Flood disaster prediction model using Remote Sensing data and geographic information system

Flood disaster prediction model using Remote Sensing data and geographic information system

Shiro Ochi, Shunji Murai
Institute of Industrial Science University of Tokyo

Suvit Vibulsresth
National Research Council of Thailand

Abstract
This paper describes about the new method to analyze the flood flow. The rainfall in the mountainous area flows partly on the surface and partly under the ground depending on the runoff ratio which is determined by the geology, the land cover, the slope gradient etc.. the water on the slope surface flows in the course of the slope aspect. The slope gradient and slope aspect are computed from DEM. The authors have developed the flood flow model using the hydrographic theory. Runoff ratio has a influence on the flow rate, so even if the precipitation is the same the quantity of the flow becomes different. In our study the runoff ratios are assumed with the slope gradient and NVI which comes from the satellite data. The flood flow model allows a simulation study for various cases of vegetation cover conditions including deforestation, which will provide prediction of flood disaster.

Introduction
In November 1989, there occurred a serious flood disaster over NAKHON-SI-TAMARAT in southern part of Thailand. The causes of this catastrophe are considered as the following points.

An abnormal rainfall.
The deforestation in the mountainous area.
The surface geology of weathered granite.

In order to predict the flood disaster and to make flood-control, many hydrological methods have been proposed, but some of them are black-box type, or some of them need much investigation. In developing countries where deforestation is advancing and the drastic landuse changes take place almost every year; it is difficult to know the up-to-date characteristics of the basin. In this sense remote sensing data is very necessary to monitor the land cover change.

Methodology

Slope Direction

The water on the slope should flow in the course of the slope aspect. The flow model can be made from DEM (digital Elevation Model) by the following procedure. On the 3x 3 grids of DEM the center pixel has the slope direction, which has the steepest to every pixel, which has the direction around the center (Fig.1a). One unit of rainfall is given toe very pixel which has the slope direction (Fig. 1b), and the water in the pixel flows tot eh direction (fig. 1c). After that, one unit of rainfall is given to every pixel again and then flows to the slope direction (Fig. 1d. e). By continuing this process unit steady state (Fig. 1i.j), we can get the quantity of the flow and pick up the stream. Fig. 2 shows the successive image during this process which generates the drainage pattern. The thickness of the line represents the quantity of the flow. The rainfall given to each pixel is changed according to the strength of the precipitation and the runoff ration.

Fig 1 Flow model on 3x3 grids

Fig.2 Drainage pattern generation

Runoff Ratio

Many examples about the runoff couldn't be found, Table-1 shows a example in which three categories re considered - the land cover, the geology and the slope gradient. The land cover is classified to four classes - dense forest, sparse forest or arable land and barren land. The penetration represents the geology in this example. Slope gradient is roughly classified to three classes.

On the other hand, the land cover image can be generated by level-slicing of NVI (Normalized vegetation Index) which is computed with the following formula.

NVI = (I.R - R) / (1.R + R)

Fog 3 shows the land cover image in 1984 on the test area using LANDSAT - MSS data and table-2 shows the relationship between land cover and NVI. From the correspondence between table 1 and table 2, the runoff ration (=a) is assumed by the following function which contains the slope gradient (=b) and NVI (=g).

a = 0.01b- 0.37g+ 0.648

Fig. 4 shows the relationship of NVI between 1984's MSS data and 1988's TM data. As can be seen in Fig. 4 it si possible to assume the runoff ratio in 988 with considering the difference between the two sensors. So the following function gives the runoff ration in 1988.

a = 0.01b- 0.26g+ 0.629

Table 1 Example on runoff ratio

Penetration slope Land cover	Good			Medium			Bad
Penetration slope Land cover	S	G	P	S	G	P	S	G	P
Dense forest	.65	.55	.45.	.55	.45	.35	.45	.45	.25
Spare forest	.75	.65	.55	.65	.55	.45	.55	.45	.35
Glass land	.85	.75	.65	.75	.65	.55	.65	.55	.45
Barren forest	.90	.80	.70	.80	.70	.35	.45	.45	.25

Table 2 Relationship between NVI and landuse

NVI	Land cover
~ - .07	Road, Baresoil
-.07 ~ .07	Glass land, paddy field
.07 ~ .45	Spare forest
.45 ~	Dence forest

Arrival Distance and Arrival Time

The velocity in the actual flood flow changes depending on the quantity of discharge. The manning formula gives the velocity of the flow as follows

v = (R ^2/3X I ^1/2)/N

R : Hydraulics radius
I : Bed gradient
N : Roughness factor

Fig.5 The supposed sectional shape

When the sectional shape of the stream is supposed as Fig 6 the velocity increases in proportion to the third power of the quantity of the flow

u µ Q^1/3
Q : Quantity of the flow

The arrival distance is defined in this study as the distance measured along the streamline starting from each pixel to the observation point. Fig6 shows how to compute the arrival distance.

Fig.7 Arrival time

Fig 6 arrival distance

The total arrival time is given by accumulating subcomponents of the arrival time from a pixel to the neighbor which is derive from division of the arrival distance by the velocity. Fig - 7 shows the difference of the arrival time depending on the strength of the precipitation. When the rainfall have a unit of mm per hour, it tales 180 unit times for the water in the farthest pixel to arrive at the observation point in the example shown in Fig 7 when the rainfall has five mm per hour, the arrival time shortens to only 100 unit times.

Simulation
The discharge of the flow on the observation point at time (=T) is given as the sum of the rainfall on the pixels of which arrival time is less than T.

A simulation study to generate the hydrograph has been made on the following for cases.

Case 1 : Dense Forest covered all over the basin
Case 2 : Deforested all forest over the basin.
Case 3 : 1984's condition using NVI in '84 MSS data.
Case 4 : 1988's condition using NVI in '88 TM data

Fig 8 shows the hydrograph in these cases. The discharge from the deforested basin (case2) increases about 44% of that from the basin covered with dense forest (case 1). And the discharge increase of 6% is seen between 1984 and 1988 (case 3 and case 4). It has been found that the vegetation coverage in the basin has a great influence on the flood flow. The advance of the deforestation increases the maximum flow and quicken the appearance of the peak.

Conclusion
The authors have succeeded to develop the flood flow model with the following functions

The model warns the dangerousness by flood damage due to the increase of runoff ratio which may be increased by land use change like deforestation.
The model monitors the real status of landuse or runoff ratio which is updated timely with satellite imagery like LANDSAT MSS, TM or SPOT.
The model simulated the peak flow accprding to change of landuse.

Fig. 8 Hydrograph simulation

Fig 9 Drainage pattern and DTM

Fig 9 is the drainage pattern image combined with the DTM. Fig. 10 is the birdeye view image using the drainage pattern and LANDSAT TM image. As can be seen in these images, this model is useful to predict the flood disaster visibly.

References

S. Murai and R. Tateishi : image Output of Land form with use of Grid Data : SEISAN KENKYU Monthly Journal of Institute of Industrial science Vol. 30 No. 7, 1978.
Sukit Viseshisin, T. Hashimoto and S. Murai : Interactive System for Automated Generation of DT, from Existing topographic Map of Thailand, Proc. Asian conference on Remote Sensing 1988.s51-56.
Isao Ymaguchi : Sabo engineering: CHUIKYU-SHA 1985.
F-MErrit (Editor) : STANDARD HANDBOOK FOR CIVIL ENGINEERS; McGRAW - HIL BOOK COMPANY 21-22, 1968.
S. Murai and S.Ochi : A Study on Integration of Remote Sensing with Geographic Information System (No.1); SEISAN - KENKYU, Monthly Journal of Institute of Industrial Science Vo. 41 No. 8, 1989.
S. Murai and S. Ochi : A Study on Integration of Remote Sensing with Geographic Information System (No.2); SEISAN - KENKYU, Monthly Journal of Institute of Industrial Science Vo. 41 No. 8, 1989.