WEB BASED SPATIAL DECISION SUPPORT SYSTEM FOR FLOOD DISASTER MITIGATION

Abstract

Water is a necessity of life. It provides support to life of plants and living being. It is a fact that moderate amount of water provides support to life but a large amount of it can leads a huge disaster and this disaster named as Flood. Generally floods occur due to the river overflows from its banks. This flood water enters the agricultural land and settlements to cause disaster in the region. A large number of people affected and economic losses resulting from flood are on rise at an alarming rate globally. The main research objective of present study is to develop a web based spatial decision support system (SDSS) for flood disaster mitigation. This research work has been divided in three sub-parts, which includes flood inundation mapping and damage assessment, developing a web-based spatial DSS for flood disaster mitigation and the third is to identify the safe and shortest route for evacuation during flood disaster. According to the Government of India-UNDP Disaster Risk Management (DRM) programme (2002-2007), Kendrapara, Jagatsinghpur, Cuttack, Jajpur and Bhadrak Districts of Orissa state are the most flood hazard prone districts. The delta region formed by the Nuna and Chitrapala rivers of lower Mahanadi region is taken as the study area for this research. It is located between East Longitude 86°16´30" to 86°30´05" and North Latitude 20°22´30" to 20°31´30". A lot of data and information has been gathered for this study. It includes: Geospatial data as DEM (Digital Elevation Model), temporal satellite images before, during and after flood, Radarsat SAR, IRS P6 LISS-III / LISS-IV, Landsat ETM+, Google Images, Bhuvan data, Toposheets from Survey of India; Socio-economic data; spatial analysis softwares as ArcGIS, ERDAS Imagine; Open source tools used for web GIS application. A field visit was performed during July 2012 for collecting various data and information. This study uses the Radarsat-1 images of 04-September 2003 and 11-September 2003 for flood inundation mapping. It shows the comparison between different flood duration images. IRS P6 LISS-III image, dated 21 January, 2006 has been used for creating land use map with the help of ERDAS Imagine software. The supervised image classification technique has been used to find ten different classes. These classes are based on field visit, ii Google imagery, Government organization data and various remote sensing data. On the basis of the Land Use Land Cover map the main elements at risk have been identified as Built Up, Agricultural Land- Kharif Crop, Agricultural Land- Rabi Crop, Agricultural Land- Double-crop area and Plantation. Flood damage assessment is done by overlaying land use map with flood map. The output map divides the classes into water and land area. It shows the area of different classes in water and land. As most of the study area is rural area, therefore the agriculture land is the prime concern for vulnerability. The crop land and built up are more susceptible to the flood water. The height of the flood and the duration of flood are important factors for the damage assessment. The duration of flood is also a very vital issue related to the crop vulnerability. So as seen in the comparision of different Radarsat data, it is observed that there is a very high damage occurrence. In terms of money it is calculated that more than INR 100 Billion loss of agriculture and INR 17 Billion loss of build-up and roads occur due to the flood. The delineation of optimal route has been carried out using three different methods and then compared to find the most suitable one. It uses five parameters for finding the priority of route using AHP, Fuzzy and ANFIS methods. The three approaches are knowledge base, data base and learning base approaches. All these three approaches are implemented and compared for the optimal path extraction. An intelligent fuzzy based decision support system has been constructed on the basis of the comparison of these three approaches. ANFIS based DSS for safe route delineation has been implemented and different flood scenarios have been explained. The elevation data plays an important role to find the optimal path in this work. The three approaches applied in this work for implementing fuzzy inference system (FIS) are: knowledge base, data base and learning base methods. In knowledge base modeling approach, FIS is implemented using expertise knowledge. FIS has been implemented in MATLAB using fuzzy logic toolbox. This modeling requires very deep understanding of input output relationship in each and every scenario. In data base modeling, FIS is modeled directly from data therefore this modeling is known as data base modeling. Clustering of numerical data forms is the basis of many classification algorithms. In this work, since it is quite difficult to calculate the number of clusters in data set, therefore Subtractive clustering method has been applied. Then fuzzy logic has been applied to extract the broad iii categories of the clusters. This method is applied for the route evacuation problem, output of the fuzzy system matches to expected value to a degree of 91% to 97%. Thus it can be seen that though clustering and fuzzy logic are effective techniques for data modeling and analysis but still optimization of result is required. For this, fine tuning of membership functions is required which can be done through learning base modeling. In the learning base modeling, the FIS needs to go through training and learning process. Adaptive learning techniques allow the fuzzy system to learn and extract information from the data they are modeling. Throughout the training, parameters associated with membership function (MF) change and finally adapt the shape and parameter values that best allow the FIS to respond in the approved manner. In this work, FIS is first implemented via all the above mentioned different approaches, then results or decision of each approach has been compared and at last most appropriate decision has been projected out. The hybrid learning method has been selected for FIS training in this study. FIS model output is to be tested against all the three data sets one by one. The average testing error for the training data set is 4.7x10-4. And the average testing error for checking data set is 15.4x10-4 and for testing data set is 8.2x10-4. The checking data has a very near values compared to the FIS values, it is found that the resultant values are accurate up to 99%. Therefore it is observed that the system is working very near to the required result. There are some limitations with the accuracy of the input data, so the result cannot be 100% accurate. But under the practical concern the error up to 6 feet can be considered with these data sets. The flood extent map is based on the DEM of area. In flood risk zone map the high elevation area is considered as low flood risk zone. The road network map of the area has been constructed on the basis of Google map, field visit, satellite imagery and open source map. The shelter zone map has shelter points on the basis of its Elevation, proximity to safe zone, capacity and convenience. These may be school buildings, commercial complexes, Government constructed shelter zones, or a shelters developed by other public organizations. iv Apache Tomcat and GeoServer has been used for constructing a web GIS based DSS for flood disaster mitigation. The GeoServer is like an application server which holds spatial data by using PostGIS for spatial database management and it retrieves data to display it with the application. It stores various GIS layers such as LULC layer, road network layer, soil layer and Google Earth layers. This web GIS can perform a number of functions on these layers. The PostgreSQL is used for information extraction from the layers. Length and area measurements, route display with names of the path are some of the key features of this web GIS. The user friendly interface is developed in JavaScript. It provides an easy way to query. This web GIS is based on open source software, therefore it also has capability to access the resources available on the Internet based on open source. This GIS is developed to support the flood disaster mitigation. The rescue operation requires shortest and safe path with detail, and this research work provides a very efficient path finder. So this web GIS can play a vital role during flood disaster. This web GIS provides some functionality to length and area measurements, route display with names and visualization of the path are some of the key features of this web GIS. This web GIS is using open source software, therefore the limitations of a freely available spatial geo processing tool is also the limitations of this web GIS. If very high resolution satellite images will use than the accuracy of damage assessment will become better compared to the LISS III images and Radarsat data. An Android based application can be developed for the flood disaster mitigation. It may integrate Google API with android application for further development of flood disaster mitigation app.