A primary estimation of surface water distribution in China

A primary estimation of surface water distribution in China

Chensheng Shi, Ryotaro Tateishi
Remote Sensing and Image Research Center
Chiba University,
1-33 Yayoi-Cho, Inage-Ku, Chiba-city, Chiba 263 Japan

Koji Kajiwaa
Institute of Industrial Science,
University of Tokyo,
7-22 Roppingi Minato-Ku, Tokyo 106 Japan

Abstract
There is an urgent need to know available water resources in every part of China in order to establish efficient usage of water resources. For this reason, we tried to develop a method for estimating the distribution of water resources using Digital Elevation Model (DEM) data and rainfall data. The inflow and outflow on every cell of mesh data have been calculated one by one, using the method of tracing rainfall water flowing on the land surface. The study mainly uses the DEM data, rainfall data, the data files of rivers network and lakes position and the amount of flowing water of the largest seven rivers in China. We have considered all the main factors which affect the distribution of water resources, while paid more attentions on the factors of rainfall and topographic expressions in our present study. The amount of flowing water of the largest seven rivers in China were used to determine the evaporative proportions of corresponding river basin. As a result of this study, a water resources distribution map of the mainland of China has been produced.

1. Introduction
In China, there is a large problem of unbalanced distribution of water resources. Besides of the seasonal unbalance of water resources distribution, there is a more malignant areal unbalance of water distribution. For the purpose of using water4 resources more efficiently, we tried to analyze the distribution of water resources primarily by using topographic data. The processing used in the study include the tracing rainfall water flowing on the land surface mainly and some other methods.

2. Study area and data

2.1 Study area
The study area spreads from 700 E to 140E0 of longitude, and from 200 N to 550 N of latidude. The files we used are 5 minute’s mesh data files and, thus, they have about 9.31 km’s grid ( exactly, the grid is about 5.345 km at 550 N latitude and about 8.435 km’s wide at 250 N latitude on the latitudinal direction. While the gird is about 9.31 km’s long at any degrees of latitude on longitudinal direction). The study area is wide extent which has 700 of longitude and 350 of latitude. And so these mesh data files have the scale of 840 pixels in one record for 420 records.

2.2. Data files
As shown in Figure I, we have used topographic data ( DEM data) , annual rainfall data, river network and lake position data in the study, and we have also used a map of Chinese river basins, for helping to make the river basin masking.

Currently, we have used DEM data chiefly, for tracing of the rainfall water flowing on land surface based on DEM data. We stressed the importance of land cover data and soil specifications for the processing of tracing flow of water on the land surface. We sill use them in further studies of this field.

Figure 1: Schematic Diagram of Processing
3. Methodology
In order to use the method of tracing rainfall water flowing on land surface for estimating the distribution of water resources, three data processing steps have been carried out.

3.1 Preparation of data files
In our study, the most important data file is topographic data, because the rainfall water will flow due to the topographic condition the directions of flowing pathway and branching flow of water are decided mainly by the elevation, exactly by the slopes of central pixel to all eight neighbors. We got topographic data from ETOPO5 data file which is a global elevation data file. This file expresses all elevation by 4260 ( pixel ) x 2160 (line ) digital values, We can consider it as a 5 minute’s mesh picture of DEM ( Digital Elevation Model ) and the data of the study area have been separated from the ETOPO5 data file.

The another important data file is rainfall data. We got it from “Grid Data of World Temperature and Precipitation” which includes annual and monthly temperature and rainfall data of all over the world. But there was a problem of resolution. Because the data files have size of 1024 ( pixel) x 480 ( line) over the world, and it is 21 minute’s mesh data. So we have to interpolate it into 5 minute’s mesh file, in order to put all files together.

In our study, we also need to know the locations of rivers and lakes. We get them from “World Data Bank 2” provided by CIA. We have selected the rivers in study areas and change it’s position into 5 minute’s mesh files coordinate and finally wrote them respectively into two data files. They are network of rivers and position of lakes respectively.

3.2 Process tracing rainfall water flowing on land surface
How does rainfall water distribute on the surface of the land? We tried to use method of tracing rainfall water flowing on land surface To answer this question. As shown in the Figural, we tracing rainfall water flowing on the land surface Using DEM and rainfall data. Using this method we can calculated the budget of rainfall water for every cell. We also used a masking file of rivers basin. From this masking file, we can know the cell is in which river basin. And so we can calculate the evaporation ratio of this river basin further.

Because the data files we used are raster files, so we can continue this processing cell by cell in order. At here, we will describe the main processing of rainfall water budget chiefly.

3.2.1. Tracing water flowing on the land surface
For calculating the slopes, a window sized 3 by 3 as Figure 2 (a) have been designed. So we can calculate the eight slopes from central cell to eight neighbors. By these slopes, we can estimate the directions of flowing path and the volumes branching to the different direction’s In the case of Figure2, we can known the flow of water in central pixel will flow on the 3,4,5,6 directions only.

Figure 2. Window for Tracing Flowing Water
The volume of water flowing on different directions can be calculated by next equation. Equation:

Here Fi is volume of flowing water along the I direction, Ft is total volume flowing out from central pixel, Si is slope of the I direction, while St is sum of slope on directions which is lower than central pixel.

The water flowing on the land surface that several branches of pathway, so the recursive processing method has been used in the program.

The main data processing flow is shown in Figure 3. When we consider one cell in the DEM data, we put the DEM value into the window and calculate slopes on 8 directions. And we get an original rainfall value of the same cell from responding position of the rainfall data file. This value is an annual rainfall value with unit of mm/year, and it can be changed into a total volume of water in this cell based on actual area of this it. As described before, we can calculate eight slopes from the central pixel to 8 directions. On tracing , we pay attention only on the directions whose elevation is lower than the central cell. The directions which has minus slope are the directions of flowing pathway of water. The amount of water flowing into next cell can be calculated by the equation.

Using this method, we continue to trace the water flowing on to next cell. We shift the window to the next cell and calculated the slope of 8 directions for finding next flowing path in new cell. At this time, we write lost volume on last cell in result file. And then we continue the processing threshold defined before. If volume of water is larger than the threshold then it must continue the processing for this pixel, and if the water volume is smaller than the threshold, the tracing on this direction must be ended and go to next direction of this cell or last cell. Like that, awe continue this processing and complete it on all directions.

3.2.2. Overlaying processing of data files
As shown in figure 3, we have used many data files in the study such as DEM data, rainfall data, network of river data and masking file. In processing, these data files were overlaid on water distribution map. First one is network of rivers, lakes. When tracing flowing water of one cell, we must search the corresponding position of river and lake files. If the cell is river or lake, we must end tracing of that pixel and write all remaining water in the same position in result file.

Water can be absorbed by soil and kept remain by plants, so we must consider these influences by the land cover and soil type. Thus, when trancing flowing water of any pixel, we must research corresponding position from the land cover and soil data files, and calculate how much water will remain in this pixel and how much water will flow down to the neighboring pixels based on the land cover and soil conditions. In this processing, we can decide what is the amount of water that will remain on this cell and what is the amount that will go to next cells which is lower than this one. We will design available model for infiltration in our further study.

Figure 3. Schematic Diagram of Estimating Water Budget
3.2.3. Using of masking file
We must consider the evaporation also, because of one part of rainfall is lost by evaporation. In our study, we have used masking file for calculating the ratio of evaporation. We draw a map of river basin by hand based on the river and lake network data files and a map of Chinese water system. Then we scan this map of river basins into the computer, produce a masking file. For one river system, we calculated the sum of rainfall using the masking file. On the other hand, we got the total annual flowing volume of the largest 7 rivers, so we can compared the total rainfall volume in one river basin and amount of flowing water of same river. Then we can get the evaporation ratio roughly. In processing of any pixel, we must research the mask file and get the lost ratio on this pixel and calculate the remain volume on this pixel.

4. Result of water resources distribution
After the processing described above, we can get a result file which shows the water distribution conditions of the study area. In figure 4, the resulted map is presented and it explains the estimated water availability of mainland China, from the highest level to the lowest level ( exact country’s boundaries are not included). The result of this study is shown here.

Actually, we can also set suitable thresholds, and change the result file into R.G.B color bands based on these thresholds. So we can produced a color map of the estimated distribution of water resources with color printer.

5. Discussion
When the rain falls on to the surface of ground, it will penetrate into soil, evaporate into air or will absorb by plants. Therefore it must consider the land cover and soil as important factors which are not considered in this study due to the lack of suitable land cover and soil data. In present study , when searching land cover and soil, we put a fixed constant of 10% as lost ratio. It means that all cells will remain the flowing water at the same percentage, even though it is not highly practical . In the future research of this field, it will intend to use land cover and soil data files, and get one return value, which is decided by the land cover and soil conditions.

The DEM data we have used in this study is 5 minute’s mesh data. So we have to interpolate other data files as same resolution. It is possible to get a more satisfactory result when the topographic data is detail. Because of that it can be conclude, the result of this method will contain a better accuracy and much details when the resolution of the data files are high. By carrying out his research, the methodology has been mainly tested and a satisfactory result have been got.

References

Sukit Viseshsin, Yoshiki Honda and Shunji Murai, Grid Data of World Temperature and Precipitation, Precipitation and Temperature Distribution of the World, Kazutaka Iwasaki, Division of Geography Faculty of Letters Hokkaido University
Susan K. Jenson, Applications of Hydrologic Information Automatically Extracted from Digital Elevation Models, HYDROLOGICAL PROCESSES, Vol. 5, 31-44 ( 1991)
P. Quinn, K. Beven, P.Chevallier and O. Planchon, The Predication of Hillslope Flow Paths for Distributed Hydrological Modeling Using Digital Terrain Models, HYDROLOGICAL PROCESSES, Vol. 5, 59-79 ( 1991)

Acknowledgement
The authors wish to express sincere gratitude to Mr. Anh, Mr. Xu Kai and Mr. Liyanage Kithsiri Perera for their helping.