Use of IRS data in the production of a hydrology map of...

Use of IRS data in the production of a hydrology map of Sri Lanka

SHARE

Bandu Adhikari and Sarath Jayatilaka
Survey Department – Sri Lanka

1.0 Introduction
Sri Lank is situated between 79 E and 82 E longitude and 6 and 10 latitudes. It has a land area of about 65,000 sq.kms with a population of about 17 million. It is hilly in the central region sloping down to the northern and coastal regions.

The annual rainfall is between 1000 mm to 5000 mm dividing the country into three zones – the wet zone in the south-west, the intermediate in the south-west including the central hills and the dry zone in the rest of the country.

Land forms a basic resource and a fundamental component in the national development strategies. Reliable information about land and natural resources is a prerequisite for effective planning and land administration.

The collection of data about the spatial distribution of significant properties of the earth’s surface had long been an important part in planning activities. These information were very often displayed on maps and charts.

The Hydrology of a country plays an very important part in the eco system of the country. Hence an attempt was made to map the hydrology using the Indian Satellite data from IRS in Sri Lanka at the Centre for Remote Sensing at the Survey Department. At the time of this study the available data was from the IRS 1A of 1992.

2.0 Use of Satellite Date for feature extraction
The potential of using satellite images for feature extraction has been provided by Remote Sensing scientist for a long time. As a result of segmentation of original grey scale image is transformed to an image in which the ground features could be independently labeled. Basically for monochrome images there are two types of general processes in image segmentation. i.e. Edge detection or line following and region growing method. In this study region growing method was employed. The principal approaches in this method were based on threshold, region growing, region splitting and merging.

2.1 Spectral Characteristics
IRS 1A satellite carries three state of the art cameras using Charge Couple Devices (CCD) as detectors. The design of the cameras is based on the concept of “push broom” scanning, using Linear Imaging Self-Scanning Sensors (LISS). Each of the three cameras will provide data in four spectral bands three of which are in the visible region and the fourth is in the near infrared region.

Band Spectral range Applications
1 0.45 – 0.52 Coastal environmental studies
ChlorophyII absorption region
2 0.52 – 0.59 Green vegetation useful for discrimination ofRokks & soil for their iron content
3 0.62 – 0.68 Strong correlation with chlorophyII absorption in vegetation, discrimination of soil & geological boundaries
4 0.77 – 0.86 Sensitive to green biomass opaque to water resulting in this contrast with vegetation

3.0 Data used in the study
The data that was used in this study was from the IRS 1A data that was acquired in 1992 in BSQ format. The band 4 was selected as it was sensitive to green biomass and water, resulting in high contrast with vegetation.

4.0 Methodology
PCI software was used in Image processing in this study.

Image Enhancement was first carried out to the raw data. IRS data was enhanced by applying Histogram equalization techniques in contrast stretching. The contract stretching is the most important operation in image enhancement since it increases the contrast of the image. Feature are thereby made easier to see and distinguishable. In the histogram equalization method, it attempts to equalize the frequencies of the pixels across the 0-255 rage. This maximizes the contrast in the central part of the histogram and compressors the data in the tails of the histogram. This contrast stretched image has advantages in interpretation.

The image was then geo-referenced by establishing a mathematical relationship between the address of the pixels in the image and the corresponding points on the ground utilizing ground control points on a map. Second order polynomial was used for GCP selection. The image pixels were resample to 40m by applying neighborhood resamplying techniques.

Grey level threshold is a technique that is employed in portioning the grey levels in the image into one of two categories as those below a user selected threshold and those above. Two levels of grey level threshold were considered in this study. The first method used was visual based on interactive modification of the threshold value. The second method was the mathematically based on the image histogram.

Tiles were made from IRS data of band 4 with equal size -3376 pixels and 2906 lines. The following diagram will illustrate the tiles and their database. Each tile contain Image channel 1 (IRS Band 4), Bitmap segment 2 (Threshold value 20-80) and bitmap segment 3 (Screen digitizing)

(0,0)

TILE 1

(3376,0)

TILE 2

(0,2906)

TILE 3

(3376,2906)

TILE 4

(0,5812)

TILE 5

(3376,5812)

TILE 6

(0,8718)

TILE 7

(3376,8718)

TILE 8

5.0 Conclusion
It was found that the threshold values used in this study could not detect some of the smaller streams and very small tanks. It was also found that while threshold the sea too was coming into the threshold thus making problems. It was proposed that the sea be taken out using a mask before the image processing was done, but however due to time constrains this was not done. Due to this perhaps there was some difficulty in obtaining the correct threshold values. Hence some of the streams that are shown in the Hydrological map have actually been screen digitized. This is one of the problems that came across during the study. However if the time constraint was not there we could have overcome this problem.

It has shown that the use of satellite images is a very quick way to map the hydrology of a country since water features appear very well on these images. The normal conventional methods of ground survey and using air photography will cost more and also the time required for such a study be very much more than using satellite data. With the better resolution satellite data like the IRS 1C data and with correct threshold one could expect that Remote Sensing can be used in this type of mapping in the future with great success.