AbstractMostly river water level and flood forecasting methods are based on gauging stations measurements at discrete locations, which limitstheir capability to provide accurate and timely data over large extent, alsolimited or no data available on remote locations.
So here we present an idea touse high resolution satellite images for real time mapping of river water level.In this project, we developed a Web based GIS system for mapping river water level,early warning and mapping for flood disasters. To improve flood forecasting/warning,we developed a decision support system (DSS) for flood monitoring and predictionthat integrates GIS, satellite image processing and hydrological modelling. Wepresent the methodology for data integration, floodplain delineation, andonline map interfaces. Our Web based GIS system can dynamically displayobserved and predicted water levels for decision makers and the general public.The users can access a Web-based GIS system which models current flood eventsand displays satellite imagery and 3D visualization integrated with the floodplain area. The output from the hydrological modeling will be used for floodingprediction for the next 1 day to 2 days (24 and 48 hours) along the lower IndusRiver. In this stage river water level analysis has been achieved, work on thehydrological modelling is in progress to acquire river water stage & floodlevel and the prediction.
Keywords: Hydrological modelling, GISbased system, morphology, satellite imagery, 3D visualization.1. IntroductionDroughts and floods are both water-relatednatural disasters which have a widespread effect on environmental factors andactivities like human life, agriculture, vegetation, local economies and wildlife. These both are natural disasters which are thought to be beyond humancontrol, but drought is the one of the most important weather-related naturaldisaster which is often aggravated by human action. Drought affects a verylarge area for months and even for years. So it has a severe impact on regionalfood production, life expectancy of the entire populations and overall economicperformance of large regions or countries. If we observe the recent datalarge-scale severe droughts have been observed in different areas of the worldencompassing all continents leading to economic and natural resources loss,this destruction lead to food shortages and starvation of masses.
On the otherhand floods are the most devastating natural hazards in the world. Floods weremost baleful than any other natural disaster both in claiming more lives andcausing more property damage.For effective management of disaster users like toplevel policy makers at the national and international organizations, middle levelpolicy makers at local levels consultants, researchers, relief agencies andlocal producers which includes farmers, water managers suppliers and tradersare interested in reliable, accurate andtimely information of drought and flood. The disaster management activities canbe grouped into three major phases: The Preparedness phase in which predictionand risk zone are identified, identification is done long before the actualdisaster event occur; in Preventionphase different activities are carried out like, monitoring, early warning& Forecasting, and preparation of contingency plans that should be taken upjust before or during the event; and the Response/Mitigation phase includes the activities which are done damageassessment and relief management.Though flood cannot be stop buttheir effect can be minimized if we have proper data, so in this project ourobjective is to map the river flow data on geographical information system. Asrivers cover a large geographical area so we can obtain data using remotesensing techniques and distribute information to control stations rapidly overlarge areas by means of satellites or transponders mounted on drones or aircraft.A satellite orbits the Earth, can explore the whole surface in a few days andrepeat its survey of the same area at regular intervals.
For these satelliteimages we will use publically available high resolution images like NASA MODIS,SPOT etc.2. MethodologyThe proposed project models a particular area on the GIS andcollect relevant data like satellite images of high resolution at differenttime period. The backend processing application will take these images asinput; image will go through multiple stages of processing and output of thiswill be quantifiable data along will the geo location. The front end GIS basedweb or mobile application will take date from the data base and model it on thegeographical information system. Fig. 1. GISbased Prediction system block diagram2.
1. Satellite Image Processing ResearchSatellite image processing techniques will be used for effectivemanagement of water bodies on land and used for decision making for disastermanagement. An extensive research has been done for identifying and designing thebest suited algorithm for analyzing and processing the high resolutionsatellite images. The process of identifying the stage of a river usingsatellite images is a difficult process, because the images captured from satelliteare remotely sensed at a very long distance.
We utilize the Support VectorMachine algorithm for this purpose. The work is divided into three phases thetraining phase, analysis phase and the testing phase. In the training phase, waterregion is extracted from the image and the stage of river is identified both bytraining two different ANNs.
In the testing process, the input raw river image waspassed through different process first image filtration to de-noised andmorphological operation was carried out on the de-noised image. After that theriver image was segmented into water regions with the aid of the ANN. Finallyin the analysis phase, the stage of the river water was categorized on scale ofthree whether river is in Normal, drought or flood condition.
2.2. SatelliteImage Data GatheringOne of the most important thing in our project is thehigh-resolution satellite image data, which is publically available ondifferent platforms some are: NASA MODIS Imagery SPOT Imagery JAXA’s Global ALOS 3D World NOAA Data Access Viewer – Discover Authoritative Datasets Global Land Cover Facility – Derived Satellite Data 2.
3. LayoutDesigning Low fidelity prototypes havebeen designed, some basic prototypes are shown below:Fig. 2. River Geographical MapView Fig. 3.
Highresolution Satellite Image2.4. ImageProcessingAn algorithm has been developed which is able toidentify the water bodies in the high resolution satellite images with anaccuracy of over 90%.
In order to understand the flood dynamics andhydrological exchange between river and related floodplains, it relies on thedata gathered of water levels at discrete locations on main channel. But due tovast expansion of the river basins we have limited gauging stations and riverpasses through locations where it’s difficult to have these stations, so someother techniques must be implied in order to improve the spatial sampling, likeradar remote sensing using interferometric phase measurements this can improvethe spatial sampling, but the phase is temporally incoherent over open water sothis technique is used to determine water levels.Satellite images are processed and passed through thesedifferent stages to extract the river stage.
Complete flow is shown with Image Processing steps shownin blue box: 2.4.1 Image AcquisitionAcquired animage from open access NASA database Fig. 6. This is a moderateresolution satellite image The image is passed throughdifferent steps to extract the river stage.2.4.2.
Image Filtration:Filtering is used to modify and enhance the image byremoving the unwanted pixel values in the image. For example, you can filter outan image to emphasize certain features or remove other features. Imageprocessing operations implemented with filtering include smoothing, sharpening,and edge enhancement.In image processing filtering is a neighborhoodoperation, in which the value of each pixel in the output image is calculated byapplying some algorithm to the values of the pixels in the neighborhood of thecorresponding input pixel. A pixel’s neighborhood is the surrounding pixels,defined by their locations relative to that pixel. 2.4.3.
Lab color space ConversionTheLab color space describes mathematically all perceivable colors in the three dimensionsL for lightness and ‘a’ and ‘b’ for the color opponents green–red andblue–yellow. The Lab color space exceeds the gamut of the RGB and CMYK colormodels. One of the most important attributes of the Lab model is deviceindependence. For the conversion of RGB image to the Lab color format belowgiver equations are used.After conversion to CIE Lab color image look like: Fig. 7. CIE color converted Image. 2.
4.4. Anisotropicdiffusion segmentation:In imageprocessing and computer vision,anisotropic diffusion, is a technique which is used to reduce image noise withoutremoving substantial parts of the image content, mostly edges, lines or other details that areimportant for the analysis of the image. In Anisotropic diffusion we creates a scale space, in whichan image generates a parameterized family of consecutively more and moreblurred images based on the result of diffusion process. Each of the resultingimages in this family are given as a convolution between theimage and a 2D isotropic. The following equation detailsthe segmentation process. Equation 2. Delta Functionapplied to Image.
Where div is short for the divergence operator, ?stands for the Laplacian, ? for the gradient and c(x,y,t) is the diffusioncoefficient. c(x,y,t) is used to control the rate of diffusion and it isselected as a function of the image gradient in order to conserve edges in the image. 2.4.5.
ImageBinarizationThe segmented image is convertedinto binary image which will be used for further processing. A binarization isa method of binarizing an image by extracting lightness (brightness & density)as a feature amount from the image. In binarization pixel image is converted intobinary image. When a pixel is selectedin an image, sensitivity is added to or subtracted from the value concerningthe Y value of the selected pixel to set a threshold value range. Next, whenanother pixel is selected, the sensitivity is added to or subtracted from thevalue concerning the Y value of the selected pixel and a new threshold valuerange is set containing the calculation result and the already setup thresholdvalue range.
The pixel with the value concerning the Y value of any pixel inthe image within the threshold value range is extracted as the same brightnessas the selected pixel and the extraction result is displayed. The output imageafter binarization contain only black and white image. 2.4.
6. Morphologicaloperation:The morphological operation is appliedto smooth the image for extracting the intended information. The impacts of themorphological operation are on the application for example in machine visionand object detection in which we may utilize it to classify the structure inimage.
Morphological operation is utilized to distinguish boundaries or objectsor object present in an image. For this, we apply the “imclose” function that isused to morphological closes the image. The following equation details theprocess. Equation. 3. Image segmentationexpression Here st is structuring element which is required by which the morphologicaloperation is applied.
Image after morphological operationlooks like: Fig. 8. Morphological Image2.
5. TrainingDataSo as to identify the stage of Support Vector Machine(SVM) technique is employed, SVM is trained by a set of water identifies images.Support Vector Machine is a machine learning method that classifies binaryclasses by finding and using a class boundary the hyper plane maximizing themargin in the given training data. The training data samples along the hyperplanes close to the class boundary are called support vectors, and the marginis the distance between the class boundary, support vectors and hyper-planes.
The SVM work on the basis of the decision boundaries defined by the decision planes.A decision plane separates between assets of objects having different classmemberships. SVM is a one of the very useful technique for data classification.For classification task training and testing date is utilized which consists ofsome data instances.
Each instance in the training set contains one “targetvalue” (class labels) and several “attributes” (features). Image withidentifies stages of river water. Fig. 9. River water levels3. ConclusionThis paper presents the idea for theintegration of satellite imagery, hydrological modeling and geographicalinformation system to come up with a decision support system which will be usedfor continuously monitoring the river water level. By processing the satelliteimagery we can extract the river stage over the entire spatial range from wherethe river passes.
Using historical date we can also predict the next stage fornext 24 to 48 hours.The system for flood eventsprediction and monitoring integrated with the web map interface willfacilitates in monitoring, prediction and decision making of the floodingevents. It works as an early warning and mapping of flood disasters.4. References1. Laurence C. Smith, “Satellite Remote Sensing ofRiver Inundation Area, Stage, And Discharge: A Review”, HydrologicalProcesses, Vol. 11, pp.
1427-1439, 19972. Flood Forecasting GIS Water-Flow VisualizationEnhancement (WaVE): A Case Study Timothy R. Petty1, Nawajish Noman2, DengDing2, John B. Gongwer3 3. DROUGHTS & FLOODS ASSESSMENT AND MONITORINGUSING REMOTE SENSING AND GIS by A.T.
Jeyaseelan4. Nipuni Hettiarachchi “Water Level Forecasting andFlood Warning System A Neuro-Fuzzy Approach”, International Conference inVideo Image and Signal Processing, At India5. Michael T. Coe, “Calculation of river dischargeand prediction of lake height from satellite radar altimetry: Example for theLake Chad basin”, Water Resources Research, Vol.
40, 2004 6. D. Mioc “EARLY WARNING AND MAPPING FOR FLOODDISASTERS”, The International Archives of the Photogrammetry, Remote Sensingand Spatial Information Sciences. Vol.
XXXVII. Part B4. Beijing 20087. HJ Barneveld “APPLICATION OF SATELLITE DATA FOR IMPROVED FLOODFORECASTING AND MAPPING”, 4th International Symposium on Flood Defence: ManagingFlood Risk, Reliability and Vulnerability8.
Toronto, Ontario, Canada, May6-8, 20089. https://modis.gsfc.nasa.gov/gallery/10.
https://earthobservatory.nasa.gov/IOTD/view.php?id=4520011. https://www.
visibleearth.nasa.gov/view.php?id=6505012. R. Loren and D.
B. Benson (eds.), Introduction to String Field Theory, 2ndedn. (Springer-Verlag, New York, 1999).13.
R. M. Karp, Reducibility among combinatorialproblems, in Complexity of ComputerComputations (Plenum, New York, 1972), pp. 85–104.14. R.
Loren and D. B. Benson, Deterministic flow-chartinterpretations, J. Comput.
System Sci.27(2) (1983) 400–433.15. R. Loren, J. Li and D. B.
Benson, Deterministicflow-chart interpretations, in Proc. 3rdInt. Conf.
Entity-Relationship Approach, eds. C. G. Davis and R. T. Yeh(North-Holland, Amsterdam, 1983), pp.
421–439.