakol, J. Lapkuntod, R. Fournie, G., J. Guitiana, S. Desvauxb, V.C. Cuongc, Tantilertcharoen, R. Thanawongnuwech, S. D.H. Dungd, D.U. Pfeiffera, P. Mangtanie, and Suradhat, K. Suwannakarn, A. Theamboonlers, A.C. Ghaniet. 2011. “Interventions for avian Y. Poovorawan. 2008. “Influenza virus (H5N1) influenza A (H5N1) risk management in live in live bird markets and food markets,Thailand”. bird market networks”. PNAS vol. 110 no. 22 Emerg Infect Dis. 2008 Nov;14(11):1739- 9177–9182 42. doi: 10.3201/eid1411.080683. PMID: Goutard, F.L, S. Tavornpanich, I. Houisse. K. 18976558; PMCID: PMC2630752. Available Chanacha, W. Thanapongtharm, A. Cameron, in: https://www.ncbi.nlm.nih.gov/pmc/articles/ K.ina D.C. Stärk, F. Roger. 2012. “Optimizing PMC2630752/ early detection of avian influenza H5N1 in backyard and free-range poultry production Azhar, M., A.S. Lubis, E.S. Siregar, R.G. Alders, E. systems in Thailand”. Preventive Veterinary Brum, J. McGrane, I. Morgan and P. Roeder. Medicine 2012; 105: 223–234. Available in: 2010. “Participatory disease surveillance and https://www.sciencedirect.com/science/article/ response in Indonesia: strengthening veterinary abs/pii/S0167587712000062 services and empowering communities Herrick, K.A., F. Huettmann, & M.A. Lindgren. 2013. to prevent and control highly pathogenic A global model of avian influenza prediction avian influenza”. Avian Dis. 2010 Mar;54(1 Suppl):749-53. doi: 10.1637/8713-031809- Reg.1. PMID: 20521726. Available in: https:// 62 Conference Proceedings

in wild birds: the importance of northern blow flies in Central Europe”.  Int J Legal regions.  Vet Res  44,  42 (2013). https://doi. Med 133, 1267–1278. Available in: https://doi. org/10.1186/1297-9716-44-42. org/10.1007/s00414-018-1967-5 Hietala, S.K, P.J. Hullinger, B.M. Crossley, H. Kindle M.A.L.F. 2007. Trinidad and Tobago Avian Influenza and A.A. Ardans. 2005. “Environmental air Contingency Plan. Ministry of Agriculture, sampling to detect exotic Newcastle disease virus Lands and Fisheries, Trinidad and Tobago. in two California commercial poultry flocks”. Malladi, Sasidhar, J., T. Weaver, T.L. Clouse, K.E. J Vet Diagn Invest. 2005 Mar;17(2):198-200. Bjork and D.W. Trampel. 2011. “Moving- doi: 10.1177/104063870501700219. PMID: Average Trigger for Early Detection of 15825506. Available in: https://pubmed.ncbi. Rapidly Increasing Mortality in Caged Table- nlm.nih.gov/15825506/ Egg Layers.” Avian diseases  55(4), 603-610. Li, Y., P. Qin, and J. Zhang. 2018. “Dynamics Analysis Available in: https://www.semanticscholar. of Avian Influenza A(H7N9) Epidemic Model” org/paper/Moving-Average-Trigger-for-Early- Discrete Dynamics in Nature and Society Detection-of-in-Malladi-Weaver/4172126f80d Volume 2018, Article ID 7485310, 12 pages. 73de6d48ce64485b8a64653b16d https://doi.org/10.1155/2018/7485310. Martínez, M.,  M.J. Muñoz, A. De la Torre, B. Linthicum, K.J., C. Liu1, S. Lin, Y. Chen, K. Chun- Martínez, I. Iglesias and J.M. Sánchez- Min Lin, T. J. Wu, C. King. 2007. “A Rift Vizcaíno. 2007. “Risk assessment applied to Valley fever risk surveillance system for Spain’s prevention strategy against highly Africa using remotely sensed data: potential pathogenic avian influenza virus H5N1”. Avian for use on other continents”. Veterinaria Diseases 2007; 51: 507–511. Available in: Italiana;43: 663–674. https://journals.plos. https://pubmed.ncbi.nlm.nih.gov/17494621/. org/plosone/article/file?id=10.1371/journal. Martinez, M., A. Perez, A. de la Torre, I. Iglesias, pone.0000191&type=printable J. Sanchez-Vizcaino and M. Munoz. 2011. Liu, C.M., S.H. Lin, Y.C. Chen, K.C. Lin, T.S. Wu “Evaluating surveillance in wild birds by the and C.C. King. 2007. “Temperature drops and application of risk assessment of avian influenza the onset of severe avian influenza A H5N1 introduction into Spain”. Epidemiology and virus outbreaks”. PLoS One. 7;2(2):e191. Infection 139: 91–98. Available at: https://www. doi: 10.1371/journal.pone.0000191. cambridge.org/core/journals/epidemiology- PMID: 17297505; PMCID: PMC1794318. and-infection/article/evaluating-surveillance- Available in: https://pubmed.ncbi.nlm.nih. in-wild-birds-by-the-application-of-risk- gov/17297505/ assessment-of-avian-influenza-introduction- Luciano, R.L., A.L.S.P. Cardoso, G.F.Z. Stoppa, into-spain/72141A5B2FF24020FB551679EC- A.M.I. Kanashiro, A.G.M. de Castro, and C0647C E.N.C. Tessariet. 2011. “Comparative Menzies, F.D., K. Burns, J.G. Fallows, I.M. McKeown, Study of Serological Tests for Mycoplasma J. Gloster, L. Burgin, E.A. Courcier, synoviae Diagnosis in Commercial Poultry P.T. McNamee, J. Robinson, P.A. Robinson, Breeders”. Veterinary Medicine International F. McNeilly, and D.A. Abernethy. 2011. Volume 2011, Article ID 304349, 5 pages “Risk assessment, targeted surveillance and doi:10.4061/2011/304349. Available in: https:// policy formulation: practical experiences with www.hindawi.com/journals/vmi/2011/304349/ bluetongue”. Epidémiologie et Santé Animale Ludwig, A., M. Bigras-Poulin, P. Michel, D. Bélanger. 59–60: 113–115. 2010. “Risk factors associated with West Nile Ministry of Agriculture, Land and Fisheries virus mortality in American Crow populations Nsubuga, F., L. Bulage, I. Ampeire, J.K.B. Matovu, in Southern Quebec”. J Wildl Dis. 46(1):195- S. Kasasa, P. Tanifum, A.A. Riolexus and B. 208. doi: 10.7589/0090-3558-46.1.195. PMID: Zhuet. 2018. “Factors contributing to measles 20090033. Available in: https://pubmed.ncbi. transmission during an outbreak in Kamwenge nlm.nih.gov/20090033/ District, Western Uganda, April to August Lutz, L., M.A. Verhoff and J. Amendt. 2019. 2015”. BMC Infect Dis 18, 21 (2018). Available “Environmental factors influencing flight in: https://www.panafrican-med-journal.com/ activity of forensically important female content/series/30/1/9/full/ Conference Proceedings 63

Odoi, A., C.N. Carter, J.W. Riley, J.L. Smith, and Slavec, B. U. Krapez, A.J. Racnik, A. Hari, J. M. R.M. Dwyer. 2009. “Application of an Wernig, A. Dovc, M. Zadravec, R. Lindtner- automated surveillance data-analysis system Knific, C. Marhold and O. Zorman-Rojs. 2012. in a laboratory-based early warning system for “Surveillance of influenza A viruses in wild detection of an abortion outbreak in mares”. birds in Slovenia from 2006 to 2010”. Avian American Journal of Veterinary Research 70: Dis. 2012 Dec;56(4 Suppl):999-1005. doi: 247–256. Available in: https://pubmed.ncbi. 10.1637/10175-041012-ResNote.1. PMID: nlm.nih.gov/19231958/ 23402126. Available at: https://pubmed.ncbi. nlm.nih.gov/23402126/ Ortiz-Pelaez, A. and D.U. Pfeiffer. 2008. “Use of data mining techniques to investigate Soares Magalhães, R., J. A. Ortiz-Pelaez, K.L. Lai disease risk classification as a proxy for Thi, Q.H. Dinh, J. Otte and D.U Pfeiffer. 2010. compromised biosecurity of cattle herds in Associations between attributes of live poultry Wales”. BMC Veterinary Research 4: 24 trade and HPAI H5N1 outbreaks: a descriptive 1-16. Available in: cholars.cityu.edu.hk/ and network analysis study in northern Vietnam. files/88112552/11897384.pdf BMC Veterinary Research 2010 6:10. 13 pages. Available at: https://bmcvetres.biomedcentral. Phillips, R. 2011. Agritrade. Informed Analysis, Expert com/track/pdf/10.1186/1746-6148-6-10.pdf Opinion. An interview with Dr Desmond Ali, Robin Phillips and Carl Khan. Technical Song, M.T., T. Oh, H.J. Moon, D. Yoo, E.H. Lee, J. Centre for Agricultural and Rural Cooperation Lee, C. Kim, G. Yoo, H and Y. Choi. 2008. ACP­EU (CTA) “Ecology of H3 avian influenza viruses in Korea and assessment of their pathogenic potentials”. Ricke, S.C., K.M, Feye, W.E. Chaney, Z. Shi, H. Journal of General Virology (2008), 89, 949– Pavlidis and Y. Yang. 2019. “Developments 957. in Rapid Detection Methods for the Detection of Foodborne Campylobacter in the United Souza-Monteiro, D.M., L. Roman Carrasco, L. J. States”. Front. Microbiol. 9:3280. doi: Moffitt, J.C. Alasdair and L.J.C Cook. 2012. 10.3389/fmicb.2018.0328. Available in: “Robust surveillance of animal diseases: an https://www.frontiersin.org/articles/10.3389/ application to the detection of bluetongue fmicb.2018.03280/full disease”. Preventive Veterinary Medicine 2012; 105: 17–24. https://cpb-us-w2.wpmucdn.com/ Saadat, Y., S.A. Ghafouri, and F. Tehrani et al. 2014. blog.nus.edu.sg/dist/a/6358/files/2016/07/1- “An active serological survey of antibodies to s2.0-S0167587712000293-main-ylglhn.pdf Newcastle disease and avian influenza (H9N2) viruses in the unvaccinated backyard poultry Van Kerkhove, M.D., E. Mumford, A.W. Mounts, in Bushehr province, Iran, 2012-2013”. Asian J. Bresee, S. Ly, C.B. Bridges and J. Otte. Pac J Trop Biomed 4 (Suppl 1): S213-S216 2011. “Highly pathogenic avian influenza (H5N1): pathways of exposure at the animal- Sanson, R.L. and R.N. Thornton. 1997. “A modelling human interface, a systematic review”. approach to the quantification of the benefits of PLoS One Jan 24;6(1):e14582. doi: 10.1371/ a national surveillance programme”. Preventive annotation/3531c496-624f-40fe-92bb- Veterinary Medicine 1997; 30: 37–47. 5fb3d7d1d894. PMID: 21283678; PMCID: PMC3025925. Available in: https://www.ncbi. Senne, D. J. E. Pearson and B. Panigrahy.1993. nlm.nih.gov/pmc/articles/PMC3025925/pdf/ Live poultry markets: A missing link in the pone.0014582.pdf epidemiology of avian influenza. Pages 50–58 in Proc. 3rd Int. Symp. Avian Influenza. US Veldhuis, A., J. Barnouin, Y. Van der Stede, L. Ren Anim. Health Assoc., St. Joseph, MO. Available and M. Dispas. 2011. “The Belgian MoSS: a in: https://www.jstor.org/stable/3298670 monitoring and surveillance system for the early detection and identification of (re-emerging Shortridge, K.F & C.H. Stuart-Harris. 1982. An animal diseases”. Epidémiologie et Santé influenza epicentre? Lancet, 2, 812–813. Animale; 59–60: 190–192. Available in: https:// fsvee.be/wp-content/uploads/2019/08/2010- proceedings.pdf. 64 Conference Proceedings

Verhagen, J.H., P. Lexmond, O. Vuong, M. Schutten, Zhao, T., Y-H. Qian, S-H. Chen, G.L. Wang, M-N. Wu, J. Guldemeester, A.D.M.E. Osterhaus et al. Y. Huang, G-Y. Ma, L.Q. Fang, G.C. Gray, B. 2017. “Discordant detection of avian influenza Lu, Y.G. Tong, M.J. Ma and W.C. Cao. 2016. virus subtypes in time and space between “Novel H7N2 and H5N6 Avian Influenza poultry and wild birds; Towards improvement A Viruses in Sentinel Chickens: A Sentinel of surveillance programs”. PLoS ONE 12(3): Chicken Surveillance Study”. Front. Microbiol. e0173470. doi:10.1371/journal.pone.0173470 7:1766. doi: 10.3389/fmicb.2016.01766 Wilking, H., M. Ziller, C. Staubach, A. Globig. T.C. Harder et al. 2009. “Chances and Limitations of Wild Bird Monitoring for the Avian Influenza Virus H5N1 — Detection of Pathogens Highly Mobile in Time and Space”. PLoS ONE 4(8): e6639. doi:10.1371/journal.pone.0006639 Conference Proceedings 65

Poultry farm vulnerability to the introduction, transmission and spread of avian influenza in Trinidad, Republic of Trinidad and Tobago R. A. Macfarlane1 and G.W. Garcia2 1Retired wildlife biologist and Head, Wildlife Section, Forestry Division, Ministry of Agriculture, Land and Fisheries, Trinidad and Tobago 2Retired Professor of Animal Production, Department of Food Production, Faculty of Food and Agriculture, University of the West Indies, St. Augustine Campus, Trinidad and Tobago A quantitative vulnerability assessment was conducted on 105 poultry farms during March to October 2019 to determine their vulnerability to the avian influenza virus. Conceptual, structural, operational and overall bio-securities were determined for each farm based on a score system. A questionnaire was administered to obtain a score by the product of an assigned weight X (ranging between 2 – 10) and a frequency factor Y (ranging between 0 - 3). Overall scores for individual farms were determined by dividing the total score (∑Xi Yi) by the total possible score. Farms which had an overall score greater than or equal to 0.50 were deemed vulnerable to the avian influenza virus. It was found that a total of 59 farms (54.2% were found to be overall vulnerable broiler farms being of most concern; structural biosecurity was deemed most critical for the effective implementation of operational biosecurity and consequently overall biosecurity; farms located in the villages of Biche and Charuma were the most vulnerable. It is recommended that construction of poultry farms should be subjected to necessary planning approval; farms should be operating with standards; quantitative vulnerability assessments should be adopted as a regular monitoring mechanism for poultry farms across Trinidad. Key words: Conceptual, structural, operational, biosecurity, vulnerability Introduction which points were accumulated to a maximum amount or part thereof as done by Souris et al. (2014). Biosecurity is one of the most important aspects of disease prevention at poultry farms. Various aspects of Avian influenza can be introduced into poultry biosecurity can be used to develop a risk profile for a farms by windborne particles (Ssematimba et al. particular disease for any individual farm. The general 2012); contaminated food and contaminated fomites objective of the methodology used is to determine the (Alexander 2007; Pepin et al. 2014), free range poultry presence/absence of a given risk factor (Brouwer et (Koch and Elbers 2006) and the movement of people al. 2018; Bouwstra et al. 2017). Such a determination (Fasina et al. 2011). is based on a complex interaction among a number of factors from conceptual circumstances impacting Infection of poultry farms with any pathogen is a two- the farm’s physical location, structural circumstances step process: a release pathway from the reservoir or resulting in the state of farm’s physical structure and vector and an exposure pathway to the victims (Sims the day-to-day operational procedures involved in et al. 2003). farm management (Chowdhury et al. 2017). A scoring system was emphasised (Sims et al 2003) to avoid The presence of rodents in some French farms bias and ensure that threats are not overlooked where was a risk factor and was found to increase sero- knowledge was lacking. Aengwanich et al. (2014) conversion for avian influenza (Duvauchelle et al. suggested that the scoring system should be one in 2013). Madsen (2012) found that farms without a rodent control programme were 2.5 times more at risk 66 Conference Proceedings

for this pathogen infection than farms which had one used by Souris et al. (2014) implemented. This programme requires farms to have ● Scores for the three categories (conceptual, a place properly constructed for the storage of feed to structural and operational), as well as the prevent contact with wild birds, rodents and arthropods overall score, were determined by dividing the (OIE 2019). Contact with waterfowl, terrestrial birds, total scores for each farm [ ∑Xi Yi ] by the total domestic waterfowl, non-waterfowl synanthropic wild possible scores birds and shore birds are also major release pathways ● Farms which had an overall score greater than for the pathogen entering into poultry farms (Pantin- or equal to 0.50 were deemed vulnerable to Jackwood and Swayne 2009; Reperant et al. 2009; avian influenza Root et al. 2015; Kosh and Elbers 2006; Singh et al. ● Farms were assigned geographical clusters. 2018; Burns et al 2012). In Trinidad the capacity of poultry pens for broiler The interview was administered by going to each farm production ranges between less than 1,000 to more and obtaining information from the persons on the than 100,000 heads, the majority of poultry pens in compound, who were either the owners, managers or Trinidad are not in the best condition for production workers. The farm locations were then mapped with a and remain the most neglected aspect of poultry hand-held GPS unit. production (Ekwue et al. 2003). The industry has been stated to be negatively affected Data analysis by such poor infrastructure (Hosein and Pun 2004) and 1. Individual farm scores for conceptual wild birds have been identified as a potential threat to biosecurity, structural biosecurity, operational backyard and commercial poultry (Brown-Jordan et al. biosecurity and overall biosecurity were 2018). done by hand using the basic mathematical calculations described above. The objectives of this study were to: 2. Mapping of farms from the respective GPS 1. Determine the percentage poultry farms eastings and northings was done with ArcGIS vulnerable to avian influenza and which 10.3. production type was of most concern 3. Z scores for GPS eastings and northings were done by descriptive statistics (using SPSS 17). 2. Determine the variation of conceptual, 4. Creating spatial clusters of farms was done by structural and operational biosecurity on overall K-means cluster (SPSS 17). biosecurity in poultry farms and consequently 5. Mean comparisons were done by one way in geographical farm clusters in Trinidad ANOVA (SPSS 17). 6. The impact of individual vulnerability aspects 3. Identify the geographical location in Trinidad was determined by linear regression (SPSS which had the farms most vulnerable to avian 17); overall biosecurity being the dependent influenza. variable while the independent variables were Methodology conceptual biosecurity, structural biosecurity A field survey instrument, adapted from Souris et al. and operational biosecurity. (2014), designed to separate aspects of conceptual, structural and operational biosecurity regarding avian Results influenza was developed. Quantitative vulnerability There were 89 broiler farms, nine layer farms and seven scores were calculated by the following process: duck farms surveyed. A total of ten clusters ranging in size from 3 – 38 farms were demarcated; hereafter ● Each question on the survey form was related to referred to as Cluster 1, Cluster 2, Cluster 3… Cluster an element of biosecurity which was evaluated 10 as seen in Figure 1. and assigned a weight X ranging from 2 to 10 ● The score for each element evaluated was a product of the assigned weight and a frequency The results of the analyses are as follows: factor Y which ranged from 0 to 3. Both the ● Over half of the farms (59, 56.2%) were overall assigned weight and the frequency factor were vulnerable to avian influenza, 19 of them were consistent with international best practices as located in Cluster 6. Conference Proceedings 67

Figure 1: Map of Trinidad showing 10 poultry farm clusters ● Cluster 8 had the highest average overall There were significant differences between some vulnerability geographical clusters regarding: ● The number of farms vulnerable to conceptual ● Conceptual biosecurity F(8, 95) = 4.066, P ≤ biosecurity was 43 (40.9%) comprising of 35 0.001. broiler farms, five layer farms and three duck farms, 27 of them were located in Cluster 6. Post hoc tests revealed significant differences between ● The number of farms vulnerable to structural biosecurity was 45 (42.8%) comprising of 36 (i) Cluster 3 and Cluster 6, P = 0.021 broiler farms, four layer farms and five duck (ii) Cluster 4 and Cluster 6, P = 0.013. farms, 11 of them located in Cluster 3. ● Structural biosecurity F(9, 95) = 5.838, P ≤ ● The number of farms vulnerable to operational 0.001. biosecurity was 94 (89.5%), 22 of them were Post hoc test revealed significant differences located in Cluster 6 and 15 located in Cluster 3. between (i) Cluster 1 and Cluster 6, P = 0.005 ● Conceptual biosecurity had the most impact (ii) Cluster 1 and Cluster 9, P = 0.038 on overall biosecurity. Overall biosecurity (iii) Cluster 1 and Cluster 10, P = 0.023 was moderately positively correlated with (iv) Cluster 2 and Cluster 9, P = 0.047 conceptual biosecurity (r = 0.527); weakly (v) Cluster 8 and Cluster 9, P = 0.047. positively correlated structural biosecurity (r = 0.037) and was weakly negatively correlated ● Overall Biosecurity F (9, 95) = 3.219, P = with operational biosecurity (r = -0.028). 0.002. Conceptual biosecurity was weakly negatively Post Hoc test revealed significant differences correlated with structural biosecurity (r - -0.130) between and fairly moderately negatively correlated (i) Cluster 2 and Cluster 6, P = 0.008 with operational security (r = -0.271). Structural (ii) Cluster 2 and Cluster 10, P = 0.006. biosecurity was fairly moderately correlated with operational biosecurity (r = 0.214). 68 Conference Proceedings

● There were no significant differences among which had been demarcated for farming under a the three different farm production types government scheme six decades prior. This area also (broiler, layer and duck). contains other types of farming including cattle, pigs and sheep. It was no surprise then that this cluster Discussion contained the most conceptually vulnerable farms; thus, highlighting the need for proper farm planning The fact that 56.2% of farms were overall vulnerable and consequent management with regards to poultry to avian influenza should be a cause for concern as farms. In that regard, there is no one intervention the poultry industry is considered to be the most solution (or set of solutions) which may work on one important subsector in the agricultural sector. Since farm that is guaranteed to work for another farm. the potential risks regarding conceptual biosecurity Consequently, there are no set protocols to address could not be addressed directly because conceptual improvement of a farm cluster other than addressing circumstances could not be changed, structural each farm individually. Therefore, enforceable laws, biosecurity and operational biosecurity had to be fully regulations, guidelines, codes of practice for poultry functional where there were negative impacts resulting farm construction and management ought to be a from conceptual circumstances. The importance of requirement. structural biosecurity on farms in Trinidad was more than what was revealed by the regression. Even though Cluster 8 (Biche and Charuma) was found to be the the results showed that structural biosecurity had the cluster of most concern with an average overall least direct impact on overall biosecurity, it facilitates biosecurity score of 0.64. It was also the cluster with some aspects of effective operational biosecurity by the highest average for operational biosecurity, the providing the infrastructure such as a gate and restricted highest average for conceptual biosecurity and the pen access without the use of a footbath. This explains second highest average for structural biosecurity. the positive correlation between structural biosecurity and operational biosecurity (r = 0.214). Conclusion In the case of duck farms, their methods of production, 1. Just over 56% of all poultry farms surveyed which sometimes include open water ponds and were overall vulnerable to avian influenza, semi-intensive settings, contributed to low structural broiler farms were of most concern. biosecurity outcomes for that production system. These circumstances are not the same for broiler farms 2. The construction of poultry farms in Trinidad and layer farms. However, there were 36 broiler farms is not standardised with regards to farm (40.4%) and four layer farms (44.4%) vulnerable location or farm physical configuration. This regarding structural biosecurity. This suggests that has led to some significant differences among standards need to be implemented for the construction individual farms and geographical clusters of of poultry farms of all production types. farms regarding various aspects of biosecurity. Therefore, it was unlikely to have a standardised approach to treat with farms in the event of an Even though 53 broiler farms were not vulnerable avian influenza outbreak. regarding structural biosecurity, 83 of them were 3. Cluster 8 was the cluster of most concern. This vulnerable regarding operational biosecurity. Therefore, there were 30 farms with adequate structural cluster encompassed villages of Biche and Charuma in south east Trinidad. biosecurity where operational biosecurity could have Recommendations been implemented but was not. The non-existent of any relevant enforceable law, regulation, guidelines 1. Quantitative vulnerability assessments for or code of practice was the likely reason for such a individual farms should be adopted as a regular situation. monitoring mechanism for poultry farms across Trinidad. The individual farm vulnerability Farm clusters were not homogenous as individual farms scores and risk profiles should be the basis in varied in sizes, ownership, conceptual circumstances which bio-exclusion should be based. and ultimate management. The largest cluster (Cluster 2. Structural biosecurity should be given priority 6) with 38 farms was in a location (Waller Field) as it would facilitate operational biosecurity. Conference Proceedings 69

3. Construction of poultry farms should be Duvauchelle, A., A. Huneau-Salaün, L. Balaine, N. subjected to necessary planning approval Rose and V. Michel (2013). “Risk factors for as it relates to spatial selection and physical the introduction of avian influenza virus in conformation. breeder duck flocks during the first 24 weeks of laying”. Avian Pathol. 2013;42(5):447-56. 4. There should be standard operating procedures for poultry farms in Trinidad. Ekwue, E.I. M. Gray and A. Browne (2003). “Poultry Farm Buildings in Trinidad: Present and Future 5. The more vulnerable clusters should be afforded Prospects”. West Indian Journal of Engineering more surveillance using detailed risk profile Volume 25 No 2 1-17. tools. Updated knowledge of the individual poultry farms will facilitate effective response Fasina,F,O.,A.L. Rivas,S.P.R.Bisschop,A.J. Stegeman action such as bio-exclusion, biocontainment and J.A. Hernandez (2011). “Identification of and depopulation in event of an avian influenza risk factors associated with highly pathogenic outbreak. avian influenza H5N1 virus infection in poultry farms, in Nigeria during the epidemic of 2006 References 2007”. Prev Vet Med. 98:204–208. Aengwanich, W., T. Boonsorn and P. Srikot (2004). Hosein, A. and K.F. Pun (2004). “Sustaining “Intervention to Improve Biosecurity System Performance Improvement in the Poultry of Poultry Production Clusters (PPCs) in Industry in Trinidad and Tobago: A Decision Thailand”. Agriculture 2014, 4, 231-238. Support Paradigm”. West Indian Journal of Engineering Volume 26 No 2 45-54. Alexander, D.J. (2007). “An overview of the epidemiology of avian influenza”. Vaccine Koch, G. and A.R.W. Elbers (2006). “Outdoor ;25(30):5637-44. ranging of poultry: A major risk factor for the introduction and development of high- Bouwstra, R., R. L. Gonzales and S. de Wit, J. Stahl, pathogenecity avian influenza”. NJAS - Wagen R. A.M. Fouchier and A.R.W. Elbers (2017). J Life Sci. 54:179–194. “Risk for Low Pathogenicity Avian Influenza, the Netherlands”. Emerging Infectious Diseases Madsen, J. M (2012). Epidemiological Analysis Vol. 23, No. 9. of Biosecurity Practices and Associated Prevalence of Diseases in Non-Commercial Brouwer, A., T. Kuiken, and C. Staubach (2018). Poultry Flocks. Thesis. DRUM, University of “Reporting Avian Influenza surveillance”. Maryland Library. In: https://drum.lib.umd. EFSA Journal;16(11):5493 edu/handle/1903/16293. Brown Jordan, A., D. Narang, S.C. Essen, S.M. OIE (2019) Terrestrial Animal Health Code. WORLD Brookes, I.H. Brown and C. Oura (2018). ORGANISATION FOR ANIMAL HEALTH. “Serological Evidence for Influenza A Virus Twenty-eighth edition, 2019. Available Exposure in Wild Birds in Trinidad & Tobago”. in: https://rr-europe.woah.org/wp-content/ Vet Sci. 2018 May 9;5(2):50. uploads/2020/08/oie-terrestrial-code-1_2019_ en.pdf. Burns, T.E., C. Ribble, C. Stephen, L. Toews, J. Osterhold and H. Wheeler (2012). “Use of Pantin-Jackwood, M. J. & S.E. Swayne (2009). observed wild bird activity on poultry farms “Pathogenesis and pathobiology of avian and a literature review to target species as high influenza virus infection in birds”. Rev. sci. priority for avian influenza testing in 2 regions tech. Off. int. Epiz. 28 (1), 113-136. of Canada”. CVJ / VOL 53. Pepin, K. M., Spackmanc, E. & Brownd, J, D (2014). Chowdhury, E. H., P.M. Das, M.R. Islam, R.C. “Using quantitative disease dynamics as a Ancog, C.D.R. Medina, M.G.V. Paraso, R.D. tool for guiding response to avian influenza Domingo and J. Dewulf. (2017). Quantification in poultry in the United States of America”. of biosecurity status in commercial poultry Preventive Veterinary Medicine 113 (2014) farms using a scoring system. Poultry Industry 376–397. Collaborative Newsletter. Available in: https://en.engormix.com/poultry-industry/ Reperant, L.A., G.F. Rimmelzwaan and T. Kuiken articles/quantification-biosecurity-status- (2009). “Avian influenza viruses in mammals”. commercial-t41111.htm Rev Sci Tech. 28:137–159. 70 Conference Proceedings

Root, J., S.A. Shriner, J.W. Ellis, K.K. Van Dalen, H.J. Souris, M., D. Selenic & S. Khaklang (2014). “Poultry Sullivan and A.B. Franklin (2015). “When Farm Vulnerability and Risk of Avian Influenza fur and feather occur together: Interclass Re-Emergence in Thailand”. Int. J. Environ. transmission of avian influenza A virus from Res. Public Health 2014, 11, 934-951; mammals to birds through common resources”. doi:10.3390/ijerph110100934 Sci Rep. 5:4354. Ssematimba, A., T.J. Hagenaars and M.C.M. de Jong Sims, L. D., T.M. Ellis, K.K. Liu, K. Dyrting, H. Wong, (2012). “Modelling the wind-borne spread M. Peiris, Y. Guan and K. F. Shortridge (2003). of highly pathogenic avian influenza virus “Avian influenza in Hong Kong 1997–2002”. between farms”. PLoS One. 7:e31114. Avian Dis., 47: 832–8. Singh, M., J.A. Toribio, A.B. Scott, P. Groves, B. Barnes, K. Glass, B. Moloney, A. Black and M. Hernandez-Jover (2018). “Assessing the probability of introduction and spread of avian influenza (AI) virus in commercial Australian poultry operations using an expert opinion elicitation”. PLOS ONE. https://doi. org/10.1371/journal.pone.0193730. Conference Proceedings 71

Effectiveness of a Hazard Analysis Critical Control Point (HACCP) plan and checklist to control food safety and quality at a poultry plant in Trinidad Lizanda Gookool, Neela Badrie, Marsha Singh Department of Food Production, Faculty of Food and Agriculture, The University of the West Indies, St. Augustine, Republic of Trinidad and Tobago Email: [email protected] Microbial food safety is increasing worldwide, and many institutions, governments and other agencies have been developing techniques to control food safety and food borne diseases in the poultry industry. Checklist is a powerful tool in ensuing food quality and standards are being met by a local body while HACCP is designed to identify, establish, eliminate or prevent hazards. This study was done at a local poultry processing plant determining the effectiveness of food safety and quality through a HACCP analysis. A decision matrix was used to identify hazards at critical control points (CCP’s) through physical, biological and chemical. The HACCP control chart was critical for cross contamination and ensured that monitoring procedures for hazards such as antibiotic residue, pathogens, physical contaminants and microbial growth had corrective measures together with proper records and verification. It also showed critical limits and the monitoring procedures during production. A local and international checklist comparison was done using the pyramid of Prerequisite Programmes (PRPs). Recommendations were made to have the plant certified as only employees are HACCP certified as well as sharing information to the public on food borne illnesses and food safety. The findings suggested that the checklist and the HACCP system are effective in controlling food safety and quality assurance in meeting with international standards. Keywords: HACCP, prerequisite programme, critical control points, hazard analysis, good manufacturing practices, food safety, chicken, foodborne, microbial pathogens, food security INTRODUCTION issue due to continuous outbreaks around the world and mechanisms are being developed each day to Chicken (Gallus gallus domesticus) is a universal food eliminate these pathogens. With the recent COVID of our era that crosses multiple cultural boundaries -19 pandemic, food safety and security were deemed and is one of the most demanded meats in the world. essential as many people were pushed into poverty In recent years, poultry meat consumption per capita in Trinidad and Tobago (T&T) (Chapman 2021). reached 58.3 kilograms in Trinidad (Williams et al. Irrespective to the surveillance systems on food-borne 2020). Millions of people in the world, fall ill after diseases in the Caribbean region, there has been poor consuming any sort of contaminated food. Of these, documentation of bacterial foodborne zoonoses in food- 420,000 people die annually, including 125,000 producing animals and their associated epidemiologic children less than 5 years of age (WHO 2020). significance (Guerra et al. 2016). Foodborne illness has become an economic burden in developing countries and due to COVID-19, it has Food Safety and Quality Assurance checklists (QA) exposed the fragility of food systems in Trinidad and are a proven mechanism to ensure quality and safe Tobago, making it a main priority. X Company Limited products in the food chain. In international studies has followed guidelines of HACCP to ensure safety of QA checklists seem to be the benchmark model for their product. business performance, bird welfare, health and safety and quality standards (Manning 2007). In 2000, food safety was deemed a prominent global 72 Conference Proceedings

An outbreak in Trinidad showed that 20.5% of chicken affect raw material, and/or affect processing or cross carcasses from cottage poultry processors were contaminate the facility. The decision matrix process positive for salmonella; the prevalence was 8.3% for steps also looked at severity of risk, controlling of risk, chicken being sold in supermarkets. (Khan et al. 2018). prevention and safety for consuming as it monitored World Food Programme (WFP 2022) in the Caribbean procedures and established corrective actions. The is looking beyond T&T for food security during this HACCP controlled chart showed potential hazards pandemic as T&T is an important trade partner in which may occur during production and established agriculture and in this case, poultry exported from X a monitoring system for each CCP, corrective actions, Company. verification procedures and documentation and record keeping. X Company was chosen for low risk associated with poultry and the consumer as well as being a RESULTS part of the Trinidad and Tobago Poultry Association and The Caribbean Poultry Association. They X Company was observed as a protocol-based produce over 35,000 broilers a week – that is more poultry plant where, proper signage, instruction and than 18 million chickens a year and has surpassed adherence were followed with adequate policies and international standards on poultry production. They are documentation. The location of the plant was in good technologically advanced and their chicks are reared in proximity to roads, transport, wind direction and a bio secure environment ensuring safe poultry meat to utilities. The area is well suited for future expansion consumers (Ramsubeik 2018). and had readily available spacing with preventative measures in case of natural disasters. This study aims to evaluate the effectiveness of a food safety using HACCP while comparing an international Prerequisite Programme (PRP) and a local food safety and quality checklist. It will then be determined if the plant meets international The PRP was implemented in accordance with the standards and conformances through Prerequisite GMP and Codex principle to establish a condition that Programmes (PRPs), Critical Control Points (CCPs) is suitable for the production and handling of food and risks that may be associated with raw material. As safety at all phases of the food chain (NBHE 2006). a study suggested the implementation of food safety The level of conformance for each PRP required by management systems such as HACCP and Good FSMS was in conformance (Annex 1). Manufacturing Practices (GMPs) are important to reduce the risk of COVID-19 infection in the industry. Critical Control Point (CCP) determination Cleaning, sanitation, good hygienic practices, and active packaging are also needed from farm to fork for The CCP is a critical control measure which was used food safety (Olaimat et al. 2020). to prevent or eliminate a hazard. If the answer to the below questions is yes you cannot proceed and if it is a MATERIALS AND METHODS no you can proceed to finish the product. A developed food safety and quality checklist was Q1. Is there a hazard associated with the raw compared using an international checklist and a local material? checklist of the Ministry of Health (MOH). This was done qualitatively for observations and records as it Q2. Is this hazard going to be exposed to any form of focused directly on PRPs. These are, location, premise, further processing? equipment, water supply, pests, waste management, maintenance / cleaning , storage / transportation, Q3. Is there a cross contamination risk to the facility traceability, sanitation personal hygiene and training. or to the other products which will not be controlled? HACCP principles were applied by conducting hazard analysis through a Decision Matrix which determined The hazard identification decision matrix chart using the critical points as well as CCPs in the production and CCPs determination processing of the carcasses. It showed risk associated chemical, biological and physical hazards that would Hazard identification is defined as the physical, biological and chemical properties that can make a product unsafe for production (Annex 2). Conference Proceedings 73

Decision Matrix for process steps (raw material) in the development of applying the HACCP principles to the food operator in eliminating food borne The decision matrix was used to identify the critical pathogens (FAO and WHO 2020). E.coli, salmonella points of the raw material process in processing phases. and campylobacter were the biggest hazards which It also gives an organised list of the hazards with the may have occurred in the production plant. below questions. PRPs Q1. Does the step involve a hazard sufficient risk and severity warrant to control? If yes proceed to Q2, if no, The PRPs were in accordance with GMP and Codex not a CCP principles to establish a condition that is suitable for the production and handling safety of food at all phases Q2.1 Does a preventative measure for the hazard exist of the food chain. They are used by auditors around the at this step? If yes proceed to Q3 if no proceed to Q2.2 world as an integrated part of the food management system. The checklist was effective in this study Q2.2 Is control at this step necessary for safety? If yes, showing compliance of X Company and the level of modify the step, process or product, if no, not a CCP conformance for each PRP required by Food Safety Modernization Act (FSMA) was in conformance as Q3. Is control at this step necessary to prevent or seen in Annex 1. Comparison of both the international eliminate/reduce the risk of hazards to safety level to checklist and the local checklist by the MOH gathered consumers? If yes, CCP, if no, not a CCP the PRPs below. The Decision Matrix process in the poultry plant The hazard identification is defined as the physical, Location biological and chemical properties that can make a product unsafe for production and cause a threat to Town and Country approval was granted for the location human health. It may contaminate the carcass and of the plant being away from industrial activities. what it produces. Physical hazards such as metals, The T & T Fire Service also granted permission as bones, dust or any foreign matter as seen in Annex 3. the plant was away from disasters and fires. Readily Biological hazards are any bacteria, microorganism, or accessible roads for transportation prevented hazards, virus. In this case with poultry, it strongly monitored breakouts and production cost. The area had none/ E.coli, Salmonella, Campylobacter and other limited pollution, odours and pests making the plant pathogens that would impact on the carcass and the less susceptible to pest and disease. plant. Chemical hazards are pest control substances, chemicals, cleaning agents, food additives, water and Premise/room biocides. In this case, it majorly focused on boiling water, ice and chlorine residue. Pathogens can multiply due to varying temperature if stored in rooms on the premises. In this case, due HACCP Control Chart to the recent pandemic, food service operators were among the first workers in frontline employment The HACCP Control Chart shows potential hazards sectors experiencing the impact of the COVID-19 which may have occurred during the production. It and such can transfer in the workplace making food uses components of the HACCP principle which are security a priority (Olaimat et al. 2020). Grease traps the critical limits, monitoring and corrective action. and footbaths at the entrance of the premises reduces The CCP’s include reception, final rinse, boiling water, the potential spread of pathogens and organisms which packaging, storage and distribution (Annex 4). may be in the atmosphere and which can survive for several days or weeks (University of Maryland DISCUSSION Extension 2021). Some of these includes E.coli, salmonella and campylobacter as well as the recent Food safety initiatives are constantly changing over acute respiratory syndrome corona virus (SARS-CoV-2) time and food manufacturers are keeping up to date with that can survive on food surfaces (Olaimat et al. 2020). trends in order to be on the national markets (Njunina Proper ventilation and fittings were intact. 2022). The HACCP system showed food safety and quality met international standards at X Company. The Equipment National Codex Committee has actively been involved Equipment was free from rust and stains on stainless steel. Stains can cause physical hazards if they are in a 74 Conference Proceedings

corrosive state, together with foreign matter. Vibrations to remove waste. When following the guidelines of can also cause injury to workers from processing, thus WHO, this can ensure bacterium cannot get into plant proper training and collaboration to use the equipment due to properly placed disposals of waste. is given to staff (Occupational Safety and Health Administration 2016). Sanitation Standard Operating Sanitation system Procedures (SSOPs) ensured that the equipment was well maintained and sanitised in the plant to prevent The requirements of the sanitation system ensure that adulteration. all sanitation and cleaning at the plant are in compliance with the MOH standards of non-corrosive chemicals Water supply being used to sanitise. Proper monitoring and recording are done to foresee any possible adulteration or cross The water supply is consistent with proper checks contamination that may occur and prevent them being done and tests are done by the local authorised from occurring. A comprehensive in-house sanitation body, the Water & Sewage Authority (WASA) every 6 programme is documented in X Company and is months. Pathogens can cause growth when they fall in rigorously implemented. temperature zones with water, that is a range between 41 and 135°F. Boiling water was fixed at 71° - 100°C Personal hygiene at the plant and literatures also show this as being the right temperature for boiling. Colour coded lines were Personal hygiene is very important as lack of it can used to ensure no cross contamination with potable spread microbial contaminants in the food plant and chlorinated water. Color-coding is important in with bacteria found on the person (USDA 2019). preventing cross-contamination due to pathogens, Campylobacter and salmonella are the most recorded allergens and foreign contaminates. It was also food safety hazards that can spread through improper developed during HACCP implementation (PRWEB hygiene. Proper signs, a footbath, hair nets and gloves 2017). Proper documentation is readily available for were used so that workers cannot exhibit behaviour auditors and systems can be monitored in the event that can result in contamination such as chewing, of contamination. Water potability ensured that X eating, drinking, smoking and coughing. Employee Company had an approved water supply as defined by health conditions and practices are vital to prevent the Food and Drug Administration (FDA) and United cross contamination of food through human contact States Department of Agriculture (USDA 2012). with surfaces and packaging material (FAO and WHO 2020). The SSOPs ensured that personal hygiene was Maintenance and cleaning monitored to ensure cleanliness from employees in preventing any cross contamination from human to Maintenance and cleaning are routinely done to ensure meat and this was found in the checklists (FAO 2006). product safety and quality at all stages from start to finish of the product. There are no possible ingress, Storage and transportation holes or spaces for the entry of pests and the building is properly inspected by the MOH, as well as a private Transport vehicles were suitably constructed with owned company with certified technicians. temperature at -18°C for frozen chicken and 4°C or less for chilled products. Air circulation and temperature Pest control is very important in the plant as temperature can increase microbial growth in the freezer, transport Proper records are available and there are hired vehicles and storage facilities. Listeria can occur in technicians responsible to ensure no insects or rodents the boiling stages (McLauchin et al. 2004) and can be can enter. There are no possible ingresses, holes carried in the product even if cooked by the consumer. or spaces for any entry and the building is properly Thermometer testing was in place to calibrate the right inspected by the MOH. temperatures to reduce bacterial contamination such as Staphylococcus aureus, Salmonella enteritidis, Waste management Escherichia coli and campylobacter which can grow to dangerous levels and cause illness (USDA 2012). A biosecurity system is used that will not affect the production of the plant. Waste management system Traceability was dispatched by Rentokil and a private farmer. The Rentokil technicians are also certified for removal of Traceability was directed to the international checklist harbourage and waste as well as employees trained but was not limited to the local checklist. Records Conference Proceedings 75

are kept for up to 3 years. Traceability allows food farm auditors. Chlorinated water was used at proper businesses to target the product(s) affected by a food temperatures to remove bacteria that may cause safety problem, minimising disruption to trade and any chemical reactions. Because of chlorine, ice was both a potential public health risks (Food Standard Gov. 2017). chemical and biological hazard but was not a CCP. The It ensures transparency and accountability throughout water was approved by the governing body to make the supply chain to meet with regulations and in the the ice. The last material was the packaging material case with X Company, international regulations. for CCP. The hazard types were foreign matter, toxic Training ingredients and inks on packages that fell under chemical and physical hazards. CCP preventative Training was documented and showed to all staff in measures were packaging approved by specific bodies, any case of emergency. There was current evidence physical inspection done and inks used that are food for HACCP training, proper food handling technique, degradable and cannot affect the meat. The critical food protection and GMP which support sanitary limits were based on preventive measures by storing at practices that may cause any form of contamination appropriate temperature 0 - 4 C֯ , water and ice testing, to the employee. According to the Canadian Institute monitoring and proper packaging material to reduce for Food Safety, 2020, training aids in reducing toxicity. The decision matrix ensured that the risks food poisoning and food allergies with legal laws in were reduced and focused on for proper monitoring and territories to operate establishments with training. This record keeping to reduce or eliminate these hazards. helps to prevent future damages or any sort of risk that Decision Matrix process can be prevented. Both the international and local checklists, with the The Decision Matrix (Annex 3) was used to identify aid of HACCP, to create PRP’s are essential for proper the CP for the raw material processing in processing documentation and record keeping especially in case of phases and gives an organised list of the hazards. a recall. The FDA together with the FSMA understands The raw materials hazard types are pathogens which how essential checklists are in ensuring food safety. are E. coli, salmonella that fell under chemical and The checklists also gave a measure of the strengths and biological. Annex 3 found that antibiotics and foreign weakness at the plant. particles can cause physical and chemical hazards in the raw material. With proper record keeping, Hazard identification using the CCP determination disinfection and sanitation, risks can be reduced or and Decision Matrix eliminated for this method reducing pathogenic growth Hazard identification (Annex 2) is defined as the in the raw meat and processing plant. Ante-mortem physical, biological and chemical properties that can inspection has potential hazards such as pest/rodents make a product unsafe for production. The Decision which is a biological hazard as well as Salmonella, Matrix shows potential hazards in raw material, E.coli and bleeding carcass. Dust and foreign matter processing water, ice and packaging which can however can create physical hazard. These were harmfully affect the processing of the poultry meat and monitored through PRP to prevent these hazards cause pathogenic growth in the plant, as well as cause a while applying the guidelines of HACCP principles. threat to human health. Raw material was identified as Metal detectors were used to locate any foreign matter a biological hazard through pathogens and antibiotics. while proper bleeding techniques were used to prevent The preventative measures used were storing the meats biological hazards. Scalding, cutting/deboning and at the proper chilled temperature as listeria and most de-feathering propose microbiological risk such as E. microbes can be formed in heated areas, ensuring proper coli, salmonella, campylobacter but is prevented with hygienic practices are used so that cross contamination the PRP’s as seen in Annex 1. Temperatures in both can be prevented while handling the meats. Lastly doing vehicles and refrigerators are also controlled by PRP’s facility audits with the daily routine check sheets used as these can contribute to microbes growing in vents as for proper traceability methods. Processed water fell well contaminate chillers causing microbial growth and under both chemical and biological hazards because of spoilage. The critical limits were done by observations the chlorine residue which can cause harmful effects to from the QA. The decision matrix addressed more human health but not a CCP. However, a preventative than one hazard and gave comprehensive measures measure used were having the water tested and records and CCPs (Njunina 2021). Proper standardisation made by government agencies and other of the local encouraged proper procedures to be practiced by 76 Conference Proceedings

using HACCP principles for proper monitoring and assurance in the processing plant as it evaluated, evaluation in these steps. monitored, recorded and verified methods that helped to Identify risk in the raw meat, process water, ice, The HACCP control chart packaging material and areas in the processing of the chicken from receiving to the end product. Measures The HACCP control chart (Annex 4) integrated used reduced risk and helped with conformity, non- components of HACCP principles such as corrective conformity and standardisation of the HACCP method. measure, critical limits and monitoring for all listed hazards. CCP 1 identified at the reception of chicken looked CONCLUSION at the residues and antibiotics. This was checked at every batch, proper recording, monitoring and lab According to the study on X Company, food safety and testing done to prevent the hazard from occurring. quality assurance can be controlled through the use For monitoring procedure, a batch compliance must of a checklist and the HACCP plan. By the company be signed with every batch and lab checks sent for adhering to third party and local auditing, the checklist verification. This method assures product is safe (Annex 1) showed risk that future risk can be prevented. from reception. CCP 2 was survival of pathogens in Proper documentation and verification were deemed boiling water. Temperature logs and testing was used essential with this plant and were readily available for monitoring to prevent the hazard from occurring. with proper record keeping. The PRP’s analysed the As stated before, verification and monitoring are very plant and showed where microbes and any form of effective at this stage as water is a vehicle for infection pathogenic contamination could occur and also showed in poultry. When these waters are less than 71°C the quality of the interior and exterior of the plant and pathogen growth can arise. Again, proper recording environs. Proper training and traceability were done and and temperature testing is a must. CP 1, the final rinse were successful in preventing any risks. The decision ensures that heat treating here is the final opportunity matrix (Annex 2) for process steps showed the risk that for reducing microbial contamination or any form of might have occurred during the processing of the raw pathogen that would/may be present on the surface or carcass and production stage that can create pathogenic the product that can contaminate the plant. The recent growth and risk contamination. It was monitored and SARS-CoV-2 is known to remain on surfaces as well verified by procedures put in place by QA. The Hazard as the listeria in water. CCP 3 is packaging where Identification Decision Matrix (Annex 3) showed risks the potential hazards had been identified as physical in the plant and how they can be reduced or eliminated contamination in the packaging of the poultry. The limits by following the HACCP system even though the plant used was to ensure no foreign matters exist and proper was not HACCP certified. recording and inspection is done to prevent this from The HACCP control chart was designed for the poultry occurring. Residue limits can be visually monitored by plant based on principles of HACCP (Annex 4). the QA, as well as recording for any metal detection Harmful potential risks which may have been present to help limit any physical contamination to get inside were verified and monitored through the CCP chart. of the product. CCP 4 looks at storage/distribution as The findings suggest that the local poultry plant passed pathogens tend to grow above 4°C. Proper temperature international standards, with proper conformance and checking, testing and calibration were done to prevent compliance through PRP’s, using checklists, together this hazard from occurring. Preventative measures used with the HACCP system. Despite the loss of 300 billion for all CCP and CP were to withhold carcass, await the dollars and more during the pandemic on the poultry vet, test temperatures and record. Corrective measures industry, mechanism is being developed daily to boost used for all CCP and CP were to revoke batches, re- production by spending over 270 million for poultry test, monitor, discard raw material that may have bad and livestock farmers in the USA by having them apply temperature reading, discard foreign matters and call for a Pandemic Assistant grant. (USDA 2022). In T&T, technician when needed. These mechanisms were used the government aids in building food security and to prevent hazards from occurring. When relating to boosting production. This will impact on the poultry conformances, all sections were in conformance with sector through an Agriculture Stimulus Package and the FSMA audit checklist and none were critical. The strengthen links between industries and agriculture food safety and quality checklist using a HACCP through Ministry of Trade and Industry (MTI), the system was effective for food safety and quality Conference Proceedings 77

Trinidad and Tobago Manufacturers’ Association Food Standard Gov. 2017. “Food Traceability.” Food (TTMA), the Ministry of Agriculture, Land and Standard & Safety Gov Austria/New Zealand, Fisheries (MALF) and the National Agricultural and January 2017. https://www.foodstandards.gov. Marketing Development Corporation (NAMDEVCO) au/industry/safetystandards/traceability/pages/ (GovTT 2022). It was recommended to have the plant default.aspx. HACCP certified and become the first poultry plant to GovTT. 2022. “Government Proactive in Mitigating have done so in T&T. Another recommendation was Impact of Rising Global Food Prices.” The for X Company Ltd to start sharing information to the Ministry of Trade & Industry| Trinidad and public for awareness on foodborne illness and food Tobago Government News, March 30, 2022. safety which will aid in consumer awareness. http://www.news.gov.tt/content/government- proactive-mitigating-impact-rising-global-food- ACKNOWLEDGEMENTS prices#.Y0rX3HbMLIU I am very grateful to the Quality Manager of X Company Guerra, M.M, Almeida, A.M. and Willingham, A.L. and to Mr S. Rahaman for assisting in the compilation 2016. “An overview of food safety and bacterial of Annex 1. Information in Annex 1 was also obtained foodborne zoonoses in food production animals from the Nestle Assessment Checklist “What does in the Caribbean region.” Tropical Animal Nestle expect from its farmers”, and from the HACCP Health Prod (2016) 48:1095–1108. Food Safety Audit Checklist. I also acknowledge the Khan A, Georges K, Rahaman S, Abdela W and information provided by Mr S. Ramsubeik during two Adesiyun A. A. 2018. “Prevalence and serotypes interviews with him in November 2018 and January of Salmonella spp. on chickens sold at retail 2019. outlets in Trinidad.” Europe PMC. DOI: 10.1371/ journal.pone.0202108. PMID: 30138324; REFERENCES PMCID: PMC6107152. Manning Louise.2007. “Food Safety & Brand Canadian Institute of Food Safety. 2020. “Why Food Equity” British Food Journal: LJM Associates Safety Culture Matters.” Canadian Institute of Limited. https://www.researchgate.net/profile/ Food Safety. June 10. https://www.foodsafety. Louise-Manning/publication/228533410_ ca/blog/why-food-safety-culture-matters. Food_safety_and_brand_equity/ links/0046352e9941befa7d000000/Food- Chapman. R. 2021. “Protecting Food Security in safety-and-brand-equity.pdf T& T.” https://Trinidadexpress.com/Opinion/ C o l u m n i s t s / P r o t e c t i n g - F o o d - S e c u r i t y - McLauchin J, Mitchell R T, Smerdon W J and Jewell K. in-t-t/article_98477582-2c81-11ec-b5a9- 2004. “Listeria monocytogenes and listeriosis: 03e3e5936c00.Html.” Protecting Food Security a review of hazard characterization for use in in T&T, October 13, 2021. microbiological risk assessment of foods.” FAO. 2006. “FAO and WHO guidance to governments International Journal of Food Microbiology, on the application of HACCP in small and/or 92: 15–33 less- developed food businesses.” FAO Food Njunina Violetta. N.V. 2021. “What Is the Difference & Nutrition Paper 86. https://www.fao.org/3/ between the HACCP Decision Tree and Hazard a0799e/a0799e00.pdf Matrix.” The HACCP Decision Tree and Hazard Matrix. November 3, 2021. FAO and WHO 2020. CODEX. “What’s next for Njunina Violetta. 2022. “Food safety trends: How standards? Private Sector looks post Covid- has the food industry changed?” https://www. 19- Safe handling food practices as important fooddocs.com/post/food-safety-trends as ever. https://www.fao.org/documents/card/ en/c/cb1502en Occupational Safety and Health Administration (OSHA) 2016. “Recommendation Practices for Safety & Health for Construction Workers.” OSHA 3886. https://www.osha.gov/sites/ default/files/OSHA3886.pdf 78 Conference Proceedings

Olaimat A N, Shahbaz H M, Fatima N, Munir S and media/press-releases/2019/08/20/washing- Holley R A. 2020. “Food Safety During and raw-poultry-our-science-your-choice After the Era of COVID-19 Pandemic” National USDA. 2022. “Food Safety and Inspection Service.” Library of Medicine, Frontiers in Microbiology. askFSIS | Food Safety and Inspection Service. https://doi.org/10.3389/fmicb.2020.01854. U.S. Department of Agriculture. Accessed March 3, 2022. https://www.fsis.usda.gov/ PRWEB.2017. “Food industry supplier sees growth contact-us/askfsis in key customer service positions.” REMCO Williams K, Mohammed E, Maharaj D, Mohammed T Corporation colour coded tool provider, and Mohammed A. 2020. “Total and Estimated adds four new positions. REMCO Products. Inorganic Arsenic Levels in Imported and Local Accessed October 2022. https://www.prweb. Chickens Sold in Trinidad and Tobago. Journal com/releases/2017/05/prweb14325574.htm of Food Production. Vol 83. No.4. 2019. Pg 710-714 University of Maryland Extension. 2021. “Footbaths”. WFP. 2022. “The Caribbean: World Food Programme.” https://extension.umd.edu/resource/footbaths. UN World Food Programme. Accessed March 1, 2022. https://www.wfp.org/countries/ USDA. 2012. “How temperature affects food?” caribbean. Accessed June 29th, 2019. https://www.fsis. WHO. 2020. “Food Safety.”World Health Organization. usda.gov/wps/portal/fsis/topics/food-safety- April 30, 2020. https://www.who.int/news- education/get-answers/food-safety-fact-sheets/ room/fact-sheets/detail/food-safety safe-food-handling/how-temperatures-affect- food/ct_index USDA.2019. “Washing raw poultry: Our Science, Your Choice.” Food Safety Education Staff Release No. 0125.19. https://www.usda.gov/ Conference Proceedings 79

Annex 1 Table 1. Comparison between local and international checklist using PRPs PRP’s International checklist Local checklist Location The plant was situated away from environmental The plant was approved by Town and Country Planning pollution, flooding and industrial activities and were Division of T & T, T & T Fire Service and MOH. It was located not prone to any infestation of pest. away from sources of strong offensive odours, smoke, dust and other air contaminants. Areas were well drained with readily accessible portable water supply. Premise/room The floor was flat for proper flow of water. There was Roads, yards and parking lots were free from excessive dust. readily available water with and without hoses, both in The plant was free from improperly stored equipment, waste, house and on walls. Proper fixtures and openings were uncut grass and litter. There were well maintained, and adequate in place to clean rooms with corners and junctions of drains, with areas to cope with maximum load. Provisions were the wall being round to ensure sanitary cleaning occurs made to exclude pest, dirt, and filth. There was a functioning and no grease is stock. All doors and fittings were made insect and rodent control programme. Walls, ceilings, floors of stainless steels. Premise was well ventilated. were in a sanitary manner. Grease traps and footbaths were installed where necessary. Equipment Equipment was made from stainless steel suitable for The design and location of equipment, containers and utensils Water supply a food plant and was easy to disassemble, clean and are such that they provide protection from contaminants and are disinfect to avoid contamination. Equipment was accessible for cleaning, sanitising, maintenance and inspection. Maintenance and routinely checked by QA in charge to ensure smooth Equipment was constructed from suitable, smooth, impervious, cleaning running and no cross contamination from occurring. nontoxic, corrosion-resistant material. Pest control Chlorination existed in house as raw poultry was Potability certificate was current and available for well water passed with the hangers. Ice was made from potable up until 6 months with weekly checks being done. There was water and stored to protect contamination. The plant proper availability of ice and ample supply of hot and cold has a water treatment system to provide potable water potable water under adequate pressures at all times during with test being done every six months to meet WASA working hours. There was ease of water through the plants and standards and the Chemistry Food & Drugs Division of lines are colour coded so that supply was protected against cross the Ministry of Health contamination or any contamination. Residual chlorination was maintained at an acceptable level for intended use. Daily checks were done by the QA. Product contact surfaces/ equipment, containers and utensils thoroughly cleaned and sanitized after use. Cleaning methods preclude contamination/ adulteration using approved chemicals. Sanitary control of raw meat sufficient to protect the product. All test and results were available for inspectors. Rooms and other areas used for receiving, processing and storing of raw materials and finished product are painted in a clean sanitary manner. A comprehensive in plant- sanitation program documented and rigorously implemented. There were programs to minimise the entry of rodent, Was not included insects and birds in the farm. The technicians were trained and certified and information readily available. Manual devices were located to trap insects. The pest control chemicals are approved for food processing areas and are stored correctly. From observation, there was no possible ingress such as in processing area, holes in walls/cracks or any evidence of infestation for rodents/pest to enter. There was a regular removal of waste and rubbish to prevent harbourage. The plant has a biosecurity programme that was monitored and had laboratory records which were sent to international auditors and Chemistry Food and Drugs Division of MOH. 80 Conference Proceedings

Waste management Waste system are disposed by Rentokil who are certified Liquid wastes were disposed in an acceptable manner. Product in harbourage wastage. The biosecurity does not affect waste and dry waste was collected in suitable containers located the production of the plant. suitably in the plant. They were then disposed of at frequent intervals in a sanitary manner. Covers were kept over while not in use. Sanitation system Sanitation and cleaning requirements used were All equipment, walls, floors, ceilings or any machinery used documented together with daily sanitation sheets. were sanitised after use and cleaning methods preclude Personal hygiene Records were available to show compliance. The contamination/ adulteration. Sanitary control of raw materials chemicals that were used are certified with food (meat and non-meat) was sufficient to protect the product. Storage and decreases. There was no evidence of dirt and debris in Sanitary control of finished product was sufficient to protect transportation production areas. product while all test results and made available to inspector/ auditors. Rooms and other areas used for receiving, processing Traceability and storing of raw materials and finished product are maintained Training in a clean, sanitary manner. There is a well comprehensive in- plant sanitation program documented and always rigorously implemented. Hygiene and security policies was in place and must Persons in direct contact with food ingredients wore clean be adhered to by workers/non workers. Any sort of garment, very hygienic, wash hands before starting work and jewellery, nail polish, false finger nails, smoking, in case of any accidents, ensure gloves were used in handling makeup eating or drinking are not accepted. Gloves, food and wore effective hair nets/restraints. They did not exhibit boots, hair nets, clothing for chilled/freezer areas any behaviour that may result in contaminating the product are provided for staff. Hand washing facilities had a such as eating, chewing, spitting, sneezing, smoking, coughing, knee push button to avoid contamination inside of the jewellery, bare hands and makeup. plant. Antibacterial soap and running water are readily available. Personal movement was restricted from raw to processed food areas where applicable to avoid cross contamination. Equipment was stored separately for dry and cold There was adequate provision for storing raw materials, materials. Vehicles well ventilated with correct packaging materials and finished products. Storage methods temperatures. used minimise deterioration and prevented contamination as arrangements precluded contamination. The cold storage area is adequate to maintain raw finished products with separate freezes for raw and finished products. The ice maker’s room was separate with proper air circulation because of stacking method. A thermometer was present with product cycling system in place. Transport vehicles were suitably constructed, properly maintained, cleaned and packed to avoid deterioration or contamination, maintaining 4°c or less for chilled products and 18°c less for frozen products. Records were maintained for a defined sufficient period of 3 year. All materials received was inspected. There were documents incase of a recall or to withdraw a product. Allergen identification and control were available for feed, chemicals (if any) and finished products. There were documents for evidence of training. Training Supervisors and production floor workers received appropriate was provided to all staff and in case of emergency with training in proper food handling techniques and food protection location, the use of the emergency stops, sufficient principles. They were aware of the dangers of poor personal skills and experience as well as training to safe guard hygiene and unsanitary practices with other sources of the welfare of chicks in the farmers care. contamination. Conference Proceedings 81

Annex 2 Table 2. The Hazard Identification Decision Matrix chart using CCP’s determination Material Hazard type Hazard Q1 Q2 Q3 CCP Preventative measure Raw Pathogens, B YY N CCP Stored at chilled temperature 0-4 c֯ . material Antibiotics NN Proper hygienic practices and handling and facility (meat) audit check sheet used Process Chlorine C, B YN water residue YY N CCP Potable water is used and tested. It is recorded by government and other audit agencies. Chlorinated Ice Chlorine C, B water (50 ppm) is used at proper temperature to remove bacteria. Water testing samples to be used Packaging Foreign C, P material matter, toxic N The portable water is used to make the ice and is ingredients, approved by the above agencies. Inks N CCP Packaging approved by QA and physical inspection done by other bodies. Inks are food gradable B-Biological; C-Chemical; P- Physical; Y-Yes; N-No; CCP- Critical Control Points 82 Conference Proceedings

Annex 3 Table 3 illustrates the Decision Matrix process results Process step Hazard type Hazard Q1 Q2 Q2.1 Q3 CCP REMARKS Raw Material Pathogens, (E.coli) C YY Y YES Inspection is done by company’s vet with a checklist of clearance. Antibiotics, PN Records kept from farmers who are Foreign particles C YY N trained by X Company Limited A n t e - m o r t e m Pest/rodents B Y NN Recorded information with proper inspection cleaning and sanitation used. Area is clean and disinfected at least twice a day; Rentokil visits and records Salmonella B Y NY MODIFY Healthy birds collected to be slaughtered Dust/soil/foreign P Y N MODIFY Cleaned and disinfected Em.acttoelri B YN N All workers must ensure they are sick free to enter plant and wear proper gears for employee hygiene. Equipment’s were properly sanitised and proper detergent food decreases used. Bleeding carcass B Y NY MODIFY Proper bleeding techniques used and is inspected by QA with proper records for other checklist/auditors Scalding Smoke P N Proper time management and temperature is monitored and controlled by QA De-feathering E.coli, Salmonella, B YY N Observe employee hygiene, proper Campylobacter temperature and water records; QA record and inspects Head and limb Metal chipping P Using stainless steel knives removal (machine). Proper cleaning was done and recorded Eviscera-tion Pathogens (faecal/ Observations for faecal contamination, Carcass gut- intestinal proper setting of equipment’s and contamination training records kept; Maintenance of (salmonella, E.coli, chlorination by QA for rinse campylobac-ter) P Visual examination for any foreign matter; X Company sheet used to Final rinse P record information. Draining P N Observation for foreign matters as it cannot go beyond this point for any recalls N Proper drainage and inspection of carcass Conference Proceedings 83

Cutting/ deboning Pathogen, B YY N NO Proper temperature as growth inhibits Sanitisers, C NY NO bacterial growth Foreign materials P Y NO PRP used PRP such as hygiene and pest control Grinding Pathogen, B YY N NO PRP used Sanitisers C N NO Sanitation system used Foreign materials P YY N NO PRP used Chopping Pathogen, B YY N NO Personal hygiene, employee hygiene Sanitisers C N N NO PRP used Foreign materials P YY Observation and metal detector NO Forming Pathogen, B YY N NO Sanitation done Foreign material P YY N NO PRP used Boiling Pathogen B YY Y YES Water temperature time was very Y YES crucial and checking of water quality Packaging Foreign material P YY NO Presence of foreign matter is very Pathogen BN Y YES crucial Distribu-tion Pathogen B YY Proper packaging and food handling practices Storage temperature and time is very important. The delivery vehicle temperature is also important and suited for products. 84 Conference Proceedings

Table 4 exposes the HACCP Control Chart for the local poultry plant Annex 4 CCP # Hazards Critical Monitoring Frequency Preventative Corrective Record Person in Verification 1.Recep-tion Limits procedure measure measure charge CCP 1 Antibiotic Delivery A batch Every Await approval Revoke batch Compli-ance Vet & Veterinary residue must be compliance delivery from the if it’s not in certificate QA testing samples within must be signed veterinarian compliance officer in lab as well as specific time each time by officer MOH official. the veterinarian Accredited official and they labs with must always be other auditors inspected available 2. Final Rinse Pathogens Adjusted Thermo- Every batch Withhold Re test and Record when QA Chlorine, pH CP 1 (E.coli, chlorine level meter used for carcass take samples and temperature Salmonella, of 50 ppm temperature, to ensure re rinsing officer testing with campylobacter) pH using strips updated reports Conference Proceedings and chlorine measure-ments 3. Boiling water Survival of 71 ° c Check Each batch Check the Set equip- Tempera- QA pathogens temperature, temperature ments in order, ture log and officer record, follow up set tempera- maintenance CCP 2 Employee and ture in order, QA factory hygiene, call technician Inspection officer 4. Packaging Physical No foreign Observations, Each pack specification to repair if reports, log Auditors, list of QA inspections, which is agreed nInesepdeecdtion, sheet for QA suppliers, sheet CCP 3 contamina-tion material supplier audits Conheck time and discard foreign detection Officer for maximum 5. Storage temperature. For materials residue limits /distribu-tion Pathogens, <4 ° C, Record storage Routine dispatch, adjust Delivery microbial maximum 3 and shelf life and monitoring temperatures Reject raw report for Calibrate ther- CCP 4 growth days, shelf record and checking to the correct meat and distribution, mometer, check life reading for the products based temperature records and adequate time. on testing, log and batch observe records internal temperatures. CCP – Critical control point, QA/C – Quality Assurance/ Control 85

Wildlife for sustainable food production, an assessment of two neo-tropical species production models Laura Tardieu and Gary W. Garcia The Open Tropical Forage-Animal Production Laboratory [OTF-APL], Department of Food Production [DFP], Faculty of Food and Agriculture [FFA], The University of the West Indies [UWI], St. Augustine, Trinidad and Tobago Email: l_tardieu@hotmail Food derived from animals in the neo-tropics are heavily dependent on the use of a limited number of domesticated livestock species (chickens, sheep, cattle, goats, pigs, turkeys). The steady growth of the population has required that novel approaches and food sources be applied to meet the needs of the populace. The use of wildlife (non-domesticated animals) for food production, is one such approach that has over the years, been overlooked for its value. Presently this type of food production system utilising local, neo-tropical species is practiced on a limited scale in the Caribbean; and yet this system has the potential to improve food and nutrition security in the region, via the provision of a steady source of protein and income, using species that are endemic to the region. Recent attention has been given to animal production of wildlife species in the neo- tropics, with the focus on highly hunted species e.g. peccary, capybara, agouti and opossum. The goal of this paper is to examine the work and research in the Caribbean, that has been conducted toward the development of wildlife production of two species, opossum (Didelphis marsupialis insularis) and agouti (Dasyprocta leporina). It will highlight the recent research that has emerged, including information on the important factors of reproduction, nutrition and behaviour. By studying these important production factors, the feasibility of wildlife production systems for the Caribbean region will be demonstrated. Finally, future areas requiring research will be highlighted to allow for more widespread use of wildlife species as additional and sustainable sources of food for the region. Keywords: Agouti, animal production, Dasyprocta leporina, Didelphis marsupialis, neo-tropics, opossum, wildlife production Introduction Presently this type of food production system is being The majority of animal protein and animal derived practiced on a limited scale in the neo-tropics and has nutrition in the neo-tropics is mainly provided by a proved fruitful for endemic species such as the capybara limited number of domesticated livestock species. (Hydrochoerus hydrochaeris) (Roth 1997, 192), lappe However, population growth and a growing demand for (Agouti paca) (Smythe and Brown de Guanti 1995) animal-based products and wild meat, combined with and collared peccary (Pecari tajacu) (de Albuquerque increased human encroachment on hitherto wild areas, et al. 2010). Despite these successes, the practice of has created unprecedented demand on our wildlife wildlife farming has not become widespread. The populations. Studies report that between 5 - 8 million main setbacks being the limited information and lack people in South America regularly rely upon wild meat of production models for wildlife species. as a source of protein (Rushton et al. 2005). As a result The goal of this review is to examine the innovative of this growing demand more utilitarian/pragmatic research that has been conducted toward the approaches must be utilised to decrease the pressure on development of production systems for two popularly our wild populations to meet the region’s growing food hunted species in the region, the opossum (Didelphis requirements particularly for animal protein and wild marsupialis insularis) and the agouti (Dasyprocta meat. The use of wildlife (non-domesticated animals) leporina). It will further highlight the new research as a resource for food production is one such approach, that has emerged from the Caribbean; and indicate that has been overlooked for its value. 86 Conference Proceedings

future areas of research towards the development of forages at night where it consumes a wide and varied production systems for these species. diet of both plant and animal material (Redford and Production system modelling Eisenberg 1992). To design a production modelling system for a The Caribbean opossum subspecies D. marsupialis wildlife animal, important factors of production need insularis is more familiarly known as the manicou in the to be understood on the chosen model species. These region. It is considered a pest species in certain islands include housing, nutrition, breeding and reproductive and is listed as ‘Vermin’ under the Conservation of information, health and diseases, the behaviour of Wildlife Act, Chapter 67:01 of the Laws of Trinidad and the wild animal and socioeconomic and economic Tobago. Despite its aggressive nature, it is commonly considerations (Garcia 2003). For traditional, hunted and can often be seen foraging and scavenging domesticated livestock species that have been reared for food in both rural and urban settings. It possesses and selectively bred over many years these factors are a fairly widespread habitat distribution within the neo- known and well-studied. However, with neo-tropical tropics, yet this common opossum species is by and wildlife species like the opossum and the agouti, many large understudied in the region. Its North American of these factors are not yet known or have limited counterpart, Didelphis virginiana conversely, has been information available. The research that follows has the focus of many reproductive studies and has been been conducted to supplement or enhance the limited lab reared and studied in captivity since the 1970s information available on these production factors for (Jurgelski 1974, Jurgelski et al. 1974). the Caribbean opossum and the agouti. With the overall end objective being the development of production Taking into account the relatively recent evolutionary models that can be effectively utilised for these species split (within the last 2 million years) of D. virginiana within the region. from D. marsupialis (Tyndale-Biscoe 2005, 116), the assumptions that have been made for the biology and behaviour of D. virginiana, may not hold true for D. Opossum (Didelphis marsupialis insularis) marsupialis species and subspecies (Tyndale-Biscoe and Mackenzie 1976, 249). Particularly for those that The opossum is a New World species belonging to the reside in smaller island ecosystems in the Caribbean order Didelphimorphia, the third largest order of neo- and occupy their own unique ecological niche. tropical South American marsupials (Harder and Fleck 1997, 173). The six recognised species in Didelphis, Opossum production modelling possess a wide geographical range that extends from Canada into South America as far south as Argentina As highlighted above, there are six important factors (Tyndale-Biscoe 2005, 116). required to design a production model for any species. The first factor to have been studied for D. m. insularis The opossum is described as bearing a similar in the region was housing design. Utilising the appearance to a rodent with dark, black eyes and a information on the related north American species naked, almost hairless tail (Nowak 1999, 12). The males of opossum, a production model for the neo-tropical are usually larger than the females with wider heads opossum was theoretically designed by Tardieu and (Cerqueira and Lemos 2000, 321, Krause and Krause Garcia (2018). This study conceptually compared and 2006, 8, Cáceres and Monteiro-Filho 1999, 468). Body contrasted three different production housing designs sizes range amongst species but D. marsupialis can using previous research on animal housing modelling weigh up to 4 - 6k g (Siciliano 2013, Shripat 2011). Like and guidelines, combined with information available all marsupials, the opossum possesses a pouch where on the species. The authors concluded with a housing the female can hold up to nine developing embryonic design that was theoretically best suited for opossum young (Tyndale-Biscoe and Mackenzie 1976, 262). production in the region. The next step for housing the The average litter size is about six (Hunsaker 1977, opossum would be a trialing of this conceptual model 142; Tyndale-Biscoe and Mackenzie 1976, 258) and to determine its feasibility. in the more tropical areas the Didelphis female can Nutrition is key to the production of any species but produce up to three litters a year (Tyndale-Biscoe and given the lack of recent information on the digestive Mackenzie 1976, 262, Hunsaker II 1977, 142; Cabello anatomy of D. m. insularis, a closer look at the 2006, 429). As a nocturnal species the opossum usually gastrointestinal tract was initiated by Tardieu et al. Conference Proceedings 87

(2019b). The results described for the first time, the this model to be truly effective future work will be micro- and macro-digestive anatomy of this species. required to determine additional captive reproductive The findings further corroborated that this species parameters including weaning interval, parasite was capable of consuming an omnivorous diet, as protocols and captive diets for the various physiological it possessed features that supported the ability to states of this species using a larger captive population digest both plant and animal material. Although this (Tardieu and Garcia 2021). study allowed for a greater understanding of the diet and nutrition needs of this species, to produce in Overall, neo-tropical, Caribbean, research on the captivity, greater observation and study will be needed opossum D. m. insularis has made many significant to determine the optimal diet for this species under strides toward the development of a production captive production conditions. model for this species under the tropical conditions of the region. Yet there remains a significant amount The next factor receiving focus by researchers was of research outstanding particularly in the areas of that of reproduction and breeding on the Caribbean health and disease, as many studies have identified the opossum. It was found that reproductive information in opossum being host to a variety of parasites in the wild the region was mostly anecdotal in nature. Researchers, (Jimenez et al. 2011; Oliveira et al. 2010, 53; Horta however, were able to tap into an often overlooked et al. 2007, 796). To reduce the spread of disease and nontraditional resource that in recent times has been improve the health and productivity in captivity, an utilised to provide important information on wildlife effective protocol for removing or reducing the parasite health and biology (Sleeman and Clark 2009). Tardieu load of the opossum will need to be developed. et al. (2019a) utilised one of these resources, in the form of admission records obtained from a local wildlife Greater examination and documentation of their rehabilitation centre located in Trinidad and Tobago, behaviour in captivity will also be required as this to obtain important reproductive and breeding details species possesses an aggressive nature when wild on this species. The admission records when analysed caught. This information will then aide in developing outlined vital breeding statistics for this species i.e. a safe and effective housing design to reduce harm litter size (average 5, range 1 - 8) and breeding seasons to the animals and to the handlers. Finally extended (February - March and August - October) in Trinidad reproductive studies will be able to identify the and Tobago. This study further identified an interesting important reproductive parameters for production of male auditory reproductive behaviour, that could aide this species to increase its use and wide utilisation by with captive reproduction, by assisting with estrous wildlife farmers in the region. detection in female opossums. Agouti (Dasyprocta leporina) The next step toward the development of a production The red rumped agouti (Dasyprocta leporina) is a system involved a trial system based on the conceptual fairly large, long-legged rodent species native of the housing design that was proposed by Tardieu and New World (Emmons and Feer 1997, 223) that can Garcia (2018). The proposed conceptual model was mainly be found within the Guianas, Brazil and the therefore combined with the information on breeding Caribbean (Emmons and Feer 1997, 227). and reproductive behaviour and the knowledge of the omnivorous nature of the species, to develop a pilot Described as weighing 3 - 5.9 kg, with a length of 490 production system. From this pilot system Tardieu – 640 mm, the agouti is often easily differentiated from and Garcia (2021) described for the first time, the other species by its large rump and strongly humped reproductive behaviour and some of the reproductive back. It is most often described as possessing a red parameters of the neo-tropical opossum in captivity. to brilliant orange coloured rump, which is where it derives its common name. However, there is much Over a 10-month period, the pilot study observed and variation in this species coat colour, which has been documented four parturitions with four litters (n = 23) found to range from dull brown olivaceous to bright successfully weaned. The research illustrated that this orange olivaceous to black on its shoulders and neck. species could be produced for production purposes in The agouti’s ears are described as short and mostly the region and resulted in a prototype model system naked with large eyes and a stub like tail (Emmons and that can be utilised for the tropics. Nevertheless, for Feer 1997, 226) 88 Conference Proceedings

The agouti is one of 11 species under the genus A method for collection of sperm via electroejaculation Dasyprocta. However, due to the genus being highly and an examination of the spermatozoa morphology variable geographically in their pelage colour, there soon followed (Mollineau et al. 2008a, Mollineau et has been some taxonomic dispute as to the number of al. 2008b). These studies were able to define standards species (Emmons and Feer 1997, 223). for identifying fertile agouti semen and developed an Diurnal mammals they are often described in the early effective method for electroejaculation of the agouti. literature as being mainly frugivorous in nature. This Both studies made strides toward the advancement of being due to most researchers observing them in the future assisted reproductive/artificial insemination for wild consuming fruits, nuts and seeds. The behaviour wildlife rodent species like the agouti. of burying seeds and nuts makes them important seed The trial and successful conception of a production dispersers in forested ecosystems. Often seen in pairs model was completed in the doctorate thesis by or sometimes solitary, the agouti is territorial with a Brown-Uddenberg (2008). This pilot research tested preferred forested habitat with good undergrowth an intensive system for production of this species using cover (Emmons and Feer 1997, 227). 14 females and further analysed captive production Agouti production modelling systems from two other sites. The result of Brown- Uddenberg’s research was a model system designed for The agouti is another highly popular wild meat in intensive production of this species in the neo-tropics. the Caribbean. In recent times, agouti production The reproductive anatomy of the male was revisited systems in the region have facilitated the generation again with a closer look at the accessory glands, and publication of considerable research. The first area which are important for reproductive function in of research for the agouti began with the description male mammals. The results provided one of the only of the digestive system by Garcia et al. (2000). Their detailed and accurate descriptions of the gross and initial anatomical digestive studies concluded that micro anatomy of this species (Mollineau et al. 2009). the agouti was likely a hindgut fermenter. It was also In Mollineau et al. (2012) the authors were able to found to possess teeth that were characteristic for an describe the anatomical stages of penile erection in herbivorous type diet and a developed cecum that the agouti utilising the electro ejaculation technique indicated the importance of plant matter digestion. developed in 2008. Combined, these studies were able Garcia et al. 2000 studies further determined that the to document the reproductive system of the male in small intestine predominated in the gastrointestinal much greater detail, increasing the knowledge store tract with an abundance of villi, which suggested a on the red-rumped agouti species in the Caribbean and more omnivorous versus a strictly frugivorous diet as thus aiding in the future captive reproduction of this had previously been reported for this species (Emmons wildlife species. and Feer 1997). Another important observation was The female agouti’s reproductive anatomy also made on the large size of the anal glands. The function received attention, with a detailed description provided at that time was largely undetermined, leading the by Singh et al. (2014). The findings of this research authors to suggest that much greater investigation added support to the use of vaginal cytology as a would be needed on these glands. method for the evaluation of reproductive stages of Reproduction was the next area of research for agouti the female agouti. Vaginal cytology already being a production and the anatomy of the male reproductive technique widely utilised for laboratory rodents (Cora system was described by Mollineau et al. (2006). This et al. 2015) and wildlife species (Asa et al. 1992) study highlighted some interesting, hitherto unknown to accurately determine the stage in the females’ features on agouti reproductive anatomy. Including reproductive cycle and often used as a tool to improve the U-shaped formation of the penis and lateral penile reproduction in captivity. cartilages, which hinted to their role as a stimulant for The most recent studies emerging from the region the female agouti and further suggested that agouti have focused on the behaviour, health and diet of this females may be reflex ovulators. Both theories, species in captivity. Health and disease were examined however, require future study to be confirmed for this via a survey study conducted by Jones and Garcia species. (2017) on gastrointestinal parasites. The results found Conference Proceedings 89

that the agouti supported a number of parasite species, of an intensive production model for the agouti and hinted to the lack of impact these parasites had on (Dasyprocta leporina) a neotropical rodent in the body condition of this species. This led to follow Trinidad, West Indies.” PhD dissertation. up study which supported this initial observation; the Cabello, D. 2006. “Reproduction of Didelphis researchers concluding that the agouti may be able to Marsupialis (Didelphimorphia: Didelphidae) tolerate parasitic loads and still retain a healthy weight in the Venezuelan Andes.” Acta Theriologica (Jones and Garcia 2018). In the tropics, high parasitism 51 (4): 427-433. can often have very negative effects on livestock and Cáceres, N. C., and E. L. A. Monteiro-Filho. 1999. animals being reared in captivity (Lamy et al. 2012). “Body Size in Natural Populations of Didelphis Animals that are able to tolerate and display parasite (Mammalia: Marsupialia) from Southern resistance are often preferred and actively sought for Brazil.” Revista Brasileira de Biologia 59 (3): traits used in production. The agouti, as suggested by 461-469. these studies, might be one of these species. Cerqueira, R., and B. Lemos. 2000. “Morphometric Differentiation between Neotropical Black- Observations on the diet of captive agouti, combined Eared Opossums and D. aurita.” Mammalia with the early anatomical observations by Garcia et 64: 319-327. al. 2000, led to nutrition and diet studies by Jones et Cora, Michelle C., Linda Kooistra, and Greg Travlos. al. (2019). The results of these studies challenged the 2015”Vaginal cytology of the laboratory rat and previously held belief that the agouti’s diet was mainly mouse: review and criteria for the staging of frugivorous, by finding evidence that this species the estrous cycle using stained vaginal smears.” consumed animal matter in captivity. These findings Toxicologic pathology 43 (6): 776-793. suggest that the agouti might be better described as an de Albuquerque, Natália Inagaki, Diva Anélie omnivore versus a frugivore, as it appears to consume Guimarães, Hilma Lúcia Tavares Dias, both plant and animal matter (Jones et al. 2019). Field Yvonnick Le Pendu, Priscila Reis Kahwage, and studies however, are needed on wild populations Alexandre Rossetto Garcia. 2010. “Intensive to determine if this omnivorous behaviour is also production system of collared peccary (Pecari displayed in wild populations. tajacu) in Brazilian Amazon.” Advances in Animal Biosciences 1 (2): 480-481. Like the opossum, agouti research still has many areas Emmons, L. H. , and F. Feer. 1997. Neotropical requiring greater study including parasite load and its Rainforest Mammals: A Field Guide. Second effect on reproduction, reproductive behaviours and ed. Chicago: University of Chicago Press. captive nutrition. Garcia, Gary Wayne, Quinn Stephan Baptiste, Conclusion Andrew Oche Adogwa, Masakazu Kakuni, Kazuyosi Arishima, and Takashi Makita. Much pioneering work on wildlife production has 2000. “The digestive system of the agouti emerged from the Caribbean region. These studies (Dasyprocta leporina)-gross anatomy and serve to illustrate that neo-tropical species like the histology.”  Japanese Journal of Zoo and agouti and opossum can be successfully produced in Wildlife Medicine 5 (1): 55-66. captivity and thus can serve as alternative sources of Garcia, Gary. 2003. “Intensive Tropical Wildlife (Non- animal protein, improving food security and conserving Domestic) Production Systems: Introduction Caribbean wildlife biodiversity by reducing hunting to Species Production Modelling.” Lecture, pressure on wild animal populations. University of the West Indies, St Augustine, September 2003. References Harder, J. D., and David W. Fleck. 1997. “Reproductive Asa, Cheryl S., Randall E. Junge, J. Stephen Bircher, Ecology of New World Marsupials.” In Gary A. Noble, Kristen J. Sarri, and Edward D. Marsupial Biology: Recent Research, New Plotka. 1992. “Assessing reproductive cycles Perspectives, edited by Norman Saunders and and pregnancy in cheetahs (Acinonyx jubatus) Lyn Hinds, 179-203. Australia: UNSW Press. by vaginal cytology.” Zoo Biology 11 (3): 139- Horta, Maurício C., Marcelo B. Labruna, Adriano 151. Pinter, Pedro M. Linardi, and Teresinha T. S. Schumaker. 2007. “Rickettsia Infection in Brown-Uddenberg,RoxannC.2008.“Conceptualisation 90 Conference Proceedings

Five Areas of the State of São Paulo, Brazil.” Mollineau, W. M., A. O. Adogwa, and G. W. Garcia. Memórias do Instituto Oswaldo Cruz 102 (7): 2008a.”Spermatozoal morphologies and 793-801. fructose and citric acid concentrations in Hunsaker II, Don, ed. 1977. The Biology of Marsupials. agouti (Dasyprocta leporina) semen.” Animal New York: Academic Press. Reproduction Science 105 (3-4): 378-383. Jimenez, F., F. Catzeflis, and S. Gardner. 2011. “Structure of Parasite Component Communities Mollineau, W. M., A. O. Adogwa, and G. W. Garcia. of Didelphis Marsupialis: Insights from A 2008b. “A preliminary technique for electro- Comparative Study.” Journal of Parasitology ejaculation of agouti (Dasyprocta leporina).” 97 (5): 779-787. Animal Reproduction Science 108 (1-2): 92-97. Jones, Kegan Romelle, and Gary Wayne Garcia. 2017. “A survey of the gastrointestinal parasites Mollineau, W. M., A. O. Adogwa, and G. W. Garcia. present in the Agouti (Dasyprocta leporina) 2009. “The gross and micro anatomy of reared intensively in Trinidad.” Livestock. the accessory sex glands of the male agouti Research for Rural Development 29 (10): 1-7. (Dasyprocta leporina).” Anatomia, histologia, Jones, Kegan Romelle, and Gary Wayne Garcia. 2018 embryologia 38 (3): 204-207. “Observations on endoparasitic load in captive reared agoutis (Dasyprocta leporina) without Mollineau, W., A. Adogwa, N. Jasper, K. Young, and G. anthelmintic exposure in Trinidad, Republic of Garcia. 2006. “The gross anatomy of the male Trinidad and Tobago” Livestock. Research for reproductive system of a neotropical rodent: Rural Development 30 (10): 1-7. the agouti (Dasyprocta leporina).” Anatomia, Jones, Kegan Romelle, Kavita Ranjeeta Lall, and histologia, embryologia 35 (1): 47-52. Gary Wayne Garcia. 2019. “Omnivorous behaviour of the agouti (Dasyprocta leporina): Mollineau, W. M., T. Sampson, A. O. Adogwa, and G. A neotropical rodent with the potential for W. Garcia. 2012. “Anatomical Stages of Penile domestication.” Scientifica. Erection in the Agouti (Dasyprocta leporina) Jurgelski, W., Jr. 1974. “The Opossum (Didelphis Induced by Electro-Ejaculation.” Anatomia, virginiana Kerr) as a Biomedical Model. Histologia, Embryologia 41 (5): 392-394. I. Research Perspective, Husbandry, and Laboratory Technics.” Laboratory Animal Nowak, R. M. 1999. Walker’s Mammals of the World. Science 24 (2): 376-403. 6th ed. Vol. 1. Baltimore, USA: The Johns Jurgelski, William Jr., William Forsythe, David Dahl, Hopkins University Press. Lyle D. Thomas, John A. Moore, Paul Kotin, Hans L. Falk, and F. Steven Vogel. 1974. “The Oliveira, Heloiza, H., Adilson B. Almeida, Raimundo Opossum (Didelphis virginiana Kerr) as a W. Carvalho, Valmir Gomes, Nicolau M. Serra- Biomedical Model. II. Breeding the Oppossum Freire, Igor Quinelato, and Acácio G. Carvalho. in Captivity: Facility Design.” Laboratory 2010. “Siphonaptera of Small Rodents and Animal Science 24 (2): 404-428. Marsupials in the Pedra Branca State Park, Krause, W.J, and W.A. Krause. 2006. “The Opossum: State of Rio De Janeiro, Brazil.” Brazilian It’s Amazing Story.” In. Columbia: University Journal of Veterinarian Parasitology 19 (1): of Missouri. 49-54. Lamy E., van Harten S., Sales-Baptista E., Guerra M.M.M., de Almeida A.M. 2012. Factors Redford, K. H. , and J. F. Eisenberg. 1992. Mammals of Influencing Livestock Productivity. In: Sejian the Neotropics. The Southern Cone. II. Chile, V., Naqvi S., Ezeji T., Lakritz J., Lal R. (eds) Argentina, Uruguay, Paraguay. Chicago, IL: Environmental Stress and Amelioration University of Chicago Press. in Livestock Production. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3- Roth, Harald H. 1997.Wildlife resources: a global 642-29205-7_2 account of economic use. New York: Springer. Rushton, Jonathan, Rommy Viscarra, Cecilia Viscarra, Frederick Basset, Rene Baptista, and David Brown. 2005. “How important is bushmeat consumption in South America: now and in the future.” ODI wildlife policy briefing. Shripat, V. 2011. Didelphis Marsupialis (Common Opossum or Manicou) In The Online Guide to the Animals of Trinidad and Tobago: The University of the West Indies. Conference Proceedings 91

Siciliano, L. 2013. Didelphis Marsupialis. In Animal Tardieu, L., and G. W. Garcia. 2021. “Breeding and Diversity Web: University of Michigan, reproductive behaviour of the neo-tropical Museum of Zoology. opossum, Didelphis marsupialis insularis, Allen 1902 under captive conditions.” Brazilian Singh, Michele D., Andrew O. Adogwa, William M. Journal of Biology 82. Mollineau, and Gary W. Garcia. 2014. “Gross and microscopic anatomy of the reproductive Tardieu, L., W. Rollock, and G. W. Garcia. 2019a. tract of the female agouti (Dasyprocta “Wildlife rehabilitation: A case study of the leporina): a neotropical rodent with potential neo-tropical, opossum Didelphis marsupialis for food production.” Tropical Agriculture insularis, Allen 1902.” Brazilian Journal of (Trinidad) 91 (1): 38-46. Biology 80: 529-534. Sleeman, J.M. And Clark Junior, E.E., 2009. Clinical Tardieu, L., Venkatesan Sundaram,Andrew O.Adogwa, Wildlife Medicine: A New Paradigm for a and Gary W. Garcia. 2019b. “Anatomy and New Century. Journal of Avian Medicine histology of the gastrointestinal tract of the and Surgery, 17 (1): 33-37. http://dx.doi. neo-tropical opossum (Didelphis marsupialis org/10.1647/1082-6742(2003)017[0033:CWM insularis, Allen 1902).” Acta Zoologica 101 ANP]2.0.CO;2 (4): 384-391. Smythe, Nicholas, and Ofelina Brown de Guanti. 1995. Tyndale-Biscoe, C.H., and R.B. Mackenzie. 1976. “The domestication and husbandry of the paca “Reproduction in Didelphis Marsupialis (Agouti paca).” FAO Conservation guide 26. and D. Albiventris in Colombia.” Journal of Mammalogy 57 (2): 249-265 Tardieu, L., and Gary Wayne Garcia. 2018. “Didelphis species, neo-tropical animals with the potential Tyndale-Biscoe, C. Hugh. 2005. Life of marsupials. for intensive production part 3: theoretical CSIRO publishing. models of a semi-intensive housing facility for the neo-tropical opossum (Didelphis marsupialis insularis, Allen, 1902).” Tropical Agriculture 95 (4): 350-368. 92 Conference Proceedings

Analytics in agriculture: GreenAlysis system in a dwarf pommecythere (Spondias dulcis Parkinson) orchard in Trinidad and Tobago T. G. Paul1, G. Eudoxie and L. Roberts-Nkrumah 1MPhil (Computer Science) student and developer of the GreenAlysis System Email: [email protected] Data analytics can help to achieve the goal of improved farming methods for increased productivity in small-scale farms. The use of data analytics is not widely reported in the Caribbean, highlighting a gap and opportunity to showcase its potential. The purpose of this paper is to evaluate the installation and performance of a low-cost GreenAlysis (Caribbean) system for environmental data in a small dwarf pomme cythere orchard at The University of the West Indies, Trinidad and Tobago. Key installation challenges included security and ease of access issues in location selection for the gateway and the effect of sensor node field placement on the signal strength of the long-range antenna. Obstacles in the path between the gateway and the sensor node affect the quality of service provided between the two radios. During the COVID-19 pandemic, limited access to the gateway antenna prevented the normal functioning of the system. However, the data can be accessed directly via a short-range wi-fi connection available from the node in the field. Data collected from May 2021 to August 2021 are presented and include GPS coordinates, temperature, relative humidity, wind speed and Photosynthetic Photon Flux Density. This system demonstrated its usefulness for collecting environmental data in the field which can facilitate interaction between farmers and researchers for decision-making for site-specific management. It can support collaboration within research teams and serve as an excellent teaching tool. Further research is warranted to increase the effectiveness of the system in a wider range of terrains where signal strength challenges can increase. Work is also needed on the application of these data to inform crop development and performance and on system design for different crop heights and architectures. Conference Proceedings 93

Breadfruit commercialisation for sustainable food and nutrition security in the Caribbean Oral Daley and Laura Roberts-Nkrumah Department of Food Production, Faculty of Food and Agriculture, The University of the west Indies, St. Augustine, Trinidad and Tobago Email: [email protected] Breadfruit (Artocarpus altilis) occupies a strategic position to contribute to food and nutrition security in the Caribbean because it is widely distributed and well adapted to local growing conditions. The fruit is a rich source of complex carbohydrate, provides several vitamins and minerals and has a complete amino acid profile. It can be cooked in several ways such as boiling, steaming, roasting, baking or frying for a wide range of dishes including those in which it may serve as a substitute for some imported starchy staples. Breadfruit may be useful to prevent or mitigate the effects of several non-communicable diseases such as type 2 diabetes, obesity and hypertension, which are of serious public health concern in the region. Furthermore, breadfruit trees are long lived, their production is environmentally friendly and will sequester carbon while producing nutrient rich fruits. Despite the many benefits offered by breadfruit, it has remained underutilised throughout the Caribbean. Recent research and innovations have provided opportunities to address challenges such as seasonal bearing, tree height management, limited germplasm, poor shelf of fruits and lack of planting materials that have long hindered the commercialisation of the crop. Furthermore, several researchers have focussed on innovation that can assist with increased consumption and utilisation. This paper argues that with increased public awareness of the benefits of breadfruit consumption, investments and policy support from the private sector and government, breadfruit can become an economically and nutritionally important crop for the Caribbean to help reduce the heavy reliance on food importation and contribute to healthier populations. It is recommended that there should be deliberate efforts aimed at commercialising breadfruit by including it into mainstream agricultural value chains through application of existing research, encouraging new research and promotion of its nutritional and environmental benefits. This will ensure sustainable production to support anticipated increased demand. Keywords: Breadfruit germplasm, cultivar evaluation, environmental benefits, non-communicable diseases. 94 Conference Proceedings