Runoff Analysis Using Satellite Data for Regional Flood
Total Page:16
File Type:pdf, Size:1020Kb
Kochi University of Technology Academic Resource Repository � Runoff Analysis using Satellite Data for Regiona Title l Flood Assessment - Spatial and Time Series Bia s Correction of Satellite Data - Author(s) PAKOKSUNG, Kwanchai Citation 高知工科大学, 博士論文. Date of issue 2016-09 URL http://hdl.handle.net/10173/1422 Rights Text version ETD � � Kochi, JAPAN http://kutarr.lib.kochi-tech.ac.jp/dspace/ Runoff Analysis using Satellite Data for Flood Reginal Assessment: Spatial and Time-series Bias Correction of Satellite Data Kwanchai Pakoksung A dissertation submitted to Kochi University of Technology in partial fulfillment of the requirements for the degree of Doctor of Engineering Graduate School of Engineering Kochi University of Technology Kochi, Japan September 2016 i ii Abstract Floods are one of all the major natural disasters, affecting to human lives and economic loss. Understanding floods behavior using simulation modelling, of magnitude and flow direction, is the challenges of hydrological community faces. Most of the floods behaviors are depended on mechanism of rainfall and surface data sets (topography and land cover) that are specific for some area on a ground observation data. Remote sensing datasets possess the potential for flood prediction systems on a spatially on datasets. However, the datasets are confined to the limitation of space-time resolution and accuracy, and the best apply of these data over hydrological model can be revealed on the uncertainties for the best flood modelling. Furthermore, it is important to recommend effective of data collecting to simulate flood phenomena. For modelling nearby the real situation of the floods mechanism with different data sources, the difficult task can be solved by using distributed hydrological models to simulate spatial flow based on grid systems. Therefore, the objective of this dissertation is to contribute the correction and evaluation of remote sensing sources for flood prediction through basin scales, and application of the model to demonstrate the approach of flood risk estimation method on small area. It also aims to present and create general bias methodology for runoff analysis using the remote sensing data sources to model the flood simulation. The geographical point of this study is the Yoshino river basin in Japan and the Upper part of Nan river basin in Thailand. Firstly, the modification and implementation of distributed hydrological modeling as the RRI model are included by VOXEL model for convenient on input and output and the GPU coding for speedup. The VOXEL model was used to integrate the input data the watershed data (DEM and Land cover) and rainfall data (spatial and temporal) and the output data runoff and inundation depth (spatial and temporal). The GPU on NVIDIA CUDA was setup for speedup about 2.6x on the complex terrain. The accuracy assessment, bias correction method evaluation, and impact of flood analysis validation were shown in the second part based on three components of flood mechanism, DEM, satellite land cover, and satellite based rainfall. The six DEM sources (GSI-DEM, ASTER GDEM, SRTM, GMTED2010, HydroSHEDS, and GTOPO30) evaluated by the referent elevation points (GCP) that are used to estimate bias correct coefficient based on spatial linear transformation, and then the data are driven by the distributed hydrologic model (RRI) to reveal the impact of the topography sources. The GSI-DEM was a high accuracy among the five DEMs and the correction algorithm could improve the accuracy responding with the coarse resolution DEMs (HydroSHEDS and GTOPO30), while the high resolution (GSI-DEM, ASTER GDEM, SRTM, and GMTED2010) had a small sensitivity. In the Shikoku Island, ASTER DEM was suitable for runoff simulation, have estimated from stereo matching. SRTM presented a performance for runoff and inundation simulation in the Nan river basin, have explored from radar laser scan with Shuttle. MODIS product outperformed iii AVHRR products that Manning’s coefficient of MODIS also showed higher performance. The MODIS roughness also presented higher performance evaluated from the hydrological modeling results. The rain gauges were interpolated into grid system with five algorithms, Inverse Distance Weight (IDW), Thiessen Polygon (TSP), Simple Kriging (SKG), Ordinary Kriging (OKG), and Surface Polynomial (SPL). The IDW outperformed as high performance algorithm in the Shikoku area that represented with the dense rain gauge network area, while the sparse rain gauge network area was the Nan river basin that the SKG was suitable algorithm. GSMaP and GPV showed the high accuracy for the Shikoku in Japan, while CMORPH outperformed among other sources in the Nan area in Thailand, on the international sources. GPV and GSMaP in Japan and GPM in Thailand as the high resolution showed the highest performance on runoff simulation, while low resolution was TRMM. The five algorithms (Mean ratio, Geometrics transformation, Linear transformation, Data assimilation, and Quantile mapping) and two schemes (Temporal and Spatial) were evaluated only GPM and TRMM in the Nan river basin, Thailand. The three algorithms (Linear, Geometrics, Data assimilation) on the spatial scheme showed the high performance, resulting from runoff validation. Finally, the application of the remote sensing data sets on flood forecasting and flood risk assessment was demonstrated. The first approach, the river basin scale simulation was used to define as the boundary condition of small area to simulate a high resolution of flood map. The second approach based on small area results, the flood risk assessment was consisted by hazard and vulnerability data. In this task, the streamflow for estimating the flood risk map was the main point for proposing. iv Acknowledgments The author would like to express my sincerely thanks to advisor professor Masataka TAKAGI for his overall supports to form this research fulfillment. With his strong supports, advising and guidance; this research is successfully done. Continuously, I would like to add my gratitude to supervisory committee members: Prof. Seigo NASU, Prof. Yoshiro KAI, Associate Prof. Takashi GOSO, and Associate Prof. Masayuki MATSUOKA for their valuable suggestions and comments to improve my research works. The author wishes to acknowledge Prof. Lawrie Hunter for English research writing. A great supports for my study is the Special Scholarship Program (SSP) of Kochi University of Technology for doctoral scholarship program in Japan. The author would like to thank all staffs of International Relationship Center for guidance on living, and many other helps. Furthermore, Takagi laboratory member are to helps and share their times and knowledge in my endurance on the research. The author would like to thank Dr. Nattakorn Bongochgetsakut for his expert suggestion in developing the GPU model, and Dr. Pongsak Suttinon for his expert suggestion in understanding the economic model. Finally, the author would like to thank my parents, and my wife and my sons, with a beautiful life. v vi Table of Content Pages Title page i Abstract iii Acknowledgments v Table of contents vii List of figures xi List of tables xvii 1. Introduction 1 1.1 Motivation 1 1.2 Overview of the thesis 7 1.2.1 Research problem and objectives 7 1.2.2 Scope of the dissertation 8 1.3 Outline of the dissertation 9 1.4 Study area 12 1.4.1 Shikoku Island, Japan 12 1.4.2 Nan river basin, Thailand 14 2. Hydrological modeling 15 2.1 Introduction 15 2.2 Hydrological model by the RRI model 18 2.2.1 Concept of RRI model 18 2.2.2 Components of the RRI model programming 19 2.3 VOXEL model assisted RRI model 21 2.3.1 VOXEL model 21 2.3.2 Application for the RRI model 22 2.4 Model setup 23 2.4.1 Setup parameter 24 2.4.2 Performance statistical 25 2.4.3 GPU on the RRI model 27 2.5 Modeled results 30 2.5.1 VOXEL model of watershed and rainfall data for the RRI model 30 2.5.2 Simulated runoff and inundation map 31 2.5.3 Speedup of the GPU-RRI model 34 2.6 Conclusion 39 vii Table of Content (continued) Pages 3. Bias correction of DEM sources and their effect on estimation of runoff and 41 inundation area 3.1 Introduction 41 3.2 Data and Methodology 43 3.2.1 Digital Elevation Model sources and Reference elevation points 43 3.2.2 Accuracy assessment 48 3.2.3 Bias correction 52 3.2.4 Hydrological simulation 53 3.3 Results and Discussion 54 3.3.1 Geomorphological property of DEMs 54 3.3.2 Accuracy assessment of DEMs 57 3.3.3 Effect of terrain morphology in the DEMs accuracy 65 3.3.4 Effect of land cover in the DEM accuracy 69 3.3.5 Evaluation of the river network 71 3.3.6 DEMs bias correction 73 3.3.7 Effect of DEM sources on runoff simulation: Shikoku Island, Japan 78 3.3.8 Effect of DEM sources on runoff simulation: Nan river basin, Thailand 81 3.3.9 Effect of DEM sources on flood area simulation: Nan river basin, Thailand 84 3.4 Conclusion 85 4. Estimation of surface roughness based on land cover datasets and their effect on 88 Distributed Rainfall-Runoff simulation 4.1 Introduction 88 4.2 Data and Methodology 90 4.2.1 Land cover based satellite and Land cover referent map 90 4.2.2 Manning’s n coefficient estimation 94 4.2.3 Hydrological simulation 95 4.2.4 Accuracy assessment 96 4.3 Results and Discussion 97 4.3.1 Accuracy assessment of Land cover based satellite data: MODIS product 97 4.3.2 Accuracy assessment of Land cover based satellite data: AVHRR product 103 4.3.3 Manning’s n coefficient based satellite sources evaluation: Shikoku Island, 108 Japan 4.3.4 Manning’s n coefficient based satellite sources evaluation: Nan river basin, 113 Thailand viii Table of Content (continued) Pages 4.3.5 Effect of Land Cover-based Surface roughness on Hydrological model 117 Results: Shikoku Island, Japan 4.3.6 Effect of Land Cover-based Surface roughness on Hydrological model 119 Results: Nan river basin, Thailand 4.4 Conclusion 122 5.