Reduce Exposure to Reduce Risk

eastern China, the southern peninsula of Korea and weaker, aftershock of stormy winds still maintained Kyushu Island of Japan. Then, it became a tropical (Figs. 4b and 6b). depression and turned toward the East Sea of Korea (the Sea of Japan). (a) Figure 2: Vertical cross section of horizontal wind (m/s) on a straight line in the west-east of typhoon center at 0900LST, September 5, 2004. At 0900LST September 7, after the main body of (b) the tropical depression passed by the Korea Strait, the split ones were divided again into four parts Figure 3: Surface atmospheric pressure change for a such as the first one (I) in the Yellow Sea in the day at (a) 0900LST, September 5, 2004 and (b) west of Korean peninsula, the second one (II) the 0900LST, September 6, respectively. Negative southern sea of the Korea, the third one (III) in the value denotes deepening of atmospheric pressure for East Sea of Korea and the fourth one (IV) in the a day. eastern coast sea of Japan near Kyusu Island. Vertical cross section of horizontal wind on a After the split one-III passed trough the Korea Strait straight line in the west-east of a typhoon center at into the East Sea of Korea, the tropical depression 0900LST, September 5, 2004 showed almost a became just a low pressure, but the surface symmetric (Fig. 2). However, after the tropical atmospheric pressure was more intensified with - depression changed from tropical cyclone intruded 15.42hPa deepening than one day before and wind in the central part of the East Sea at 2100LST, the speed was also reinforced into stormy winds, due to vertical structure of wind became asymmetric from the terrain-induced channel effect of the high the typhoon center (Fig. 7), showing more stronger eastern mountainous coast of Korean peninsula, the eastern Russia near Vladivostok and the high western mountainous coast of Honshu, Japan (Figs. 4a and 6a). Then, at 2100LST, September 7, after the typhoon passed by the central part of Korean peninsula, the parts I and II were vanished, but part III was reinforced in the East Sea of Korea showing very stormy wind of 14~22m/s in the eastern coast and Part IV also maintained the aftershock of stormy winds. Even though the low pressure became slowly Reduce Exposure to Reduce Risk 140

in the horizontal and vertical in the east side of its center. When the tropical depression was changed into an extra-tropical cyclone (i.e., a low pressure), geopotential tendency (/t; m/day) at 500hPa for 24 hours at 2100LST, September 7, 2004 had a negative value of 276m/day, which implied the shrunken of atmospheric depth from 500 hPa level to the ground surface for 24 hours in Fig. 8. The compressed atmospheric in the East of Korea caused the increase of channel flow in the layer, resulting in horizontal wind speed near the sea surface in the East Sea of Korea. Thus, aftershock of stormy wind could maintain continuously until (a) (a) (b) Figure 5: Surface winds (m/s) in (a) a coarse-mesh domain with a 27 km horizontal grid interval at 0900LST, September 5, 2004, before a typhoon- Songda passed by Cheju Island (a small circle in a box), and (b) at 0900 LST, September 6. A small circle in the typhoon boundary in the south of Cheju Island in (a) denotes typhoon eye and large white and curved arrows denote wind driven current and cyclonic winds, respectively. (b) Figure 4: As shown in Fig. 2, except for at (a) 0900LST, September 7, 2004 and (b) 2100LST, respectively. Reduce Exposure to Reduce Risk 141

(a) Figure 7: As shown in Fig. 2, except for 2100LST, September 7, 2004. (b) Figure 8: Geopotential tendency (/t; m/day) at 2100LST, Spetember 7, 2004. Negative value Figure 6: As shown in Fig. 2, except for (a) at 0900 denotes the shrunken of atmospheric depth from 500 LST, September 7, 2004, when the typhoon just hPa level to the ground surface, resulting in the passed by the Korea Straight and (b) at 2100 LST, increase of horizontal wind speed in the East Sea of September 7, after the weakening and splitting of Korea. the typhoon due to land and shallower sea depth and ACKNOWLEDGEMENTS becoming an extra-tropical cyclone (i.e., a low Authors would like to express our best thanks to two reviewers on very helpful comment and suggestions. pressure) in the East Sea of Korea. Large white and This work was funded by the Korea Meteorological curved arrows denote wind driven current and Administration Research and development Program cyclonic winds, respectively. under Grant CATER 2006-2308 for 2010~2012- “Generation mechanism and prediction of windstorm in the mountainous coast”. Reduce Exposure to Reduce Risk 142

REFERENCES Implications with respect to intensity change. Anthes, R. A. and Chang, S. W., 1978, Response of Monthly Weather Review, 131, 1783-1796. Elsner, J. B., 2003, Tracking hurricanes. Bulletin of the hurricane boundary layer to changes in sea American Meteorological Society, 84 (3), 353- surface temperature in a numerical model. 356. Journal of Atmospheric Sciences, 35, 1240- Gilbes, F, Armstrong, R. A., Webb, R. M. T. and 1255. Muller-Karger, F. E., 2001, SeaWiFs helps asses Babin, S. M., Carton, J. A., Dickey, T. D. and hurricane impact on phytoplankton in Caribbean Wiggert, J. D., 2002, Hurricane-induced Sea, Eos, Transactions. American Geophysical phytoplankton blooms in the Sargasso Sea. Union, 82, 529-533. Proceeding of the 2002 AGU/ASLO Ocean Hong, S. Y. and Lim, J. O., 2006, The WRF single- Sciences Meeting, American Geophysical moment 6-class microphysics scheme (WSM6). Union, Honolulu, Hawaii. Journal of Korean Meteorological Society, 42, Cheung, T. C. and Chan, P. W., 2009, Improving 129-131. wind and rain simulations for tropical cyclones Jian, G. J. and Wu, C. C., 2008, A numerical study with the assimilation of Doppler radar data. of the track deflection of supertyphoon Haitang Proceeding of the 10th Annual WRF Users’ (2005) prior to its landfall in Taiwan. Workshop, Boulder, Colorado, 1-835. Knauss, J. A., 2005, Introduction to physical Choi, H., 2010, Extreme cold sea outbreak near oceanography, 2nd Ed.,Waveland Press, Illinois. Cheju Island, Korea by strong wind and Monaldo, F. M., Sikora, T. D., Babin, S. M. and atmospheric pressure change under typhoon Sterner, R. E., 1997, Satellite imagery of sea Rusa. Disaster Advances, 3(1), 32-41. surface temperature cooling in the wake of Choi, H., Choi, D. S. and Choi, M. S., 2009, Heavy Hurricane Eddouard. Monthly Weather Review, snowfall by orography and wind shift in cold 125, 2716-2721. front crossing Korean eastern coast. Disaster Advances, 2(4), 48-60. Cione, J. J. and Uhlhorn, E. W., 2003, Sea surface temperature variability in hurricanes: Reduce Exposure to Reduce Risk 143

A CONCEPTUAL MODEL FOR LANDSLIDE PREDICTION IN SRILANKA L.D.C. S. Subhashin and H.L. Premaratne Lecturer (Probationary)Department of Information Technology, Faculty of Management studies and commerceUniversity of Sri JayewardenepuraSri [email protected], Senior LecturerUniversity of ColomboSchool of Computing (UCSC)Sri [email protected] ABSTRACT Landslides are most recurrent and prominent disasters in Sri Lanka. Sri Lanka has been subjected to a number of extreme landslide disasters that resulted in a significant loss of life, material damage, and distress. It is required to explore a solution towards preparedness and mitigation to reduce recurrent losses associated with landslides. This research is aimed to identify a conceptual model for landslide prediction. Two conceptual models were identified based on the lessons learned through a literature review. The two models are the Artificial Neural Network (ANN) model and the Hidden Markov model (HMM). Both models function on training based on past (or historical) data. The data for training known as the training set consist of factors that contribute to landslides. Many factors contribute to landslides. The conceptual models which consist of all these factors initially recognize specific factors that might influence to occur landslides in Sri Lanka. A questionnaire was designed based on the information in previous research around the world. Numerous factors were considered when defining the research sample, but mainly focused on the lessons learned through a literature review. Data was gathered from one hundred and twenty six experts in the area of landslides from four national universities in Sri Lanka. Twelve factors were initially identified as the most relevant and it is important to note that the most significant factor that influence to occur landslides is rainfall which has the highest mean value in analysis and other more prominent factors were Soil Material, Geology, Land Use, Curvature, No of Previous Occurrence, Soil Texture, Slope, Aspect, Influence of Construction, Soil Drainage, and Soil Effective Thickness. These two conceptual models which consist of the identified factors will be trained to predict three classes namely, „landslide occurs‟, „landslide does not occur‟ and „landslide likely to occur‟. Once trained, the model will be able to predict the most likely class for the prevailing data. The research is continued to implement a system which will be thoroughly tested in order to predict landslides with a high accuracy. KEY WORDS: Conceptual Model, Landslides, Prominent Factors, Neural Network Model, Hidden Markov Model 1. INTRODUCTION in the perceptible downward and outward movement of soil, rock, and vegetation under the influence of Landslides are a common natural phenomenon in gravity. The materials may move by falling, toppling, many part of the world, especially in hilly or sliding, spreading, or flowing. Some landslides are mountainous terrains. A landslide event is defined as rapid, occurring in seconds, whereas others may take “the movement of a mass of rock, debris, or earth hours, weeks, or even longer to develop. (soil) down a slope (under the influence of gravity) (cruden, 1990).The word “landslide” also refers to Many factors contribute to slides, including geology, the geomorphic feature those results from the event. gravity, weather, groundwater, wave action, and Other terms used to refer to landslide events include; human actions. Although landslides usually occur on mass movements, slope failures, slope instability and steep slopes, they also can occur in areas of low relief terrain instability. In spite of the simple definition, according to bandara et al. (1994). Landslides can landslide events are complex geological/ occur as ground failure of river bluffs, cut and-fill geomorphologic processes and therefore difficult to failures that may accompany highway and building classify. The most commonly used landslide excavations, collapse of mine-waste piles, and slope classification system is based upon the material failures associated with quarries and open-pit mines. type and the type of movement described by Underwater landslides usually involve areas of low (cruden and varnes , 1996). relief and small slope gradients in lakes and reservoirs or in offshore marine settings. Landslide is a general term for a wide variety of down slope movements of earth materials that result Reduce Exposure to Reduce Risk 144

Typically, a landslide occurs when several of these and postulated that the three peniplains are the result factors converge. Every year in Sri Lanka especially of successive bedrock uplifts. The recent detailed in mountain regions like Badulla district, landslides structural and tectonic mapping of central highlands damage many houses and cause millions of rupees indicates that in addition to the vertical epirogenic damage to buildings, roads, railways, pipelines, movements of the southerly drifting manipulate of agricultural land and crops. Sri Lanka, there are horizontal thrust developed regionally and a series of strike slip faults along An effective solution to minimize the damage from a mega-lineaments (Vitanage, 1994). Some of the landslide will be a mechanism for early prediction. If lineaments appear to be active and some of the older we can come up with a prediction model, it will be highly weathered lineaments are commonly more benefit to individuals as well as the whole associated with large destructive recurrent landslides. society. This research is to develop a conceptual Since there are no simple mechanisms to foresee model to predict landslide disasters and the study is instabilities, monitoring is the most appropriate mainly focussed on the landslides in the Badulla mechanism to understand their behavior. district. The government during this period was aroused the 1.1 Goals and objectives need of early warning systems and it came up with The overall aim of the survey and research is to atmospheric models that forecast the temporal and develop a conceptual model to predict landslide spatial distribution of rainfall which is a useful disasters. alternative source of rainfall information. This is now increasingly being used in hydrological applications The specific objectives of the research are enunciated though it mainly focuses on rainfall information. below; Landslide disaster risk reduction demands integration 1. Find the factors affected for landslides. of a number of disciplines associated with various aspects of physical and hydro-meteorological 2. Come up with conceptual models for landslides characteristics of a region as well as social and prediction. cultural dimensions. Therefore, landslide disaster mitigation requires collective and corporative efforts 1.2 METHODOLOGY of all relevant r&d institutions lead by strong The Research problem was defined and a research executing organizations of the country. design was prepared accordingly. A theoretical background was reviewed through a literature survey. One of the mechanisms that evolved during the The scope of the research was defined through a disaster in May 2003 and an implementation of further literature survey. To achieve the objective one landslide disaster mitigation works, was the given above, literature, data gathered from the establishment of a ad-hoc group named “operation experts who are experienced in landslide and professional combine” established at Rathnapura, the detailed ground survey of the priority areas of most affected area. This was carried out in landslide prone districts of the hill country were used. accordance with the action plan proposed by the The collected data are analyzed and the output will Central Engineering Consultancy Bureau (CECB) at be finalized. the council meeting of the Disaster Management Committee held in Colombo, 28th May 2003. Existing models are investigated through a literature review study existing models and conceptual models An early study on this subject was carried out by Lee are proposed. Finally a landslide prediction system et al. (2004). The purpose of this study was to will be implemented. develop landslide susceptibility analysis techniques using an Artificial Neural Network (ANN) and to 1. LITERATURE REVIEW apply the newly developed techniques to the study Sri Lanka is an island in the northern Indian Ocean area of Boun in Korea. Landslide locations were just south of southernmost part of India and extends identified in the study area from the interpretation of in latitude form approximately 060 N to 100 E and in aerial photographs, field survey data, a spatial longitude from approximately 800 N to 820 E with database of the topography, soil type, timber cover, an extent of about 65,000 km2. geology and land use. The landslide-related factors (slope, aspect, curvature, topographic type, soil Adams (Adams, 1929) was the first to draw attention texture, soil material, soil drainage, soil effective to the existences of three well-marked plains of thickness, timber type, timber age, and, timber erosion cut in the precambrian rocks of Sri Lanka. diameter, timber density, geology and land use) were These “three terraces” present three successive stages extracted from the database. Using those factors, of denudation brought about by successive uplift of landslide susceptibility was analyzed by artificial Island as a whole. On morphological grounds Wadia neural network methods. Maps constructed in a (Wadia, 1943) rejected this “erosion terrace theory” vector format spatial database using the GIS software Reduce Exposure to Reduce Risk 145

ARC/INFO were used for the application of ANN Table 1: Variables 2)Code methods. T1 1)Topography T2 The Geological Society of London (Fuxiang and Lan, Slope T3 2002) did a research to come up with a prediction Aspect T4 model for landslide disasters using Artificial Neural Curvature Networks and Geographic Information Systems. Topographic Type 4) According to this research it was not possible to use S1 ANNs and GISs effectively to predict landsides 3)Soli S2 objectively but it showed the significance of spatial Soil Texture S3 distribution of the occurrence of landslides resulting Soil Material S4 in the development of a landslide distribution Soil Drainage database. Soil Effective Thickness 6) W1 In Philippines, (Lee1 and Evangelista, 2006) did a 5)Wood W2 research about landslide-susceptibility analysis Timber Age W3 techniques using an Artificial neural network and a Timber Diameter W4 Geographic Information System (GIS) applied to the Timber Type W5 Baguio City. Data preparation involved the Timber Density W6 digitization or creation of a GIS database which Geology included the topographical, geomorphic- logical, Land Use 8) geological and land cover data. In the verification of OT1 landslide-susceptibility maps, the Artificial Neural 7)Other Factors OT2 Network showed a very high prediction accuracy of No of previous Occurrence OT3 93.20% in the case of 0 slope. Earth Quakes OT4 Rain Fall Landslides are one of the most hazardous natural Influence of Construction disasters, not only in Sri Lanka but also in worldwide. The Governments and research Before proceeding, all the variables were laid down institutions worldwide have attempted for years to before a panel of experts for evaluation. The panel of assess landslide hazards and risks and to determine experts selected the most influential and relevant their spatial distribution. In this research it is factors and selected the most relevant and important expected to develop a model to predict landslides factors affecting for landslides. using ANNs and GISs. A likert scale was used for this rating. Respondents 2. DATA COLLECTION AND ANALYSIS were allowed to rate each factor in a scale from 1 to 5 according to their preference. Number 1 was for The primary objective of the study was to identify „totally disagree‟ variables and number 5 was for factors which effect to occur landslides, through a „strongly agree‟ variables. quantitative methodology with a survey based approach. Numerous factors were considered when Selecting respondents for the data collection survey defining the research sample, but mainly focused on was also important to maintain the quality and the lessons learned through the literature review. As standards of the findings. Therefore, a dedicated the main focus of the research was on landslides in approach was practiced when selecting respondents Sri Lanka, GIS experts in several universities were from each university. Because their perception was selected as the resource persons for the survey. evaluating the variables and the output was totally Data collection was done via the Likert scale questionnaire filled by the representatives of the selected universities. Data were collected under four categories namely, Topography, Soil, Wood and Other factors. By referring to researches done before, variables were identified. A questionnaire was prepared to gather information for these variables. The variables defined describe the factors that affect for landslides. All the variables were selected from past researches (see Table 1) Reduce Exposure to Reduce Risk 146

based on it. In total 200 copies of the Questionnaire According to the Figure 1, the mean values of the were distributed among four universities, and 126 of variables were of „distributed‟ nature. Landslides which were duly filled and returned to the authors. were occurring due to all the factors that we have Professors, Senior Lectures, Lectures, Civil identified above. The most prominent factor for Engineers and PhD students were among the experts occurring landslides in Sri Lankan context is the rain who responded to the survey. fall which scores the highest mean value. Soil material is also highly significant. In factor analysis A reliability analysis was carried out to determine the we ignore the variable „earth quakes‟ which had a consistency and stability of data. Cronbach‟s alpha minus component in the matrix because in Sri which is the coefficient for reliability was analyzed Lankan context it is not significant. Finally it was on each data set. Reliability of a measure was tested possible to identify twelve most prominent factors for both consistency and stability. Consistency which affect to occur landslides. These factors are indicates how well the items measurr a concept hanging together as a set. Therefore, to ensure the Rain Fall reliability of the data Cronbach‟s alpha coefficient Soil Material was calculated. Geology Land Use Submitting the data for factor analysis establishes a Curvature factorial validity. This confirmed that, the theorized No of Previous Occurrence dimensions have been emerged. Though variables Soil Texture indicated in each category were selected through Slope literature and confirmed by a panel of experts in Aspect order to identify the most relevant factors, a factor Influence of Construction analysis was carried out. Soil Drainage Soil Effective Thickness 3.1 Findings of the study Figure 1: Mean value and landslide factors The first objective of the research was achieved from descriptive statistics. A descriptive frequency Designing conceptual models analysis was done and the Table 2 shows the mean value for each factor after doing factor analysis. Since the main focus of the research is to identify a conceptual model for landslide prediction, two Table 2: Mean value for landslide factors conceptual models are initially identified namely, the Artificial Neural Network (ANN) model and the Variable Name Variable Mean Hidden Markov model (HMM). Rain Fall OT3 4.72 Both models function on training based on past (or historical) data. The data for training also known as Soil Material S2 4.34 the training set consist of actual data on factors that contribute to landslides. Out of the possible candidate Geology W5 4.25 factors only the most prominent ones have been identified which will be the inputs to each of the Land Use W6 4.24 models. Curvature T3 4.20 No of Previous Occurrence OT1 4.12 Soil Texture S1 4.11 Slope T1 4.10 Aspect T2 4.00 Influence of Construction OT4 3.83 Soil Drainage S3 3.82 Soil Effective Thickness S4 3.72 Reduce Exposure to Reduce Risk 147

Each of the two models will map the input data to Promin Landslide three possible classes. The output classes will be ent occurs „landslide occurs‟, „landslide does not occur‟ and factors „landslide likely to occur‟. Once trained, the model ANN will be able to predict the most likely class for the Landslide prevailing data. does not occur The ANN Model Landslide The first step towards Artificial Neural Networks was likely to introduced in 1943 when Warren McCulloch, a neurophysiologist, and a young mathematician, occur Walter Pitts, published a paper on how neurons might work. They modelled a simple neural network with Figure 2: Artificial Neural Network Model electrical circuits. The Hidden Markov Model Since a very high uncertainty is involved in predicting landslides, a Neural Network is a better The Hidden Markov Model offers an approach for solution as it handles uncertainty to a very high modelling dynamic systems. Processes are often degree to prediction of Landslide Disasters as a viewed as signals, which come from either one or dynamic prediction model. This is an effective multiple sources. These signals may have fixed or solution when it is difficult to build the relationship. unfixed parameters and may even be corrupted at First of all, integration of remote sensing data or GIS times. Therefore, a mathematical signal model can data is convenient. Secondly there is no need a be used to process the signal in hopes of deriving a specific statistical variable. Thirdly, accurate statistical process that describes the source. This first analysis is possible through training area datasets leads to the theory of Markov Chains, which can be which is a few because of pixel based calculations. extended further to Hidden Markov Models. Leonard Neural Network can process simultaneously E. Baum and several colleagues first published the qualitative and quantitative data. Therefore, a Neural theory of HMMs during the 1960s. In 1989, Network is a good candidate to establish these Lawrence Rabiner published his tutorial on HMMs, models. which explained the theory in a more general context (Lawrence). A geographic information system (GIS) integrates hardware, software, and data for capturing, While a Markov Process does model a stochastic managing, analyzing, and displaying all forms of process, it observes, rather than describes or predict geographically referenced information. A GIS was the process, which generates the observations. Here used to efficiently analyze the vast amount of data. in lies the importance of the Hidden Markov Model. This survey focused on exploring the methods of The HMM records observations as probability construction of a specific database using GIS. functions of the state, where the source is described as a hidden stochastic process (Taylor Sauder, 2011). The benefits of integrating GIS and artificial neural network are efficiency and ease of management, This model behaves in the same way as ANN (see input, display and analysis of spatial data for Figure 3. Identified factors will be trained to predict landslide susceptibility analysis. the same three classes as in the ANN „landslide occurs‟, „landslide does not occur‟ and „landslide Figure 2 shows conceptual models of Artificial likely to occur‟. Once trained, the model will be able Neural Network Model. This model which consist of to predict the most likely class for the prevailing data. the identified factors will be trained to predict three classes namely, „landslide occurs‟, „landslide does not occur‟ and „landslide likely to occur‟. Once trained, the model will be able to predict the most likely class for the prevailing data. Reduce Exposure to Reduce Risk 148

Landslide 5. Wadiya, D.N, (1943). The three superposed occurs peneplains in Ceylon. Records of the Geological Survey Department, Paper I: 25-38. Promine HMM Landslide nt does not 6. Vitange, P.W.(1984). Seismicity Neglected factors Aspects of Sri Lankan Landslide Studies, occur Proceedings National symposium of Landslide in Sri Lanka, Vol.1, pp 17-18. Landslide likely to 7. Atapattu, S., Herath, S. &Rupasinghe, N. (2008). Towards landslide risk reduction in Sri occur Lanka. Figure 3: Hidden Markov Model 8. Sauder, T.(2011). The Implementation of a Hidden Markov Model in MATLAB for the 3. CONCLUSION Prediction of Commodity Prices. Based on this research twelve factors were initially 9. Lee, M., Lee, S. & J. Wonand. (2008). identified as the most relevant to influence landslides. Landslide Susceptibility Analysis using Gis and It is important to note that the most significant factor Artificial Neural Network. that influence to occur landslides is rainfall which has the highest mean value in analysis and the other more 10. Fuxiang, Y., Lan, G. & Tao.(2002). The prominent factors were Soil Material, Geology, Land prediction of local landslide based on gis and Use, Curvature, No of Previous Occurrence, Soil neural networks. Texture, Slope, Aspect, Influence of Construction, Soil Drainage, and Soil Effective Thickness. 11. Evangelista, D. G. & Lee1S. (2006). Earthquake-induced landslide-susceptibility Two conceptual models were identified based on the mapping using an artificial neural network. lessons learned through a literature review. The two models are the Artificial Neural Network (ANN) model and the Hidden Markov model (HMM). 4. FUTURE WORK The research will be continued to implement the two models which will be thoroughly tested in order to predict landslides with a high accuracy. D. REFERENCES 1. Cruden, D.M. (1990). A Simple Definition Of A Landslide. Bulletin IAEG, 43:27-29 2. Cruden, D.M & Varnes, D. J. (1996). A Suggested Method for Reporting a Landslide. Bandara, R..M..S., Haridharan& Cruickshank, R.D. (1994), Landslides in Badulla District of Sri Lanka, Proceedings of National Symposium on Landslides in Sri Lanka, Vol 1, pp 127-132. 3. Varnes, D. J. (1978). Slope movement types and processes. 4. Adams, F.D. (1929). The geology of Ceylon. Canadian Journal of Research , Reduce Exposure to Reduce Risk 149

LANDSLIDE VULNERABILITY ASSESSMENT ALONG FOUR LANE ROAD EXPANSION INBETWEEN VISAKHAPATNAM TO BHIMUNIPATNAM, ANDHRA PRADESH – A GIS APPROACH P.Jagadeeswara Rao and P.V.V.Satyanarayana College of Engineering (A), Andhra University, Visakhapatnam-530 003, India E-mail: [email protected], [email protected] ABSTRACT The 22 km long coastal strip between Visakhapatnam city and Bhimunipatnam experiences rapid developmental activities, especially laying of a four lane road popularly known as the Beach Road, without regard to the structural fabric of the various geological formations. Two major rock headlands protruding into the sea have been cut to lay the road, are now became the major potential zones for landslides. A number of foot hill areas have also been severely altered leading to rock creep/soil creep on to the road. Huge rock blocks are often rolling down on to the road creating panic to the passersby. Owing to compositional variations, soil creep is also become active at places. The road is also passing through the intensely gullied Quaternary red sand deposits which are known for geological and archeological importance. In this study, IRS P6-LISS-III of January, 2011 and SRTM-90 satellite data have been used to decipher topographical, drainage parameters, land use/land cover and slope. The major land use is four lane road expansion besides several constructions on top of the hills altering the existing topography. A number of non-perennial streams in the low-lying areas are crossing the road leading to flooding during heavy rains. Soil samples have been collected along the road alignment to assess California Bearing Ratio and Permeability to study erodability and strength characteristics revealing high erodability and low strength. At places, migration of coastal sands on to the road is causing road accidents. The existing road has two potential threats: on the seaward side there may be a chance of breaching the road by the wave attack and on the landward side manmade landslides/soil creep may takes place on to the road. This study recommended retaining walls, grouting, rock bolting, geo-netting/iron-netting and turf as remedial measures to overcome the problems. KEY WORDS: Geographic Information Systems, Land use/land cover, grouting, geo-netting, California Bearing Ratio GEOMATICS BASED INTEGRATED SLOPE AND LAND USE/COVER MODELLING AND LANDSLIDE RISKS OF NILGIRI MOUNTAINS, SOUTH INDIA M. Muthukumar Spatial Technologies, Department of Rural Development, Gandhigram Rural Institute-Deemed University Gandhigram, Dindigul District, Tamilnadu, India, Phone: +91 9865505886 E-mail: [email protected], [email protected] ABSTRACT Landslides have become a fast spreading disaster in most of the mountainous systems of the world including India. As this phenomenon is the combined effect of natural earth system dynamics and the triggering parameters including the anthropogenic activities. The scientists & technocrats from all over the world have started studying involving different above parameters and various advanced scientific tools. In this connection the present study has been carried out taking two important and diagnostic natural and anthropogenic parameters, namely the Slope and Landuse / Land cover for assessing Landslide Vulnerability in part of the Nilgiri mountains. The same has been accomplished by taking the Nilgiri mountains of Western Ghats, South India as the test site which is chronically prone for Landslides. On the said study detailed landslide inventory was made in the study area and over 144 major and significant landslides were located and GIS database was generated accordingly. Similar GIS database was generated on the slope of the area, classifying the slope into Reduce Exposure to Reduce Risk 150

four categories viz: steep ( more than 450), moderate(200 to 450), shallow (30 to 200), and rolling(Less than 30). The Landuse/cover GIS layer was prepared using LANDSAT data and bringing out ten classes. These two GIS layers on slope and Landuse / cover were integrated using Intersect function in overlay tools in Analysis tool of Arc GIS software which has been lead into 37 number of integrated polygon classes. Over this, the GIS layer on Landslide was superposed and the number of Landslide was covered in each polygon class were calculated. The same was shown that, out of 144 landslides 86 Landslides fall in four polygon classes namely Plantation + Moderate (33), Plantation + Shallow (21), Dense Forest + Moderate (16) Forest Blank + Moderate (16) and the remaining 58 slides fall in 14 integrated slope and landuse/ cover polygon classes in a scattered fashion. Accordingly these four polygon classes were buffered out as most vulnerable areas to landslides. The papers deals in details about the technique used, the outputs and mitigation measures. KEY WORDS: Geomatics, Landslides, Slope, Landuse / Land cover, Nilgiris. POST COLLISION TECTONICS AND THE PHENOMENON OF LANDSLIDES IN PENINSULAR INDIA SM.RAMASAMY Gandhigram Rural University, Gandhigram – 624302, Tamilnadu, India E-mail: [email protected], www.ruraluniv.ac.in ABSTRACT The recent Geo- information technology based studies show that the Indian plate is tectonically active and resultantly the earth surface processes too are vibrant and controlling the natural disasters. The paper deals with the post collision tectonics and it’s contribution to landslides. Due to the still prevalent northerly directed compressive force which has originally drifted the Indian plate towards northerly and it’s obstruction by the Himalayas from the north, the Indian plate is whirling with East-West trending alternate arches and deeps and fracturing with North-South extension, Northeast-Southwest sinistral, Northwest-Southeast dextral and the East-West release fracture swarms. Such post collision tectonics significantly control landslides in many mountains in South India. For example, the Tirumala-Tiruppathi hills exposing the Precambrian Cuddapah quartzites is resting over one of the arches aligned along Mangalore –Chennai. As a result of such arching ,the Tirumala hills is also getting uplifted as revealed by deep fracture valleys ,shearing of the quartzites along the escarpments ,protruding dykes along the obsequent slopes of the granitic basement and active slope processes like debris flow, talus accumulation, colluvial fills , alluvial fan etc., all leading to and also symbolizing landslides. Similarly in the Nilgiris of south India, the landslides dominantly fell along NNW-SSE and ENE- WSW fractures and the landslides of the period from 1900 to 2000 AD seem to gradually shift from NNW-SSE towards NW-SE fractures indicating the anticlockwise rotation of the Nilgiris due to post collision tectonics. The landslides, palaeo scars and the slump scars of Shevroy-Kalrayan hills also seem to fall along ENE-WSW /NE-SW post collision sinistral faults. Thus the paper discusses the same in detail. KEY WORDS: Geo-information Technology, Tectonically induced Landslides, South India. Reduce Exposure to Reduce Risk 151

IMPROVING EXISTING LANDSLIDE HAZARD ZONATION MAP IN KMC AREA, SRI LANKA Oshadee Lasitha Potuhera1 and Vithanage Primali Anuruddhika Weerasinghe2 University of Kelaniya, Sri Lanka, E-mail: [email protected], [email protected] ABSTRACT In Sri Lanka, presently used landslide hazard zonation (LHZ) map which was developed by National Building and Research Organization (NBRO) is based entirely on geological, geomorphological and hydrological factors. As development expands into unstable hill slope areas under the pressures of increasing population and urbanization, human activities such as deforestation or excavation of slopes for road cuts and building sites, etc., have become important triggers for landslide occurrence. The present study was undertaken in highly urbanized Kandy Municipal Council (KMC) area in Sri Lanka. Main objective of the study was to validate the existing LHZ map with current active landslides and the improvement of the LHZ map for further use for management purposes. Validation of the existing LHZ map shows lowest percentage of landslide occurrence in the landslides most likely to occur zone and highest in the landslides are to be expected zone. To evaluate this unreasonable situation building density and transport lines were used. The relationship between building density and landslide occurrence was 97.1% and the relationship between distance from transport lines and landslide occurrence was 88.3% till 50 m. An improved LHZ map was developed including the effect of building density and distance from transport lines using frequency ratio method and improved LHZ map has an accuracy of 98.5%. KEY WORDS: Buildings, Hazards, Map, Landslide, Roads LANDSLIDE HAZARD AND RISK ANALYSIS OF KANDY MUNICIPAL COUNCIL AREA USIND GEOINFORMATICS TECHNIQUES Sivapatham Thavavathani1, N.D.K. Dayawansa2 and Ranjith Premalal De Silva3 1Survey department of Sri Lanka, Colombo, Sri Lanka 2Department of Agricultural Engineering, University of Peradeniya, Sri Lanka 3Uva Wellassa University, Badulla, Sri Lanka ABSTRACT Landslide hazard is a serious concern in mountainous areas of Sri Lanka. Urban Multi-hazard Disaster Mitigation Project has identified seven landslide prone districts in Sri Lanka based on the distribution and nature of past landslides. Kandy also has been identified as a district which is prone to landslides. This study was conducted with the objective of analyzing landslide hazard in Kandy municipal area of central Sri Lanka and producing a map of landslide hazard risk using Geo-informatics techniques. Topography, geology, land use, land form, hydrological network and inventory of past landslides were used to assess the landslide hazard risk in the study area. Landslide risk was identified in four categories as no risk, low, moderate and high risk. Spatial maps were developed for the identified factors in GIS environment and the maps were classified into above four classes based on the expert knowledge. As an example, above 30 degree slope areas were identified as high risk areas while 0-5 degree slopes were identified as no risk areas. Land slope was identified as the most influential factor on landslide hazard while geology was identified as the second most important. The distance from the hydrological network was identified as the least important factor in causing landslides among the selected factors. Rainfall was not considered as it is more or less uniform throughout the study area. Weights were given to the above spatial layers according to their relative importance. Those weights were decided based on the expert knowledge especially of the National Building Research Organization of Sri Lanka. Multi criteria decision making system was used in the process and weighted overlay analysis was performed to obtain the final outcome. The final hard zonation map was compared with the map of the landslide risk developed based on inventory of past landslides. According to the landslide hazard zonation map, majority of the study area (73%) comes under moderate hazard while high hazard is limited to 4% of the study area. It is important there only 1% of the area is Reduce Exposure to Reduce Risk 152

under no hazard category. When compared with the developed landslide hazards zonation map with the old landslide risk zones, majority of the study area (85%) comes under moderate hazard and low hazard is limited to 15% of the study area. Study identified the usefulness of GIS in mapping possible landslide hazard and spatial comparison of map of landslide risk developed using past incidences. The results would be more accurate if it included factors such as population density and built up density into the model. The final risk map can be used in formulating infrastructure development plans to minimize the hazards in the future. Reduce Exposure to Reduce Risk 153

Reduce Exposure to Reduce Risk 154

Technical Session-7 [Hall B]: GIT4 Infrastructure M anagement GIS-Integrated AHP Approach for Land Suitability Evaluation for Biofuel Crop 156 in Thailand 163 Naruemon Phongaksorn, Nitin K. Tripathi 168 171 Spatial Variability of Climatic Water Balance Under Rainfed Rice Based on Agro- 178 Ecological Zones in Kurunegala District 183 WMUK Rathnayake, DN Sirisena 183 184 Variability of Soil PH, P, K and ZN in Rice Breeding Fields at Rice Research and Development Institute DN Sirisena, WMUK Ratnayake, RMK Atapattu Importance of Incorporating Vegetation into Urban Infrastructure to Reduce Heat Island Effect: A Study in Colombo City of Sri Lanka N.L. Ukwattage, N.D.K. Dayawansa Application of Geographic Information System for Tea Plantation Management as Decision supporting tool A Case study at St Coombs Estate, Talawakelle, Sri Lanka N.N.K.Wellala, J.Gunatilake, H.W. Shyamale Risk Map Assessment Methodology – A Pilot Study for Central Region of Singapore J. Chandrasekar, D.K. Raju Identification of Urban Population Distribution and Spatial Pattern of Urbanization in the city of Colombo, Sri Lanka K.A.S.S. Wijesekera, Ranjith Premalal De Silva Mapping the Spatial Pattern of Rainfall and Elephant Intrution In Sri Lanka R M C W M Rathnayake Reduce Exposure to Reduce Risk 155

GIS-Integrated AHP Approach for Land Suitability Evaluation for Biofuel Crop in Thailand Naruemon Phongaksorn1, Nitin K. Tripathi1 1 Remote Sensing and Geographic Information Systems, School of Engineering and Technology, Asian Institute of Technology, Thailand E-mail: Naruemon [email protected], [email protected] ABSTRACT: The aim of this research is to integrate GIS and AHP approach for land suitability evaluation for biofuel crop, namely, cassava, at Nakhon Ratchasima province, Thailand. The objectives are to find suitable sites for biofuel crop using geo- information technology and estimate the potential availability of raw materials for biofuel production to meet the biofuel needs. The methodology is divided into three main phases: physical, socio-economic land suitability evaluation, and estimate the potential availability of raw materials for biofuel production. The results showed that the AHP was capable of classifying and mapping the physical land suitability with the 58.33% higher overall accuracy comparing to the classical maximum limitation approach. The majority of the area for the study area was highly physical suitable while the whole province had highly socio-economic land suitable for cassava plantation. Finally, their suitable areas for cassava plantation have performed the potential ethanol capacity matched with the ethanol plant capacity and ethanol consumption of the province. KEY WORDS: Land suitability evaluation; biofuel crop; GIS; AHP 1. INTRODUCTION The use of bio-ethanol is beneficial social, Malczewski, 2004). AHP method is a powerful and environmental, and economic aspects, particularly to flexible decision making process to help people set Thailand, as it reduces greenhouse gas (GHG) emissions priorities and make the best decisions (Saaty, 1980). In and fuel imports; it also contributes to job creation, consequence, this study employed the GIS integrated the income generation, and the stabilization of crop prices AHP to develop a model of the land suitability evaluation for farmers (Bell et al., 2011; Sorda et al., 2011). In for biofuel crops, namely, cassava that have been consequence, the biofuel consumption has increased increasingly used for biofuels in Thailand (Bell et al., dramatically for decade (Sorda et al., 2011) resulting in 2011). This study was conducted with three main the requests on agricultural land for biofuel feedstocks objectives, (a) to compare physical land suitability which also increase (Fischer et al., 2010). In addition, the evaluation between AHP approach and ground truth data, biofuel feedstocks are low productivity because the (b) to evaluate socio-economic land suitability, and (c) to agricultural practices are not good enough (Silalertruksa evaluate land suitability and estimate the expected and Gheewala, 2012). The integrated land suitability availability of raw materials for biofuel production to evaluation for biofuel crops and the current policy of the meet the biofuel needs. Government is one practical method to achieve the above mentioned issue (FAO, 1976; Fischer et al., 2010; 2. Study Area Haugen, 2010). The land suitability evaluation is to estimate suitability of the available land for a specific use There are significant characteristics that should be based on the physical parameters and/or socio-economic considered in study area in order to develop biofuel crop conditions of an area (FAO, 1976). It can be carried out suitability modeling. For instance, an area should have on higher productivity, profitability, and friendly wealthy biofuel crop cover and a biofuel plant location, environment as the available land without expansion. only the existing biofuel crops without expansion should There are several conventional methodologies for the be considered (MOAC, 2011) based on ideally being land suitability evaluation, which include maximum relatively high productivity, and the socio-economic limitation, parametric, and modeling. Numerous studies benefits and the nature conservation should be have been done to develop limitations of the above maximized. Nakhon Ratchasima Province (14°58′20″N mentioned method using GIS (Van Ranst, 1996; and 102°6′0″E) of Thailand, with approximately Samranpong et al., 2009). 2,126,888 hectares was selected as the study area (Figure 1). The province has already been dominated by cassava GIS-integrated multicriteria decision making (MCDM) (OAE, 2009) and mainly the biofuel plants with full method based on Analytic Hierarchy Process (AHP) capacity of 340,000 liters/day and fresh root cassava as approach is an effective tool that has been widely the feedstock supply (MOE, 2009). implemented for land suitability evaluation (Saaty, 1980; Reduce Exposure to Reduce Risk 156

of the matrix A. Then, calculate the factor weights ( wi ) using equation (3). wi  1  m  xij  (3) n   n   j1 i1xij   Table 1 Membership functions of linguistic scale for AHP (Saaty, 1980). Figure 1: Geographical location of study areas at Nakhon Definition Intensity of Ratchasima Province, Thailand importance Equal importance 3. Analytic Hierarchy Process (AHP) approach Equal to moderate importance 1 Moderate importance 2 The purpose of the AHP is to express the importance of Moderate to strong importance 3 each factor relative to other factors. The procedure of the Strong importance 4 AHP for assigning priority weights of the various criteria Strong to very strong importance 5 consists of three major steps (Saaty, 1980): Very strong importance 6 Very to extremely strong importance 7 Step1: Construct the PCM among the entire factors. Let n Extreme importance 8 be the number of the factors and aij be the pairwise 9 comparison value of dimension i factor to j for all factors i,j ≡{1,2,….n} which are given by one expert. Then, Step 3: Estimate the consistency ratio (CR) to ensure that assign the AHP linguistic scale (Table 1) to each aij of the judgments of experts are consistent. Let n be the the matrix A = (aij)nxm as shown in equation (1). number of the factors and max be the average value of 1 a12  a1 j  a1n  1 a12 1  a2 j  a2n  the consistency vector (CV). Then, calculate the  CV, max , and Consistency Index (CI) as shown in the following equations (4), (5), and (6) respectively: A     1    1   m  m  xij    1 a1 j 1 a2 j  aij  n 1   n    ain  (1) CV      j1 i1xij  xij  (4)  wi        1    j1   1 a1n 1 a2n  1 ain  1   Then, use a geometric mean technique to define the max  CV (5) geometric mean of each aij for the final matrix A. Let k n be the amount of experts and aijb be the pairwise comparison value of dimension i factor to j given by CI  max  n (6) expert b, where b = 1, 2,…, k and i,j ≡{1,2,….n}. After n 1 that, calculate the final matrix A as shown in equation Let RI be the random inconsistency indices, RI depends (2), where xij is a geometric mean of the AHP on the number of the factors (n) according to Saaty (1980). After that, CRs are calculated in the following comparison value of dimension i factor to j for all equation (7). If the CR is less than 0.1, the judgment is consistent. Therefore, the derived weights can be used. experts, where i,j ≡{1,2,….n}. CI (7)  xij  aij1  aij2  aijb ... aijk 1 k (2) CR  RI Step 2: Compute the factor weights. Let n and m be the 4. Materials and Methodology number of the factors, xij be a geometric mean of the 4.1 Conceptual frameworks AHP comparison value of dimension i factor to j for The methodology is divided into two phases (Figure 2): n Phase A: a comparison of the conventional maximum all experts, where i,j ≡{1,2,….n}, i1xij be the sum of limitation and the AHP approaches based on physical land suitability evaluation. Phase B: GIS-integrated AHP m column j of the matrix A, and  be the sum of row i j1 Reduce Exposure to Reduce Risk 157

approach based on socio-economic land suitability distance map for each socio-economic factor were evaluation. standardized or normalized by dividing by the maximum value for each map using “Map Calculation” operation in Phase A: Phase B: GIS as following in equation (8) (Ma et al., 2005): Physical factor Socio-economic factor Identify a set of factor Identify a set of factor STDDistance   Distance - Max Distance   1 (8) for the biofuel crop and constraint for the  Max Distance  biofuel crop Maximum AHP Then, all socio-economic factor maps were reclassified limitation Buffer zones for into four suitability classes following the FAO all factors framework as follows: S1, S2, S3 and N based on a higher value for a cell to suggest higher suitability. Then Linear Combination AHP assign the score 1, 0.8, 0.5, and 0 for S1, S2, S3, and N Method respectively into each classified socio-economic factor map. Maximum AHP Linear 4.2.3 Ground truth map limitation map Combination map Erased by constraint map Reference Map A ground truth map is to combine the existing cassava Socio-economic land planted area and an average cassava yield statistics data Physical land suitability map for the in Market Year (MY) 2008/09 of the study area. The suitability map for the existing cassava planted area was obtained from biofuel crop classification of THEOS and Landsat 5 TM images biofuel crop which were done following Phongaksorn, et al. (2012). Land suitability map Then, the ground truth map was classified into four for the biofuel crop suitability classes as follows: S1, S2, S3, and N based on the biofuel crop strategy of Thailand (MOAC, 2011) as Figure 2: Framework of methodology shown in Table 3. Finally, the ground truth map obtained was compared with the physical land suitability map 4.2 Map data based on the conventional maximum limitation and AHP approaches for cassava in section 5.1.4. Map data were collected along with a set of the identified physical, socio-economic factors, one constraint as a Table 3 Classification of average cassava yield forest zone, and a reference map for the biofuel crops at Nakhon Ratchasima Province. The map projection is Class Average yield (ton/ha) WGS 84, zone 47, and scale 1:50,000. Highly suitable (S1) >31 4.2.1 Physical factor map Moderately suitable (S2) 20.67 - 31 Eight physical factors were identified based on the manual of qualitative land evaluations for economic plant Marginally suitable (S3) 10.33 - 20.67 of Thailand following the FAO framework (Tansiri and Saifak, 1999). Then, all physical factor maps were Not suitable (N) 0.00 - 10.33 reclassified for assigning the score into four suitability classes as follows: S1 (highly suitable), S2 (moderately 4.3 Data analysis suitable), S3 (marginally suitable) and N (not suitable), according to the cassava requirements as shown in Table 2. 4.3.1 Phase A: Physical land suitability evaluation 4.2.2 Socio-economic factor map Eight physical factors were considered in both the classical maximum limitation and AHP approaches for Four socio-economic factors maps were represented by biofuel crop suitability evaluation as follows: mean distance from suitable biofuel crop plantations to roads, temperature (F1), annual rainfall (F2), soil drainage (F3), water resources, markets, and labors. These factors were nutrient status (F4), soil depth (F5), reaction (F6), soil identified under the maximize socio-economic benefits; texture (F7), and slope (F8). Physical land suitability the physical suitability area for biofuel crop plantation evaluation for biofuel crop was identified using a linear closer to their socio-economic factor feature. A distance combination method. map for each socio-economic factor was developed using “Find Distance” operation in GIS. These maps are in a I. Classical maximum limitation approach raster format with a 15 meter cell size. The obtained The conventional maximum limitation method, the land Reduce Exposure to Reduce Risk suitability is defined according to the most severe limitation of land qualities and the land index is calculated by multiplying the rating scores of each factor as shown in Table 2. 158

Table 2 Land use requirements for cassava (Tansiri and Saifak, 1999). Land use requirement Factor rating classes and scores Factor Unit S1 S2 S3 N 1 Mean temperature °c 0.8 0.5 0 Annual rainfall mm 25-29 30-32 33-35 >35 Soil drainage class 24-14 13-10 <10 Nutrient status class 1200-1500 1500-2500 2500-4000 >4000 Soil depth cm. Well to excessively 900-1200 500-900 <500 Reaction pH Moderately well Very to Somewhat Soil texture class drained drained - poorly drained Moderate to very Low - - high >100 50-100 25-50 <25 7.4-7.8 7.9-8.4 >8.4 6.1-7.3 5.1-6.0 4.0-5.0 <4.0 Loam Silty clay Clay Sand >20 Slope % 0-5 5-12 12-20 II. AHP approach Step 1: Construct the PCM of the eight physical factors water resources (C2), distance to markets (C3), and by assigning the AHP linguistic scale (Table 1) to each distance to labors (C4). The PCM of the socio-economic aij of the matrix A = (aij)nxm as shown in equation (1). factors for cassava was assigned as shown in Table 6. Then, use a geometric mean technique, as shown in Socio-economic land suitability evaluation for biofuel equation (2), to define the geometric mean of each aij for crop was identified using the linear combination method. the final matrix A for cassava (Table 4). Step 2: Compute the priority weights (W) for cassava 5. Results and Discussions (Table 4), as shown in equation (3). Step 3: Estimate the consistency ratio (CR), as shown in 5.3.1 Phase A: Comparison of the classical maximum equations (4)-(7). limitation and AHP approaches for the physical land suitability evaluation III. Linear combination method The matrix A (Table 4) indicated that the soil drainage Let n be the number of the factors, Wi be the weight value was the most important factor since a root crop as of ith factor, and Si be the score of ith factor. Then, the cassava needed to grow in well soil drainage for root value of land suitability (L) is defined as shown in decay protection. The soil drainage was rated between moderate to strong and strong importance (4-5) more equation (9): than the mean temperature. The soil drainage was also rated between equal to moderate and moderate L    nWi Si (9) importance (2-3) more than the annual rainfall, soil i1 texture, and slope. In addition, the soil drainage was rated between equal and equal to moderate importance (1-2) For example, the value of land suitability for cassava more than the remaining factors. The nutrient status was based on the AHP approach (L) as shown in the equation considered as the next most important factor because the below: nutrient status was a common factor in high productivity. The soil depth was the third important one since it was a L  0.0411 0.117  0.8 0.218 0.5 0.206  0 basic variable in a root crop. The last important factor 0.132 1 0.111 0.8 0.100  0.5 0.076  0.8 was the mean temperature because there was no much difference in the mean temperature of the study area (27- Then, these L values of the AHP approach were 28°C). reclassified into four suitability classes based on an equal interval as follows: S1 = 0.75-1, S2 = 0.50-0.75, S3 = 0.25- Physical land suitability map based on the conventional 0.50, N = 0-0.25. maximum limitation and AHP approaches (Figure 3 (b) and (c)) were compared with the ground truth map 4.3.2 Phase B: Socio-economic land suitability (Figure 3 (a)). The overall accuracy of the classical evaluation maximum limitation and AHP approaches for cassava was 5.67% and 64% respectively (Table 5). The results The four socio-economic factors were considered in the were indicative of the AHP approach that was capable of AHP approach for biofuel crop suitability evaluation as classifying and mapping the physical land suitability for follows: as follows: distance to roads (C1), distance to cassava with the 58.33% higher overall accuracy than the conventional maximum limitation approach. Reduce Exposure to Reduce Risk 159

Table 4 The matrix A and weights W of physical factors based on the AHP approach for cassava. F1 F2 F3 F4 F5 F6 F7 F8 W 0.041 F1 1 1/3-1/4 1/4-1/5 1/4-1/5 1/3-1/4 1/2-1/3 1/2-1/3 1-1/2 0.117 0.218 F2 3-4 1 1/2-1/3 1-1/2 1-1/2 1-2 1-2 1-2 0.206 0.132 F3 4-5 2-3 1 1-2 1-2 1-2 2-3 2-3 0.111 0.100 F4 4-5 1-2 1-1/2 1 2-3 2-3 1-2 2-3 0.076 F5 3-4 1-2 1-1/2 1/2-1/3 1 1-2 1-2 1-2 F6 2-3 1-1/2 1-1/2 1/2-1/3 1-1/2 1 1-2 2-3 F7 2-3 1-1/2 1/2-1/3 1-1/2 1-1/2 1-1/2 1 1-2 F8 1-2 1-1/2 1/2-1/3 1/2-1/3 1-1/2 1/2-1/3 1-1/2 1 Figure 3: Physical land suitability maps based on the maximum limitation and AHP approaches for cassava at Nakhon Ratchasima Province: (a) Ground truth map. (b) Classical maximum limitation map. (c) AHP map. Table 5 Error matrix of suitability classification accuracy assessment for maximum limitation and AHP maps of cassava at Nakhon Ratchasima Province. Classic maximum limitation approach Overall Accuracy = 5.67% Ground truth map Maximum S1 S2 S3 N Total limitation map S1 - - -- - (a) S2 163,165 20,335 - - 183,500 S3 29,174 2,265 - - 31,439 N 134,272 9,570 - - 143,842 Total 326,611 32,170 - - 358,781 AHP approach: Overall Accuracy = 64% Ground truth map S1 S2 S N Total 3 AHP map S1 211,42 103,42 - 11,76 326,61 4 7 12 S2 27,205 3,637 - 1,327 32,169 S3 - - -- - N- - -- - (b) Tota 238,62 107,06 13,08 358,78 71 l9 4 - (c) 5.2.2 Phase B: Socio-economic land suitability Reduce Exposure to Reduce Risk evaluation As a positive result of the AHP in the previous part, the AHP approach is the standard for the socio-economic analysis. Socio-economic land suitability map based on the AHP approach for cassava (Figure 4) was created from the four socio-economic factors. The matrix  (Table 6) indicated that the distance to markets was the most important factor because the factor was to save product transportation costs. Therefore, the distance to markets was rated between moderate and moderate to strong importance (3-4) more than the distance to roads. The distance to markets was also rated 160

between equal and equal to moderate importance (1-2) reported that the ethanol plant capacity was 112.2 million more than the distance to water resources. In addition, liters and the ethanol consumption was 8.61 million liters the distance to markets was rated between equal to for Nakhon Ratchasima Province in 2009. moderate importance and moderate importance (2-3) more than the distance to labors. The distance to labors was considered as the next most important factor. The distance to water resources was the third important one since cassava was tolerates drought. The last important factor was the distance to roads because there were a lot of roads in the province resulting in easy access to the product transportation. Table 6 The matrix  and weights Ŵ of socio-economic factors based on the AHP approach for cassava. C1 C2 C3 C4 Ŵ C1 1 1/2-1 1/3-1/4 1/2-1 0.126 C2 1-2 1 1/2-1 1/2-1 0.223 C3 3-4 1-2 1 2-3 0.414 C4 1-2 1-2 1/2-1/3 1 0.237 Figure 5: Land suitability map based on AHP approach for cassava at Nakhon Ratchasima Province Table 7 Potential ethanol capacity for cassava at Nakhon Ratchasima Province Suitability Area Potential Potential Classification* (ha) Cassava Ethanol Production* Capacity** (million (million tons/year) liters/year) Highly suitable (S1) 309,166 >9.58 >1,597 Moderately suitable (S2) 37,607 0.97 162 Marginally suitable (S3) 670 0.01 2 Not suitability (N) 11,337 0.06 10 Total 358,781 >10.62 >1,771 Figure 4: Socio-economic land suitability map based on Note: * the median of Table 3 AHP approach for cassava at Nakhon Ratchasima ** 6 kg of cassava fresh roots can be fermented Province. to 1 liter ethanol (Silalertruksa and Gheewala, 2010) 5.2.3 Land suitability evaluation for the biofuel crop 6. Conclusions The suitable areas for the biofuel crop could ideally be relatively high productivity, maximizing the socio- The study presents the integrated GIS and AHP approach economic benefits and the nature conservation under the for the land suitability evaluation for the biofuel crop in current policy of the Government to control agricultural Thailand, namely cassava, at Nakhon Ratchasima land expansion for biofuel feedstocks. In consequence, a Province. It confirms that the AHP approach is a suitable land suitability map based on the AHP approach for approach for the land suitability evaluation for the cassava (Figure 5) was conducted from each of the biofuel crop. The outcome of the study has established physical land suitability map for cassava, the socio- that: economic land suitability map for cassava, the forest map, and the existing cassava map which was extracted (a) Performance of the conventional maximum from the ground truth map. limitation and AHP approaches in classifying the physical land suitability classes was The result indicated that the suitable areas for cassava different. The AHP approach performed better. were 358,781 ha. The potential cassava production was It was found that Nakhon Ratchasima Province more than 10.62 million tons/year that could be had highly physical land suitability for cassava fermented the potential ethanol capacity for more than plantation. 1,771 million liters/year (Table 7). The potential of ethanol capacity for cassava performance has matched (b) Nakhon Ratchasima Province had large socio- with the ethanol plant capacity and ethanol consumption economic land suitability for cassava plantation. of each province. The Ministry of Energy (MOE, 2012) (c) Nakhon Ratchasima Province had land Reduce Exposure to Reduce Risk suitability for cassava plantation. Their suitable 161

area for cassava plantation has performed the Silalertruksa, T. and Gheewala, S. H., 2010, Security of expected ethanol capacity matching with the feedstocks supply for future bio-ethanol production ethanol plant capacity and the ethanol in Thailand, Energy Policy, 38, 7476-7486. consumption of the province. Sorda, G., Banse, M. And Kemfert, C., 2011, An Further studies are needed for the other decision making overview of biofuel policies across the world, approach such as Fuzzy AHP, neural networks, Genetic Energy Policy, 38, 6977-6988. Algorithms (GA), and Cellular Automata (CA), could be included in the methodology to ensure more combined Tunsiri, B. and Saifuk, K., 1999, Manual of qualitative land evaluations for economic plant. (Bangkok: and/or comparative studies. The factor for suitable site Land Development Department, Ministry of agriculture and Cooperatives). for cassava cultivation would be centered on more socio- economic factors such as gross margin, discounted cash Van Ranst, E., Tang, H., Groenemans, G., Sinthurahat, flow analysis such as Benefit/Cost ratio (B/C ratio), Net S., 1996, Application of fuzzy logic to land Present Value (NPV), and Internal Rate of Return (IRR), suitability for rubber production in peninsular to help increasing accurate of evaluate suitable sites. Thailand, Geodema, 70, l-19 ACKNOWLEDGEMENTS This research is funded by the Agricultural Research Development Agency, Thailand. REFERENCE Bell, R. D., Silalertruksa, T., Gheewala, H. S. and Kamens, R., 2011, The net cost of biofuels in Thailand - An economic analysis, Energy Policy, 32, 834-843. Food and Agriculture Organization of the United Nations (FAO), 1976, A framework for land evaluation, Soils bulletin, No. 32, (Rome: FAO). Fischer, G., Prieler, S., Velthuizen, H., Berndes, G., Faaij, A., Londo, M., et al., 2010, Biofuels production potentials in Europe: Sustainable use of cultivated land and pastures, Part II: Land use scenarios, Biomass and Bioenergy, 34, 173-187. Ma, J., Scotta, N. R., DeGloriab, S. D. And Lembob, A. J., 2005, Siting analysis of farm-based centralized anaerobic digester systems for distributed generation using GIS, Biomass and Bioenergy, 28, 591-600. Malczewski, J. (2004). GIS-based land-use suitability analysis_a critical overview, Progress in Planning, 62, 3-65. Ministry of Agriculture and Cooperatives (MOAC), 2011, Strategy of Energy Crops, (Bangkok: Ministry of Agriculture and Cooperatives). Ministry of Energy (MOE), 2012, Energy Consumption Statistics, http://www.energy.go.th/?q=node/315. Office of Agricultral Economics (OAE), 2009, Agricultural Statistics of Thailand 2009, http://www.oae.go.th/download/download_journal/y earbook2552.pdf Phongaksorn, N., Tripathi, N. K., Kumar S. and Soni, P., 2012, Inter-Sensor Comparison between THEOS and Landsat 5 TM Data in a Study of Two Crops Related to Biofuel in Thailand, Remote Sensing, 4, 354-376. Saaty, T. L., 1980, The analytic hierarchy process, (New York: McGraw-Hill). Samranpong, C., Ekasingh, B. and Ekasingh, M., 2009, Economic land evaluation for agricultural resource management in Northern Thailand, Environmental Modelling & Software, 24, 1381-1390. Reduce Exposure to Reduce Risk 162

SPATIAL VARIABILITY OF CLIMATIC WATER BALANCE UNDER RAINFED RICE BASED ON AGRO-ECOLOGICAL ZONES IN KURUNEGALA DISTRICT WMUK Rathnayake, DN Sirisena Rice Research and Development Institute, Batalagoda. E.mail: [email protected] ABSTRACT Kurunegala district which is one of the major rice growing districts in Sri Lanka covers nine agro-ecological regions (DL1b, IL3, IL1b, IL1a, IM3b, WM3b, WM3a, WL2b, WL3) implying vast climatic differences within the district. Larger extent of paddy lands in the district is cultivated under rainfed conditions. Rainfall is perhaps the most limiting factor under rainfed condition where the temperature is in the suitable range. In low land rainfed rice fields, water gain from rainfall and water losses largely result from transpiration and evaporation. Water lost through seepage and percolation is insignificant under rainfed conditions. Therefore, under rainfed condition, water balance could be expressed by reducing potential evapo-transpiration from the actual rainfall, which implies the climatic water balance. It gives a rough estimate of water shortage and surplus. Therefore, this study was conducted to study the spatial variability of climatic water balance during the cropping season of 3 month rice variety in wet (maha) and dry (yala) seasons in Kurunegala district. Actual rainfall and potential evapo-transpiration were calculated on monthly basis of five years data from 2002 to 2007 for the nine agro-ecological zones. Climatic water balance was determined for dry season from May to July and for wet season from November to January and calculated for the whole cropping period of three month age as well as for different growth phases (vegetative, reproductive and maturity phases). Maps were produced in Geographic Information System capabilities. When total rainfall in wet season is considered, it is sufficient for whole cropping period for all agro-ecological zones. If distribution is taken into account in three cropping phases, water is surplus in all agro ecological zones during the vegetative phase. In the reproductive pahse, surplus water appeared in DL1b, IL3, IM3b, WM3b regions only. During the maturity phase, water surplus was showed only in IM3b region. When total rainfall in the dry season was considered, agro-ecological zones of WL2b, WL3, WM3a showed surplus water during all three cropping phases. DL1b, IL3, IM3b regions showed water deficit during the vegetative phase while whole DL and IL regions showed water deficit during the reproductive phase. In the maturity phase, only WL2b, WM3a regions had surplus water. According to this results it could be concluded that although the total rain water is sufficient for the whole cropping duration, water shortages occurs during the different growing phases of rice. Therefore, both total amounts of rainfall as well as rainfall distribution pattern throughout the season should be considered rather than considering the total amount of rainfall for the whole season to get optimum productivity from rainfed rice cultivation. Key words: agro-ecological zones, climatic water balance, rainfed rice, spatial variability Reduce Exposure to Reduce Risk 163

INTRODUCTION balance was calculated by reducing potential evapo- transpiration from the actual rainfall on monthly basis for Kurunegala district covers nine agro-ecological regions two different seasons. For three month age group of rice, which belong to Dry Zone (DL1b), Intermediate Zone dry season was considered from May to July and wet (IL3, IL1b, IL1a, IM3b) and Wet Zone (WM3b, WM3a, season was considered from November to January. It WL2b, WL3) (Punyawardana et al, 2003). These regions was determined for the whole cropping period of three imply vast climatic differences within the district. month age as well as for different growth phases Almost all the area belongs to low country (less than 300 (vegetative, reproductive and maturity phases). One m in elevation) and the amount of rainfall received in the month is taken for maturity phase and one month is dry zone, intermediate zone and wet zone is less than assigned for reproductive phase while remaining days are 1750 mm, between 1750 mm and 2500 mm, and more assigned to vegetative phase. If the age class is three than 2500 mm respectively (Panabokke, 1996). Rice month, vegetative phase is one month. Therefore, in this cultivation is done in all nine agro-ecological zones and study, vegetative, reproductive and maturity phases were as such Kurunegala district is identified as one of the assigned to May, June, July months in the dry season and major rice growing districts in Sri Lanka. Rice November, December, January for the wet season cultivation is done twice a year as in the wet (maha) respectively. Maps were produced in Geographic season which lies from October to February and dry Information System capabilities. (yala) season which lies from March to September. Rice cultivation is done under major irrigation, minor Figure 1: Nine agro-ecological zones in Kurunegala irrigation and rainfed condition. Cultivated extent under major, minor, rainfed conditions in wet season is 17%, district 41%, and 42% respectively. Under dry season, it is 21%, 40% and 39% respectively. These statistics confirm that RESULTS AND DISCUSSION large extent of paddy lands in the district is cultivated under rainfed conditions. The average rice yield in The calculated values of climatic water balance for nine rainfed condition (in wet season as 3.5 t/ha and dry season as 2.9 t/ha) is lower than in minor irrigated agro-ecological zones in wet and dry seasons are shown condition (in wet season as 3.7 t/ha and dry season as 3.1 t/ha), and as well as major irrigated condition (in wet in Table 1. season as 4.4 t/ha and dry season as 3.6 t/ha) in the district. Since water is limiting, 56% percent of the Table1. Climatic water balance for nine agro-ecological rainfed area is cultivated in the dry season while that in the wet season is 89%. Rainfall is perhaps the most zones in wet and dry seasons limiting factor under rainfed condition where the temperatures are in suitable range (Yoshida, 1981). The Wet Season Dry Season amount and pattern of rainfall varies widely from one place to another and from year to year. In low land DL1b 84.7 - 311.6 rainfed rice fields, water gain from rainfall and water losses largely result from transpiration and evaporation. IL1a 15.6 - 65.6 Considering water gains and losses, water balance for these areas could be calculated. Under rainfed condition, IL1b 24.8 - 137.3 water balance could be expressed by reducing potential evapo-transpiration from the actual rainfall, which IL3 37.4 - 226.2 implies the climatic water balance (Yoshida, 1981). It gives a rough estimate of water shortage and surplus. IM3b 264.1 - 187.6 Therefore, this study was conducted to study the spatial variability of climatic water balance during the cropping WL2b 87.2 196.33 season of 3 month rice variety in wet (maha) and dry (yala) seasons in Kurunegala district. WL3 46.6 136.4 MATERIALS AND METHODS Distribution of nine agro-ecological zones in Kurunegala WM3a 98.3 85.2 district is shown Figure 1. Monthly rainfall data and evaporation data from 2002 to 2007 were collected from WM3b 238.4 - 8.1 the Department of Meteorology, Sri Lanka. The potential evapotranspitation was computed by multiplying pan Figure 2 shows the spatial variability of climatic water evaporation with 0.9 which was the pan factor of Class A balance within the district. When total rainfall in the wet pan (Munasinghe et al, 2004). Actual rainfall and season is considered, it is sufficient for whole cropping potential evapotranspiration data were assigned to nine period for all agro-ecological zones. agro-ecological regions accordingly. Climatic water Reduce Exposure to Reduce Risk 164

Figure 2: Spatial variability of climatic water balance Figure 4: Spatial variability of climatic water balance in wet season (maha) within the district. during reproductive phase in wet season (maha) within the district When water distribution was considered in three cropping phases separately, water is surplus in all agro ecological zones during the vegetative phase (Figure 3). In the reproductive stage surplus water appeared in DL1b, IL3, IM3b, WM3b regions only (Figure 4). During the maturity, water surplus was showed only in IM3b region (Figure 5). Figure 3: Spatial variability of climatic water balance Figure 5: Spatial variability of climatic water balance during vegetative phase in wet season (maha) during maturity phase in wet season (maha) within the within the district When total rainfall in the dry season is taken into consideration, agro-ecological zones of WL2b, WL3, WM3a showed water surplus during three month cropping period (Figure 6). Reduce Exposure to Reduce Risk 165

Figure 6: Spatial variability of climatic water balance in Figure 8: Spatial variability of climatic water balance dry season (yala) within the district during reproductive phase in dry season (yala) within the district DL1b, IL3, IM3b regions showed water deficit during the vegetative phase (Figure 7) while whole DL and IL regions showed deficit of water during the reproductive phase (Figure 8). In the maturity phase, only WL2b, WM3a regions had surplus of water (Figure 9). Figure 9: Spatial variability of climatic water balance during maturity phase in dry season (yala) within the district Figure 7: Spatial variability of climatic water balance In this study, normal cropping season which farmers used during vegetative phase in dry season (yala) to cultivate and monthly climatic data were considered. within the district Results showed that total rainfall received were not enough during the three growth phases for many of the agro-ecological zones except in the vegetative phase during wet season. If climatic water balance was calculated weekly basis for different cropping periods, better information could be derived to find out the best period for sowing rice in rainfed cultivation. CONCLUSION According to this results it could be concluded that although the total rain water availability is sufficient for the whole cropping duration, water shortages occurs during the different growth phases of rice. Therefore, both total amounts of rainfall as well as rainfall distribution pattern throughout the season should be Reduce Exposure to Reduce Risk 166

considered rather than considering the total amount of 2.Natural Resources, Energy and Science Authority, rainfall for the whole season to get optimum productivity Colobmo. Sri Lanka from rainfed cultivation. Punyawardana, BVR, TMJ Bandara, MAK Munasinghe, NJ Banda and SMV Pushpakumara, 2003. Agro- REFERENCES ecological regions of Sri Lanka. Natural Resources Munasinghe, MAK and RD Chithranayana, 2004. Management Center, Department of Agriculture, Spatial-temporal variability of potential Peradeniya, Sri Lanka evapotranspiration in Sri lankaand its application in Yoshida S, 1981. Fundamentals of rice crop science. The agricultural planning. Journal of the Soil Science Society International Rice Research Institute, Philippines. Pp 94- of Sri Lanka. 110. Panabokk, CR, 1996. Soils and agro-ecological environments of Sri Lanka. Natural Resources Series-No Reduce Exposure to Reduce Risk 167

VARIABILITY OF SOIL PH, P, K AND ZN IN RICE BREEDING FIELDS AT RICE RESEARCH AND DEVELOPMENT INSTITUTE DN Sirisena*, WMUK Ratnayake, RMK Atapattu Rice Research and Development Institute, Batalagoda, Ibbagamuwa* [email protected] ABSTRACT Rice variety development is the major mandate given to Rice Research and Development Institute. After hybridization, second generation (F2) seed onwards are grown in fields. Plant characters and yield performances which are considered as important in selection and depend on the soil fertility. Therefore, uniformity in soil fertility in the paddy filed is paramount important in the selection purpose or otherwise fertility variation may hide the performance of better lines. Therefore, this study was conducted to study the homogeneity of soils in the rice breeding paddy fields and to identify limiting factors. Ninety one soil sampling points and the field boundary were demarcated using Global Positioning System (GPS). Soils were collected up to 15 cm depth and analyzed for pH, P, K and Zn. Analyzed data of pH, P, K, Zn were used to prepare thematic maps using Geographical Information System (GIS). Variability of pH within the field is minimum and optimum for rice plants. Soil K levels of entire paddy fields are below 78 mg kg-1 and soil P and Zn is highly variable within the field. Majority of the field having soil P below 10 mg kg1 while significant extent of that having P above 24 mg kg-1 aswell. Soil Zn levels are above the critical levels but around 10% of the area having below the critical level of 1 mg kg-1. Due to spatial variability of K, P and Zn, better rice lines may not be performed well and selection of lines may be a problem in these fields unless site specific fertilizer management is practiced. Key Words: Soil fertility, Spatial Variability, Rice Breeding 1. INTRODUCTION Potassium (K) and Exchangeable Zinc (Zn) were Rice variety development is the major mandate given determined by atomic absorption spectrophotometer. to Rice Research and Development Institute at Batalagoda. Therefore, majority of paddy fields at Analyzed data of pH, available P, exchangeable K., Batalagoda is used for this purpose. After and Exchangeable Zn were used to prepare thematic hybridization, second generation (F2) seeds onwards maps in the Geographical Information System (GIS) are grown in this fields and selection is done by environment. Legends were assigned for each map looking at the plant characters. Above all characters, based on the critical level of each fertility parameter yield performance is considered as important in separately. selection purpose (Abeysiriwardena, Sandanayake. 2000). The yield performance of rice is depending 3. RESULTS AND DISCUSSION on the soil fertility and the quantity of fertilizer The total area selected was 7.3 ha. Distribution application especially N, P, K and Zn (Dobermann, pattern of soil sampling points in the selected paddy and Fairhurst. (2000). Therefore, uniform in soil field is shown in Figure 1. fertility in the breeding paddy filed is paramount important in the selection purpose or otherwise fertility variation may hide the performance of better lines. Therefore, objective of this study was to study the homogeneity of soils in breeding paddy fields and to identify the limiting factors. 2. MATERIALS AND METHOD Figure 1: Distribution of soil sampling points in the Boundary of the research field was demarcated by selected paddy field using Global Positioning System (GPS). Ninety one sampling points were selected from the entire breeding field and soil samples were collected from each sampling point up to 15 cm depth randomly Soil sampling points were also demarcated by using GPS. Soil samples were analyzed for pH using 1:5 H2O methods. Available soil Phosphorus (P) was determined by Olsen’s method. Exchangeable Reduce Exposure to Reduce Risk 168

Soil K levels were varied from 6.7 mg kg-1 to 87 mg Spatial variability of Zn is shown in Figure 4 and kg-1. Results of the Figure 2 shows that exchangeable soil K levels of the entire paddy fields accordingly it shows that majority of the paddy fields are below the critical level of 78 mg kg-1 (Bandara et are having Zn above the critical level of 1 mg kg-1 al, 2005). Therefore, application of potassium fertilizer is a must to all paddy fields to obtain an (Bandara et al, 2005). Therefore, application of Zn optimum performance of rice lines. fertilizer as done at present may also enhances the Available soil P is highly variable within the field variability within the field. (Figure 3). The minimum value of available P was 1.5 mg kg-1 and the maximum value was 32 mg kg1 Critical P level for optimum rice production is 10 mg kg-1 (Bandara et al, 2005) and majority of the area is having soil P levels below 10 mg kg1. Around ten percent of the area is having soil P levels more than 24 mg kg-1 P. Therefore, blanket application of P fertilizer as done at present increases the variability of soil P levels further and affect the performances of the plants. Figure 4: Spatial variabity of exchangeable Zn in reseach field pH of soil is an important criterion which determine the availability of most of the nutrients. Variability of pH within the field at RRDI is shown in the Figure 5. Accordingly pH is above the critical level as such it is not a problem for optimum performance of rice plants. Optimum level of pH for rice is in between 5.5 and 6.5 (Bandara et al, 2005) and values are in the range from 5.5 to 7.5. Figure 2: Spatial variabity of exchangeable K in reseach field Figure 4. Spatial variabity of exchangeable pH in reseach fields Figure 3: Spatial variabity of available P in reseach 4. CONCLUSION field Spatial variability of P and Zn is a problem for performance of rice plants in the breeding field. Reduce Exposure to Reduce Risk 169

Therefore, better rice lines may not be performed well and selection of rice plants may be a problem in these fields even if P and Zn fertilizer are applied at recommended rates. Therefore action should be taken to homogenous P and Zn in this fields before cultivation of breeding lines. REFERENCES  Bandara,W.M.J,D.KumaragamaD.B.Wickrama singhe and S.B.A.Weerawarna. 2005. Site specific nutrient management strategy to increase rice yields in Low Country Intermediate Zone. J. Soil Sci. Soc, of Sri Lanka 17: 32-43  Dobermann, A. and T. Fairhurst. (2000). Plant sampling for diagnosis of nutritional disorders. Nutrient Disorders & Nutrient management. International Rice Research Institute. Pp 175  Abeysiriwardena, D. S. de Z. and S. Sandanayake. 2000. Future rice research as directed by trends in cultivated extent and yield of rice during the recent past. Proceedings of the Annual Symposium of the Department of Agriculture. 2: 371-381, Reduce Exposure to Reduce Risk 170

IMPORTANCE OF INCORPORATING VEGETATION INTO URBAN INFRASTRUCTURE TO REDUCE HEAT ISLAND EFFECT: A STUDY IN COLOMBO CITY OF SRI LANKA N.L. Ukwattage& N.D.K. Dayawansa Department of Agricultural Engineering, Faculty of Agriculture, University of Peradeniya, Sri Lanka Abstract: The trend of increasing temperature in urban core areas or the formation of urban heat islands has been experienced in many cities around the world. Thermal discomfort and increased energy consumption have been identified as the most critical impacts of this heat island formation. This temperature development is mainly attributed to the conversion of soft land cover types such as vegetation to hard land cover types such as building, roads etc. This study was carried out in Colombo city, the commercial capital of Sri Lanka with remotely sensed data to assess the impact of land cover on land surface temperature development and to identify the importance of incorporating vegetation into urban environment to reduce temperature development. Landsat TM and ETM+ images acquired on 14 March 2001 and 13 February 2005, respectively were used in the study. Land use maps of two dates were prepared with supervised classification of the imagery. The indices namely Normalized Difference Vegetation Index (NDVI), Normalized Difference Built-up Index (NDBI) and Normalized Difference Water Index (NDWI) were also derived to support the identification of vegetation, built up areas and water. Images of surface temperature were prepared with the thermal bands of the Landsat TM and ETM+ images and subsequently the heat islands were identified in both images. Attempts were made to explore relationships between land cover types and land surface temperature. Results confirmed that the Colombo city is an area dominated by built-up. From 2001 to 2005, the built-up areas have been increased from 66% to 73% of the total land extent. Colombo Fort, Kochchikade, Aluthkade, Kotahena areas were identified as the areas having low built-up density and high temperature. Also majority of the heat islands are concentrated in this area. A negative linear relationship was obtained between NDVI and Land Surface Temperature (LST) while a positive linear relationship was obtained between the NDBI and LST with a correlation coefficient of -0.75144 and +0.575798, respectively. The usefulness of the vegetation to reduce temperature development was identified using two areas with similar extent of built-up and different extent of vegetation cover. This study demonstrated the usefulness of incorporating vegetation into urban infrastructure to reduce the impact of heat development. It will eventually help to reduce the thermal discomfort and energy demand for cooling in urban areas. KEYWORDS: Landsat satellite data, heat islands, NDVI, Colombo Reduce Exposure to Reduce Risk 171

1. INTRODUCTION using satellite derived land surface temperature (LST) measurements have been conducted using Urbanization; the conversion of other types of lands various remote sensing data such as NOAA to uses associated with growth of population and AVHRR with 1.1 km spatial resolutions, ASTER economy is a main type of land use and land cover with 90 m spatial resolution, Landsat Thematic change that has occurred in the human history. As a Mapper (TM) and Enhanced Thematic Mapper Plus result of urban expansion, soft land cover such as (ETM+) thermal infrared (TIR) data with 120m and vegetation and water areas are removed and 60m spatial resolutions, respectively (Zhao-ming et replaced by hard land cover surfaces like buildings, al., undated). roads and paved areas. This alteration will inevitably result in the redistribution of incoming Also, since remote sensing is widely used to solar radiation and induce the urban rural contrast in monitor land cover types and to estimate surface radiation and air temperature. Therefore, biophysical characteristics of land surface; it excess warmth is experienced in urban atmosphere provides the chance to link land surface temperature and surfaces compared to none urbanized rural information easily with land cover type (Sarkar, surroundings. This phenomenon is called as Urban 2004). Heat Island (UHI) formation (Shahmohmadi et al., 2011; maloley, 2010). 1.1 Problem Statement Ill effects caused by urban heat island formation are Heat islands formation in urban areas occurs both on experienced in most of the urbanized cities in many the urban surfaces like buildings, bare lands, roads countries in the world. The developed cities in Sri and water bodies and in the urban atmosphere Lanka including Colombo, Excess warmth during (Caselles et al., 1997). Surface urban heat islands day and night time has become a serious problem are formed on urban surfaces. which adversely affect the quality of life. Elevated temperatures from urban heat islands can With its urbanization, land cover types of Colombo affect the urban environment and quality of life in city has changed over the time. Soft land cover many negative ways. Thermal discomfort and areas have been reduced and hard land cover areas increased energy consumption have been identified have been increased, affecting the bioclimatic as the most critical impacts of this heat island conditions of the city (Emmanuel, 2004). It has formation. brought about an increasing trend in thermal discomfort which correlates well with hard land Mitigation of heat island effect has become a topic cover changes (www.questia.com). Colombo city of interest, due to those negative impacts on quality area is now experiencing high level of discomfort of life. Although, all the major causes of urban heat from heat during all the months of the year except in island formation are associated with the December as a result of its environmental change urbanization, it is not possible to stop this process as (www.bbc.co.uk). urbanization is part of the development process of a country. As a result, other alternatives have been Increment of temperature has increased the energy found to mitigate urban heat island effect. demand especially in the form of electricity as most of the temperature regulating measures are Many communities are taking actions to reduce electrically driven (e.g.: Air conditioners, fans). It urban heat islands using several strategies. Out of also might be contributed to record the country’s these strategies, improvement of vegetative cover of highest electricity consumption from the city of the area is considered as the most sustainable Colombo (www.sundaytimes.lk). Therefore, measure (Bretz et al., 1994). This requires planned necessity of controlling urban heat island effect has urbanization with integration of green component in become more and more important as it is a way of to city plans as it provides the opportunity of saving energy. controlling temperature development in urban areas. Urban heat island studies are generally conducted in 1.2 Objectives two ways; measuring the UHI in air temperature The main objective of this study was to explore the through the use of automobile transects and weather relationship between land cover type and land station networks, and measuring the UHI in surface surface temperature in Colombo city. It also temperature through the use of airborne or satellite remote sensing. Studies on the UHI phenomenon Reduce Exposure to Reduce Risk 172

attempted to identify the importance of vegetation a. Conversion of thermal band digital cover to reduce the temperature built up. numbers into spectral radiance L (Wm- 2Sr-1µm-1) 2. METHODOLOGY The Landsat TM and ETM+ thermal bands have 2.1 The Study Area different gain and offset values. Therefore, for the The selected study area is the Colombo city of Sri calculation of their radiance values different Lanka which is located in the western province and formulas are used. For the calculation of the lies within Low Country Wet zone. Mean annual radiance values of the TM data, the Equation 3 was ambient temperature of the area is about 290C used. showing little variation among years. Rainfall in the city averages around 2450 mm in a year. It covers Radiance = DN*Gain + Offset an area of approximately 37.31 km2 and population ………………………Equation 3 is around 653,000 (www.cmb.ac.lk). The high rate Where, of urbanization, concentration of secondary and Gain = 0.0551582 tertiary economic activities, large number of vehicle Offset = 1.2377996 population and fuel consumption, artificial surfaces and low vegetation cover are some of the main For the calculation of the radiance of ETM+ data, characteristics of this area. These characteristics Equation 4 was used: have lead to a trend of increasing temperature of the city in recent past (Liyanage and Manawadu, 2008). Radiance = ((LMAX-LMIN) / (DNMAX- DNMIN)) * (DN- DNMIN) + LMIN …Equation 4 2.2 Materials and methods Where for band 61, Medium resolution multi temporal satellite images LMAX = 17.04 LMIN = 0.0 from Landsat ETM+ and TM sensors were used in DNMAX=255 DNMIN =1 the study to extract land surface temperature and land cover information. b. Conversion of spectral radiance (L) into at-sensor brightness temperature BT 2.2.1 Extraction of land use/ cover (K) information: Major land cover types of Colombo city area were identified using land use map of the BT  K2 area and supervised classification was employed to extract the areas under each of these land use/ cover ln  K1  1 type.   L Information on major land cover types as vegetation ……………………………… Equation 5 and built-up was extracted by deriving Normalized Difference Vegetation Index (NDVI) and BT= sensor brightness temperature Normalized Difference Built up Index (NDBI) models for each respectively according to the K1 is 607.76 (for TM) or 666.09 (for ETM+) Wm−2 Equations 1 & 2. sr−1 μm−1 NDVI= NIR-Red/ NIR + Red……….Equation 1 K2 is 1260.56 (for TM) or 1282.71 (for ETM+) NDBI= MIR-NIR/ MIR+NIR………Equation 2 c. Determination of surface emissivity 2.2.2 Determination of land surface Surface emissivity values for different land cover temperature and identification of heat island types were obtained based on their NDVI as given areas: Digital numbers of thermal infrared bands in Table 1. were processed using different mathematical models Table 1: NDVI and corresponding surface to extract heat related properties of the image area at emissivity values the time of image acquisition. NDVI Value Surface Emissivity Between-1 and -0.18 0.985 0.955 Between -0.18 and - 0.157 1.0094 + Ln (NDVI) Between 0.157 and 0.727 0.99 Between 0.727 and 1 Reduce Exposure to Reduce Risk 173

d. Land surface temperature (Ts) calculation land use/ cover types in Colombo city during the years 2001 and 2005. Ts  BT Table 2: Distribution of land use/ cover in Colombo city in 2001 and 2005 1   .BT .ln      Type of land use % Area … ……………. Equation 6 Year Year 2001 2005 λ= wave length of emitted radiance 11.5 μm Built-up area 66% 74% ρ= hc/σ = 1.438* 10-2 mK Vegetated area 32% 25% Water area 2% 1% e. Identification of urban heat islands Vegetated area and water area have been reduced Derived land surface temperature values were whereas built-up areas have been increased. normalized using Equation 7 to obtain normalized temperature (T) values. NDVI and NDBI surfaces of any given image reflects the strength of vegetation cover and built-up T  T  Tmin area. These index values very between +1 and -1 Tmax  Tmin scale and +1 represents high density and -1 represents very low or no density of the respective ……………………….…………Equation 7 cover class. T is a value in between 0 and 1.This value is used to discriminate heat islands from other areas. According to two NDVI surfaces highest NDVI having T ≥ 0.6 are considered as urban heat islands. values are associated with Colombo eastern and Based on the above criteria, urban heat islands were southern areas which consist of more home gardens. separated and mapped. Cinnamon garden area, Narahenpita and Thimbrigasyaya area produced NDVI values f. Data interpretation and derivation of between 0.1 to 0.4 showing relatively high density relationships of vegetation. Kochchikade, Fort, New Bazzar and Relationships between land surface Aluthkade areas show the lowest NDVI values (-0.1 to -0.3) as they have closely built-up areas. Heavily temperature and land cover type (vegetation, built- built up areas which show highest NDBI values are up area) were identified using several samples of coincide with the areas having lowest vegetation urban heat islands and other areas using stratified cover in both years. random sampling. Regression analysis and correlation coefficient 3.2 Changes in land use/ land cover during 2001- were used to explore strength of relationships of 2005 NDVI and NDBI with Land surface temperature. The problem in difference in the resolutions Changes occurred in land cover of Colombo city between thermal band and other bands was avoided were detected using two Landsat images and the by resampling thermal band into 30m pixels. results are briefed in this section. The NDVI and NDBI variations in few randomly selected areas for 3. RESULTS AND DISCUSSION 2001 and 2005 are presented in Figure 2. 3.1 Status of land use/ cover in Colombo According to the Figure 2 (a) NDVI of the area has reduced in 2005 compared to 2001 in the same area. According to the study, built-up has been remained Many reasons can be attributed to these changes as the major land cover type in Colombo city during including the rainfall received prior to image both years of study. Distribution pattern of different acquisition, date of image acquisition and the land uses have been remained more or less similar changes that have occurred in vegetation. Since the during the period. Colombo fort, Kochchikade, images are acquired almost during the same period Kotahena and Aluthkade areas can be identified as of the two years, it can be suggested that there are heavily built-up areas and Cinnamon garden, some changes in vegetation which has reduced the Narahenpita, Thimbirigasyaya and Kirillopone areas vegetation vigour. This is also confirmed by the are identified as high vegetated areas in the city results of the NDBI surfaces. When comparing (a) during both years. Table 2 illustrates the major three Reduce Exposure to Reduce Risk 174

and (b) it is seen that strength of built up areas has LST in C 45 also increased slightly in 2005. 40 35 3.3 Impact of land cover on land surface 30 temperature 25 Relationship between land cover and land surface 20 temperature was assessed using samples selected by stratified random sampling from heat island areas y = -11.97x + 30.95 and other areas. Relationships were explored using 15 R² = 0.780 regression and correlation analysis. 10 5 0 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 NDVI Figure 3: Relationship between NDVI and LST NDVI 0.6 Negative linear relationship was found between NDVI 2001 NDVI and LST in the regression analysis with a correlation coefficient of -0.75144. Accordingly, 0.4 NDVI 2005 temperature built up is high in low vegetative areas 0.2 and is low in highly vegetative areas. Vegetation can cool the surrounding area by absorbing heat as 0 latent heat of evaporation. The study shows that -0.2 areas with heat islands have very low NDVI values -0.4 (-0.5 to -0.2). -0.6 Relationship between NDBI (which is an indication of the density of buildings) and LST is given in Pixels in the sample Figure 4. (a) Changes in NDVI over time 45 0.5 LST in C 40 0.4 35 0.3 30 0.2 NDBI 2005 25 NDBI 2001 0.1 20 0NDBI 15 y = 9.353x + 31.14 R² = 0.746 -0.1 Pixels in the sample 10 (b) Changes in NDBI over time 5 Figure 2: Changes in land use/land cover over time 0 Pixel based comparison between land surface -1 temperature and strength of vegetation cover -0.8 extracted as NDVI is presented in Figure 3. -0.6 -0.4 -0.2 0 0.2 0.4 0.6 NDBI Figure 4: Relationship between NDBI and land surface temperature Regression analysis gave a positive linear relationship between NDBI and LST. Correlation coefficient was +0.575958. Accordingly when the built up area density increases, the LST also increases. Since built up areas lower the reflection of short wave radiation and have high heat capacity, LST tends to increase with the increasing building density. It leads to create heat islands in heavily built up areas consist of metal, concrete and asphalt. Similar results have been obtained by many other scientists who carried out similar research. Marzen and Wersinger, (2004) have got a correlation coefficient value ranges from -0.6556 to -0.8264 for the relationship between NDVI and LST. Also Reduce Exposure to Reduce Risk 175

Dwyer et al., (2004) have investigated a correlation c. Comparison of LST in two sites of – 0.780 between NDVI and LST. Figure 5: Impact of land use planning on land Impact of land use planning on temperature was surface temperature assessed by comparing the vegetation cover, built- up area and land surface temperature of two 4. CONCLUSIONS collected samples as given in the Figure 5. Sample 1 The study revealed that Colombo city is an area is referred to the unplanned urban area (from dominated by built-up areas. The extent of built-up Bambalapitiya) and sample 2 is referred to the has been increased from year 2001 to 2005 from properly planned area (from Cinnamon garden) in 66% to 74%, where as extent under vegetation and the city of Colombo. water has been reduced from 32% to 25 % and 2% According to the Figure 5, NDVI values of majority to 1% respectively. of the pixels in planned area lay in the range of 0.4 to -0.1 and in unplanned area in the range of -0.4 to Land surface temperature (LST) development is low -0.1. NDBI values of both samples lay in the range with the increasing vegetation cover and high with of 0.1 to 0.4 reflecting a similar built-up fraction in the increasing building density. A negative linear both areas. Land surface temperature of well relationship was established between strength of planned area averages around 300C whereas in vegetation cover (NDVI) and the LST and a positive unplanned area it averages around 340C. It allows linear relationship was established between the justifying that difference of vegetated fraction of building density and the LST in Colombo city for two areas has given rise to LST difference. the study dates. It was found that the vegetation Therefore, it also indicates the fact that proper land plays a considerable role in reducing temperature use planning in urban areas can reduce the even in areas with similar built up density. Hence, temperature development. The ambient temperature inclusion of vegetation component in urban values can be different due to sea breeze at the site 1 infrastructure planning can bring environmental and which is located closer to the sea. economic benefits. a. Comparison of NDVI in two sites REFERENCES b. Comparison of NDBI in two sites Bretz, S., Akbari, H., Rosenfeld, A. 1998. Practical issues for using solar-reflective material to mitigate urban heat islands, Atmospheric Environment, 32:1, 95-101. Caselles, V., Coll, C., & Valor, E. 1997. Land Surface Emissivity and Temperature Determination in the Whole HAPEX-Sahel Area from AVHRR data. International Journal of Remote Sensing, 18, 1009–1027 Emmanuel, R. 2004. Assessment of Impact of Land Cover Changes on Urban Bioclimate: The Case of Reduce Exposure to Reduce Risk 176

Colombo Sri Lanka. Architectural Science Review, 007_AL_Annual_Report.pdf, last accessed on 23rd 46, 131-157. February 2009 Manawadu, L., & Liyanage, N. 2008. Identifying Maloley, M.J. (2010). Thermal remote sensing of Surface Temperature Pattern of City of Colombo. urban heat island effects: Greater Toronto area, Engineer, 41, 133-140 Geological Survey of Canada, Open file NO. 6283, Natural resources Canada. 39p. Marzen, L. J., & Wersinger, J. M. 2004. Auburn University Environmental Institute Annual Sarkar, H. 2004. Study of Landcover and Technical Report, Population Density Influences on Urban Heat http://water.usgs.gov/wrri/AnnualReports/2007/FY2 Island in Tropical Cities by Using Remote Sensing Article ID 497524, 9 pages, 2011. and GIS: A Methodological Consideration, 3rd doi:10.1155/2011/497524 FIG Regional Conference, Jakarta, Indonesia, October 3-7, 2004 Zhao-ming, Z., Guo-jin, H., Rong- bo, X., Wei, W., & ZhiYun, O. (Undated). A Study on the Changes Shahmohamadi, P., Che-Ani, A. I., Maulud, K. N. of Urban Heat Islands in Beijing Based on Satellite A. Tawil, N. M. and Abdullah, N. A. G. (2011). Remote Sensing, available online at The Impact of Anthropogenic Heat on Formation of http://www.aars- Urban Heat Island and Energy Consumption acrs.org/acrs/proceeding/ACRS2005/Papers/D4- Balance, Urban Studies Research. vol. 2011, P13.pdf, last accessed on 06th June Reduce Exposure to Reduce Risk 177

Application of Geographic Information System for Tea Plantation Management as Decision supporting tool A Case study at St Coombs Estate, Talawakelle, Sri Lanka N.N.K.Wellala1, 2, J.Gunatilake2, H.W. Shyamale1 1 Agricultural Economics Division, Tea Research Institute, Talawakelle, Sri Lanka 2Postgraduate institute of science, University of Peradeniya, Peradeniya, Sri Lanka Abstract Tea (Camellia sinensis) is the most popular beverage in the world and Sri Lanka is one of the top global producers. Sri Lankan tea industry is now in a decisive situation where a need of solutions to prevail over challenges such as low productivity, high cost of production comparatively other main producing countries and declining competitiveness in the world tea market. Geo-Information Technologies (GIT) can be used effectively to find out solutions to above challenges. A case study was undertaken with the objective of creating a Geographic Information System (GIS) based Decision Supporting System (DSS) for the St Coombs estate, Talawakelle. High resolution satellite images, topographic maps (1:10,000) and field data were used to prepare a base data set. These base layers were integrated in order to create, a digital map for St. Coombs estate.By utilizing these maps and other socio – economic data, GIS based DSS was created. Finally some analysis was carried out to investigate the applicability of the DSS for day-to-day plantation management. Key words: Geo-Information Technologies, Decision Supporting System, Productivity, Cost of Production 1. Introduction from traditional methods by providing alternative tools which can monitor and analyze data. GIS can be use to 1.1 General Introduction increase productivity and profitability in the plantation Tea is a one of agricultural product of importance to the sector (Khairi Razi and Ismai, 2010) Sri Lankan economy and the world tea market.Sri Lanka is the world’s third largest exporter in tea market and the The essential parts of GIS implementation to plantation world’s fourth largest producer of tea. The productivity management shall include. Mapping. Spatial in the tea industry is remainining stagnant with yields management, data management, fertilization and pruning barely exceeding 1,400 kilograms per hectare. It is a low programs, production prediction and forecasting value in comparison; the productivity in India is around income.GIS is used to effectively manage these facilities. 1,800 kilograms and Kenya around 2,400 kilograms per Mapping of an estate is fundamental to the GIS system. hectare (Tea market update 2009). Utilize available lands Following mapping of the estate, data can be analyzed to on estates towards sustainable development by ensuring quantify and qualify plantation resources with the GIS land use efficiency with proper soil conservation software. Due to the dynamic nature of a plantation site, measures and environment protection is mandatory to from initial land clearing, growing stage through to increase yields. replanting, it is important for the information database to be current. Changes to the spatial information have to be Geo-Information Technologies (GIT) provides an easily modified in the GIS. important tool for the management of plantations. Prior to the introduction of global positioning systems (GPS), This study, which is an attempt to introduce GIS and RS Geographic Information Systems (GIS) and Remote techniques in order to create a Decision Supporting Sensing (RS), obtaining data on the field was difficult System (DSS) to maximize estate profits by utilizing and in many cases inaccurate. (Khairi Razi and Ismai, available lands on estates towards sustainable 2010) Furthermore, plantation management has to development by ensuring land use efficiency with proper consider the changing nature of an estate that extends soil conservation measures and environmental from initial land clearing, the production stage and protection . The strength of IT in the form of GIS finally the re-planting or conversion phase. GIS differs Reduce Exposure to Reduce Risk 178

solutions is expected to give excellent results towards of the feature are then stored, these data were stored as achieving this goal. layers. 1.2 Objectives Generated maps could be published by using publisher extension ArcGIS 10 software. Estate management can 1. To create a GIS based Decision Support System utilize these published maps with the help of ArcReader (DSS). software. ArcReader is a free, easy to use desktop mapping application that allows users to view, explore, 2. To demonstrate GIS based DSS applications for query and print maps. plantation management. This system permits generation of some reports to I. Identification of suitable lands for tea support decision-making in agronomic planning in a cultivation rapid and simple way. Thus this will lead to a Decision Support System (DSS) where record of almost every II. Identification of environmentally individual tea plant and all corners of the field will be sensitive areas. documented. 2 Methodology. 2.2 Investigating the applicability of the DSS 2.1 Development of GIS Decision Support System In order to demonstrate the application of DSS, (DSS) following analysis were carried out. Remote Sensing Images A. Identification of land suitability levels to tea Digital image processing cultivation: The main land factors or land characteristics like landform class or slope, soil Digitization Thematic maps depth and rockiness which are known to Vector layers Grid layers influence productivity of tea lands were considered in this classification. Estate Field Data base Data B. Identification of environmentally sensitive areas: According to the selected factors of GIS based DSS steepness and the proximity to water resources environmentally sensitive areas were identified. Figure 1: Diagram showing the procedure of Generating DSS 3. Results and Discussion Geo information technologies are novel field of 3.1 Generation of DSS for St.Coombs Estate application to plantation sector. Therefore most of the planters are less familiar with GIS and spatial data.Hence The study was conducted in St Coombs estate, it is necessary to designan extremely easy to use Talawakelle, Nuwara Eliya District as a case study. applicationfor this nontechnical audience. Satellite images (GEOEYE 1 Geo 0.5m) was used and various digital image processing techniques were DSS can be defined as, “A spatially based computer adopted. application or data assists a researcher or manager in An estate database was prepared including important general attributes which gives information on socio- making decisions” economic and environmental factors. The base data set consists of primary layers (layers which could be directly (http://www.umesc.usgs.gov/dss.html). In order to create extracted from the ground features) E.g. Road network, land use, building, teafields, stream network and DSS a GIS based Estate database was created. Spatial secondary layers (The layers which should be generated by processing the primary data sets) E.g. DEM, and non-spatial data were stored. To prepare the GIS elevation, soil, slope classes, stream reservation, environmentally sensitive area. Series of maps and database these spatial and non-spatial data should be information were generated and results are discussed below. linked, by giving every geographic feature a name or Building map: In order to increasing productivity of an number (usually just called its ID). Non-spatial attributes estate it is very much necessary to increase the labour productivity. This map provides important information on estate housing facilities up to individual house level. Management can utilize this map to identify development needs of the estate in relation to housing Reduce Exposure to Reduce Risk 179

welfare and sanitary conditions with the assistance of were further classified to provide cultivar level GIS specialists whenever necessary. information. Management can identify production trends, susceptibility or tolerance to pest, disease conditions and Road map: Road network was mapped. Roads were coarse climatic conditions like drought, heavy rainfall categorized into four groups. This will be a guide to the etc. with respect to spatial distribution. This will be more visitors. In addition, this map will be useful to plan informative in managerial prospects. transport of green leaf efficiently to the weighing shed or to the factory by saving time as well as the cost. Digital Elevation Model (DEM): Based on the existing contour maps and Satellite image a Digital Elevation Field map: Plantation management requires accurate Model (DEM) of the entire estate was generated. DEM information of estate boundaries and estate divisions for was used to generate slope map. management purposes. Therefore the field map of the St. Coombs estate was prepared using high resolution Slope map: Estate slope map was generated by using satellite image and existing secondary data prospecting DEM. The map was classified into five slope classes. to provide true distance, direction and area.Furthermore Proximity analysis techniques were used to identify land information on age, yield, pruning, labour requirement use pattern under different slope classes. Table 4 shows for major field operations (weeding, fertilizing, pruning the percentage of area under different slope classes. etc.) were stored in the data base, this data can be utilized for decision making or planning process. Similarly Table 4: Percentage of area according to different slope manager can use these kinds of maps to understand the classes relationship between different factors. All these data were presented in an interactive manner to offer more Value Class of Land Limitatio Percentage information to the user. Figure 2 shows the labour n of total requirement for fertilizer application while Figure 3 showing Pruning schedule of St Coombs estate. area 02- Gently undulating none 31% 04% plain 04- Undulating to None to 6% 16% rolling minor 16- Hilly terrain minor 27% 30% 30- Steep terrain moderate 29% 60% Figure 2: Labour requirement for fertilizer application >60% Very steep terrain severe 8% 3.2 Applications of the DSS: 3.2.1 Identification of land suitability levels for tea cultivation Figure 3: Pruning Schedule Landform class or slope, soil depth and rockiness were considered as the main land characteristics known to Land use and Land cover map: Visual classification influence productivity of tea lands. By considering above method was used to generate land use map. Estate land factors land suitability map was generated, according to cover was divided into six land use classes. Tea lands criterion shown in Table 5, Figure 6 illustrates the land suitability level for tea cultivation. Reduce Exposure to Reduce Risk 180

Table 5: Criterion to identify suitability level for tea cultivation Figure 7 - Map of sensitive areas Stream Reservation: The width of 20m distance from each bank of Stream/River is reserved as protected areas for water bodies, according to the Act of National Environment under Crown Reservations. This identification is done by using proximity functions, which demarcates the stream buffer from each side of the bank. These protected areas are not supposed to clear due to any reason as the area is required of preserving water. Stream reservatione t a l . , 2 0 0 8 ) Figure 6: Land suitability for tea cultivation 3.2.2 Identifying environmentally Sensitive areas The criterion for identifying the sensitive areas is as follows; by combining the areas reserved for the water bodies and the steep areas having the slope value greater than 60%, will be identified as the sensitive areas. Figure 7 displays locations identified as sensitive areas.

to minimize soil erosion in this area. The removal of any http://www.gisdevelopment.net/application/agriculture/o kind of tree/ plant in such identified sensitive areas verview/mi04111.htm should not be under taken at all, due to any reason. An attempt was made to provide specific GIS functions Khairi M, Razil M, Ismail M.H, 2010. Applying GIS for in an easy-to-use package to the plantation management Mapping Agricultural Roads Network in Felda Trolak in order to carry out analysis to solve spatial issues using Utara for Oil Palm Plantation Management, Information GIS tools. All sorts of plantation information, scheduling Management and Business Review Vol. 1, No. 1, pp. 11- information etc. will be stored in a centralized location, 15. which would help the management to take effective managerial decisions. All together maps and information Miller D implementation of GIS to palm Oil Plantation was published using arc reader software in order to bring Management in Indonesia. all the information into a user friendly environment. This http://www.gisdevelopment.net/application/agriculture/o will be the GIS based decision supporting system. verview/ma07201.htm In this study, visual interpretation techniques were used Murai, Shunji, 1999. GIS workbook, Vol. 2, Asian to extract ground features. As an example for the Institute of Technology, Bangkok, Thailand. classification process visual classification method was used in this study. Although automated classification Panabokke C R, Amarasinghe L A, Pathiranage S R W, methods (supervised and unsupervised classification) are Wijeratne M A, Amarathunga S L D and widely recognized, they have some limitations when Anandacoomaraswamy A 2008. Land Suitability applying to high spatial resolution imagery (Carlos Classification and Mapping of Tea Lands in Ratnapura Glenn and Sandra, 2002) as these images contain more District. S.L.J Tea Sci.. 73(1),1-10 details to recognize and the result will be more complicated. Nowadays computer, GIS and remote Rafieyan O, Gashasi J and Sani A 2008. Updating the sensing technology offers novel possibilities for land cover map using satellite data in order to integrated managing, editing and generating raster and vector data, management of natural resources. World applied facilitating the visual interpretation methods, thereby, sciences journal.3(1):48-5 this methodology results more informative and error free maps (Carlos Glenn and Sandra, 2002). Rajapakse R.M.S.S, Jayakody J.A.A.M Jayawardena MP Prepare sustainable strategic development plans for 4. Conclusions some selected tea estate in mid country region in Sri Lanka using GIS  It is necessary to take policy decisions to protect http://www.gisdevelopment.net/application/agriculture/p identified environmentally sensitive areas. roduction/agric0010.htm.  Immediate actions should take to remove tea Statistical Bulletin of the Tea Board of Sri Lanka cultivation from identified unsuitable area in 2010.Sri Lanka Tea Board, Sri Lanka. pp 38-39. order to increase land productivity. 6. Acknowledgements  A GIS has the power to integrate different information and visualize scenarios, present Mr. U.C. Oliver (Former manager of the St. Coombs ideas, and provide solutions for complicated estate) problems. Therefore it can be suggested that GIS can be successfully used in tea sector for its Manager & the staff of the St. Coombs estate managerial purposes. Head of Division & the staff of Agricultural Economics  GIS and RS techniques can be effectively used Division ,Tea Research Institute of Sri Lanka to create a system to support decision-making process in the plantation sector. 5. References De Alwis K A., Fernando L H., Jayasooriya S E., Kulasegaram S., Perera M B A., Sandanam S., Sivasubramaniam S and Wettasinghe D T 1980, Simplified Land Suitability Classification for tea.Tea Quarterly.49 (2), 5-12 De Silva R L 1979. Proposals for diversification of agriculture on tea plantations.1.The concept and underlying principles.Tea Quarterly.48, 26-31. Ghosh A. GIS Anchored Integrated Plantation Management Tea. Reduce Exposure to Reduce Risk 182

RISK MAP ASSESSMENT METHODOLOGY – A PILOT STUDY FOR CENTRAL REGION OF SINGAPORE J. Chandrasekar1 and D.K. Raju2 Tropical Marine Science Institute, National University of Singapore, Singapore E-mail: [email protected], [email protected] ABSTRACT GIS is being applied in all fields of research. This emerging technology has become the backbone for providing information and analyses for assessing flood risk to properties and human lives. This pilot study was carried out for Central Region of Singapore to assess land loss and associated damages to buildings under inundation levels of 2.5m, 3.0m and 3.5m. A digital elevation model (DEM) is used to compute flood and flood-depth map for given inundation levels. Automated tools were developed in ArcGIS using the NOAA’s Bath-tub approach for generating flood and flood-depth maps. To estimate flood loss, the buildings were classified with the help of Comprehensive Data Management System (CDMS) data dictionary for floods. CDMS is a data dictionary which defines the datasets required for Earth Quake, Flood and Hurricane in FEMA’s (Federal Emergency Management Agency, USA) HAZUS-MH model. ArcGIS and FME (Field Manipulation Engine) tools were used to customize the building inventories into the CDMS format. This study uses the flood-depth grid, combined with building inventories in order to estimate the flood damage factor. The ArcGIS spatial overlay techniques were used to identify flood-affected buildings and the US Army Corps damage function were used to derive damage for different types of residential facilities. KEY WORDS: ArcGIS, Bath-tub, CDMS, DEM, FME, NOAA, US Army Corps IDENTIFICATION OF URBAN POPULATION DISTRIBUTION AND SPATIAL PATTERN OF URBANIZATION IN THE CITY OF COLOMBO, SRI LANKA K.A.S.S. Wijesekera1 and Ranjith Premalal De Silva 1Department of Geography, university of Kelaniya, Sri Lanka, E-mail: [email protected] 2Uva Wellassa university, Badulla, E-mail: [email protected] ABSTRACT Urban population distribution and the spatial expansion of the urban areas show a considerable relationship. Usually, the urban population shows a very dynamic and unevenly distributed pattern. This study attempted to identify urban population distribution and spatial pattern of urbanization in the city of Colombo with the help of Geo Informatics. The main objective, of the study considered urban population distribution with the concept of “Concentric Zone Model (Burgess Model)” for the period of 1891 to 2001. To fulfill this objective, census data for the considered time period was gathered at city ward level. ArcGIS 9.3 software was used to analyze the specific features of spatial population distribution into concentric zones. The results showed that population density movement from city center or Central Business District (CBD) to city periphery occurs in a pattern of concentric zone. Geo informatics provides geostatistical background to identify and compare the pixel distribution in a given themes. To build up an application for identifying population activities in a certain area, satellite data can also be used. It is visible that complex constructions usually relate with highly populated areas like urban centers. Urban Index (UI) is an important index for presenting urbanization and constructions in an area. UI normally relates with population density and vegetation cover. To identify the spatial pattern of urbanization, Landsat ETM + satellite data in year 2001 for Colombo city area were used. With the help of ERDAS Imagine software, UI and Normalize Difference Vegetation Index (NDVI) were constructed using the satellite image. Regression analysis was used to identify the relationship between Population Density and NDVI with UI. The regression analysis shows moderate positive linear relationship between Population Density and UI. Also moderate negative linear relationship between NDVI and UI. KEYWORDS: Population density, urbanization, Colombo, urban index, Satellite data, GIS Reduce Exposure to Reduce Risk 183

MAPPING THE SPATIAL PATTERN OF RAINFALL AND ELEPHANT INTRUTION IN SRI LANKA Rathnayake Mudiyanselage Chithrangani Wasantha Menike Rathnayake, Uva Wellassa University, Badulla, Sri Lanka, E-mail: [email protected] ABSTRACT Elephant intrusion in human settlement is one of the key issues in Sri Lanka which have the devastating impact on humans and the elephants which have been continued for several decades. Elephants arrive to the human settlement areas mainly to fulfill the requirement of food as the lack of the food and the water sources in the protected areas. The amount of forest area in the island has decreased and remains only about 18% out of the total land area of the country. Therefore as a lager species that required larger amount of food and water it is required to roam in a larger roam range to need the requirement of food and water outside the protected areas. It is very simply known as human elephant conflict (HEC). The conflict between the humans and the elephant will continue for a long time because of the socio physical changes happen Sri Lanka. Elephant intrusion highlighted as a top issue in Sri Lanka as a result of socio physical changes which has happened in last few decades along with the rapid development activities took place around the country. The changing climate also has direct influence on intensifying the issue with the changes of the climatic elements specially with the changes in rainfall. The global climate in the world has undergone changes resulting unbelievable impact to both physical and the human environment. The seasonal pattern has changed and the amount and the intensity of the rainfall have decreased according to the studies carried out by the scientists. And along with the decreasing rainfall the atmospheric temperature has increased. These changes were dangerously impacted on plant and the animals on the earth. The elephant that need larger home range to survive has brutal effect with the changes in the climate. This study has carried out to investigate the relationship between monthly rainfall distribution and the elephant intrusion into human settlement. It has mapped the distribution monthly rainfall pattern and the elephant intrusion from January 2003 to December 2009. It has shown that the distribution has an inverse relationship with the rainfall. If there are enough rainfall where the elephant can have enough fodder and the intrusion keeps less and if the rainfall is less then the intrusion keeps higher. KEY WORDS: Elephant intrusion, climate, climate change, distribution, rainfall. Reduce Exposure to Reduce Risk 184

Technical Session-8 [Hall A]: Coastal Hazard / Forest Fire & Haze Monitoring Shoreline Change in the Coastal Area of Quang Tri, Vietnam by using 186 Remote Sensing and GIS 193 Pham Thi Phuong Thao, Ho Dinh Duan 198 198 Characterization of the February to April 2011 Muang Phayao District, Thailand Fire 199 Disaster Using Satellite Remote Sensing and GIS 199 200 Phaisarn Jeefoo 200 201 Modelling the Effect of Coastal Landuse on Tsunami Inundations Patterns along the 202 Selected Stretches of South Indian Coast K.Srinivasa Raju, G.Gopinath, M .Ramalingam, Bhoop Singh Coastal Line Changed Detection in Thailand W utjanun M uttitanon Generation of Katabatic Wind along the Lee Side of The Mountain Adjacent to the East Sea of Korea Doo Sun Choi, Hyo Choi and Soo M in Cho Cooling of Sea Surface Waters Induced by Strong Marine Wind Under the Intensification of A Coastal Low Pressure Near the East Sea of Korea Hyo Choi, Soo M in Choi Coastal Erosion Hazard Assessment, South West Coast, Sri Lanka Bandula W ickramarachchi, Talpe Liyanage Chanaka Vinodh Developing a Decision Support System for Forest Fire Risk Modeling in Praviyangala Area of Walawe Basin N. Saranga Vithanage, K.V.D. Edirisooriya Regressive Prediction of PM10, PM2.5 and PM1 Concentrations at a Korean Eastern Coastal City Under the Dust Transportation from Gobi Desert M i Sook Lee, Hyo Choi Formation of Sea and Coastal Fogs by the Intrusion of Warm Air Over Cool Sea Surface by Cyclogenesis in the Yellow Sea Hyo Choi Reduce Exposure to Reduce Risk 185

MONITORING SHORELINE CHANGE IN THE COASTAL AREA OF QUANG TRI, VIETNAM BY USING REMOTE SENSING AND GIS Pham Thi Phuong Thao(1) and Ho Dinh Duan(2) (1)Institute of Oceanography, Vietnam. Email: [email protected] . (2)Hue Institute of Resources, Environment and Sustainable Development. Email: [email protected] Abstract Shoreline evolution is one of great concerns for coastal engineers and managers. Nowadays, it is easy and effective when using remote sensing combining with GIS tools to monitor erosion/deposition and calculate rates of change along a coast. In this paper, these techniques were applied in Quang Tri province, Vietnam from 1986 to 2010 by using Landsat images. Shoreline change rate was calculated by Digital Shoreline Analysis System (DSAS), an ArcGIS extension. The results showed that shoreline tended towards accretion in this area in general. At two estuaries, shorelines were observed to have changed remarkably due to a complex littoral processes and sediment discharge from rivers. Keywords: Landsat, remote sensing, GIS, DSAS, shoreline change 1. Introduction Sea, cliff coast in Suffolk, England [Tuncay, 2011; Brooks and Spencer, 2010]. In developing Shoreline evolution is one of great countries, achievements in this direction have also concerns for coastal engineers and managers been obtained. Using Landsat imageries, tend of because it can make a lot of damage for local erosion/deposition in 10 years was analyzed in economy and people. The erosion, for example, can Accra, Ghana and from Kanyakumari to Tuticorin affect strongly to coastal constructions, tourism coast, India [Frederick, 2011; Sheik and beaches, and fishing villages while the deposition Chandrasekar, 2011]. In Vietnam, shorelines were can make trouble for ship to transport near estuary monitored at Nam Dinh, Binh Thuan, and Kien because of dunes, sand bars or a narrow width of Giang province [To and Thao, 2008, Pham Thi river mouth. Therefore, monitoring shoreline Phuong Thao et al. , 2008, Hai Hoa Nguyen et al., change is very necessary and remote sensing is 2010]. very effective for such purpose. Besides, with nowaday GIS tools, rate of change can be In this paper, Landsat images were used calculated quickly after having shoreline data. for monitoring shoreline change at Quang Tri area, These techniques have proven efficient and cost- especially at two river mouths. First, shorelines at effective compared with traditional approaches. different time slices were extracted from analysis of Landsat images. After that, extracted shoreline In the United States, historical shoreline data was imported into DSAS to calculate rates of data are used for calculating rates of change by change in the period of 1989-2010. using Digital Shoreline Analysis System (DSAS), an ArcGIS extension, along US coast such as 2. An overview of the study area - Quang Tri Texas, Gulf of Mexico, Southeast Atlantic, province coastlines California, Boston, New England, Mid-Atlantic, and Alaska [Cesar, 2004; Morton et al., 2004; The study area is Quang Tri coast Morton et al., 2005; Hapke et al., 2006; Hapke and stretching from Mui Lay to Hai Duong with 50km Reid, 2007; Maio, 2009; Himmelstoss et. al., 2011; in length. There are two estuaries, namely, Tung Yuri and Leiserowiz, 2012]. Using the same estuary and Viet estuary in this area. Therefore, method and GIS tool, rates of shoreline change in four regions are divided as follows (Figure 1): some other countries have been calculated as well. In Europe, such projects were carried out in • Region 1 is Tung estuary Ramsar wetland, Turkey and at southern of North • Region 2 is from southern Tung estuary to northern Viet estuary (~14 km) Reduce Exposure to Reduce Risk 186

• Region 3 is Viet estuary • Region 4 is from southern Viet estuary to Hai Duong (~30 km) Region 1 - Tung estuary Region 2 Region 3 - Viet estuary Region 4 Figure 1. Quang Tri coast – the study area. Quang Tri is affected by two monsoon 59-116 cm. Maximum amplitude low tides is seasons, the winter and the summer. The winter nearly monsoon begins from October to March next year. The predominant wind direction in winter is from the same to the high tide north and northwest with the mean wind speeds of [http://gioithieu.quangtri.gov.vn/]. 3.1-4.4 m/s and 3.4-4.0 m/s, respectively. The summer monsoon begins from April to September Quang Tri coast is influenced directly by with hot and dry wind from southwest. The mean one or two tropical stroms every year. Tropical wind speed is approximately 4.5-5.2 m/s [Nguyen storms occur from June to October, especially in Van Cu et al., 2008]. August-October with the frecency of 0.3-0.7 storms/year [Nguyen Van Cu et al., 2008]. About wave regime, predominant direction is from northeast offshore and changes to 3. Methodology and Data east nearshore in winter. The wave height is about 0.5-1.5 m. In summer, predominant wave is from 3.1 Methodology southwest and west offshore with 0.5-0.75 m in height and changes to southeast nearshore with 0.3- In remote sensing, waterlines were 0.5 m in height [http://moitruong.quangtri.gov.vn/]. extracted from Landsat images for Quang Tri area. Band threshold was applied in this study. Band 7 Hydrologically, this area has a mixed was used because its spectrum is good to semi-diurnal tide with a small amplitude. discriminate water from land [Jensen, 1996; US Maximum amplitude high tides varies in range of Army Corps of Engineers, 2003]. Tide adjustment was neglected. The first reason is because the resolution of Landsat image is not very high (30m) and tide amplitude is small (approximately 1.2m) in this region. Second, the precise data of beach Reduce Exposure to Reduce Risk 187

profile are not always available. Therefore, only EPR and LRR are shown. We need more input extracted waterlines were assumed as shorelines. data for WLR calculation. After having shoreline data, rates of 3.2 Data change were calculated by using DSAS [Thieler et al., 2009]. The process includes three main steps: Source of imagery data is from the website of United States Geological Survey - setting up baseline and shorelines (USGS). All Landsat images have the same - choosing parameters for transects projection of WGS-84, zone 48 and the same 30m - calculating shoreline change rates resolution. In addition, sea level data was collected from WXTide32 (http://www.wxtide32.com) for In DSAS version 4.0, there are some Nhat Le River station. All data are listed in Table statistical methods, such as End Point Rate (EPR), 1. Linear Regression Rate (LRR), and Weighted Least Squares Regression Rate (WLR). In this paper, . Tabe 1: List of Landsat scenes used to extract Quang Tri shorelines and sea level data N Sensor Date Local Time Sea level (cm) o. 08/01/1989 09h45 66 1 Landsat 4TM 2 Landsat 5TM 28/06/1996 09h25 61 3 Landsat 7ETM+ 30/05/2000 10h04 44 4 Landsat 5TM 04/05/2005 10h00 29 5 Landsat 5TM 11/02/2010 10h03 46 Note: Mean sea level at Nhat Le station (for Quang Tri area) is about 60 cm. All images are from January to June in year to tended to spread up to northern estuary, which prevent sudden great shoreline changes in a very decreased the width of the river mouth. This could short time due to storm period. make trouble for ships to go to Viet estuary port which is near the river mouth. After 2005 with jetty 4. Results and Discussions construction, sediment deposited at southern estuary, which increased the width of beach and There are four regions in this study area: made the river mouth much wider (Figure 2). The two river mouths at region1 and 3; and two straight presence of the jetty was not so clear in Landsat shorelines at region 2 and 4. Monitoring erosion image in 2005 and 2010 because the width of the and deposition was carried out at estuary areas jetty (15m) is so small when comparing to the while calculating rates of shoreline change was resolution of image (30m). In region 3, both applied in other areas with a simple shape of northern and southern of estuary tended to shoreline. accretion in 1989-1996 but tended to erosion in 1996-2000 and accretion again in 2000-2005 In region 1 and 3, shoreline change has (Figure 3). However, in the period of 2005-2010, occurred so complex because of hydrodynamic northern part was eroded while southern part was processes in the estuaries and the interactions deposited. The reason may be due to the effect of between the estuaries and coastal zone as well. In alongshore current combining to sediment source region 1, there was no jetty at southern Tung flowing down from river. estuary before 2005. The width of beach was so narrow. In addition, sand bar at southern estuary Reduce Exposure to Reduce Risk 188

1989-1996 1996-2000 2000-2005 2005-2010 Figure 2. Shoreline change in region 1 – Tung estuary in the period of 1989-2010. 1989-1996 1996-2000 2000-2005 2005-2010 Figure 3. Shoreline change in region 3 – Viet estuary in the period of 1989-2010. In region 2 and 4, shorelines are quite straight, so it Transects are perpendicular to shoreline as shown is convenient to calculate rates of change by using in Figure 4. DSAS. Distance between transects is 100m. Reduce Exposure to Reduce Risk 189