Geoinformation, monitoring and assessment Environment

Posted December 1999

Inventory and monitoring of shrimp farms in Sri Lanka by ERS SAR data, Part 1

by
Carlo Travaglia
FAO Environment and Natural Resources Service
James McDaid Kapetsky
FAO Inland Water Resources and Aquaculture Service
Giuliana Profeti
Consultant, Digital Image Processing
Environment and Natural Resources Working Paper No.1
FAO, Rome, 1999


For a printed copy of this publication, e-mail: Carlo.Travaglia@fao.org


Abstract

Inventory and monitoring of shrimp farms are essential tools for decision-making on aquaculture development, including regulatory laws, environmental protection and revenue collection. In the context of government aquaculture development policy, much attention has to be focused on the identification and monitoring of the expansion of shrimp farms. Therefore, the availability of an accurate, fast and mainly objective methodology that also allows the observation of remote areas assumes a great value. The satellite remote sensing approach is also economically viable, as the value of shrimps more than justifies an accurate inventory and monitoring of the development of the farms.

SAR data are unique for mapping shrimp farms, not only for their inherent all-weather capabilities, very important as shrimp farms occur in tropical and sub-tropical areas, but mainly because the backscatter from surrounding dykes allows for recognition and separation of shrimp ponds from all other water-covered surfaces.

The study is based on interpretation of SAR satellite data and a detailed image analysis procedure is described. Although hardware and software needed for the extraction of useful information from SAR data are currently available at most remote sensing laboratories, good knowledge in imaging radar theory and practice in handling and processing SAR data are not. The report aims at the necessary technology transfer for an operational use of the approach indicated in other similar environments.

The methodology reported in this study has been tested under operative conditions in north-western Sri Lanka with the support of FAO project TCP/SRL/6712. The mapping accuracy achieved for shrimp farms, after field verification of preliminary results and refining of interpretation keys, is estimated to be more than 90 percent.


Acknowledgements

The authors are greatly indebted to all who assisted in the implementation and completion of this study by providing information, advice and facilities.

The participation in and contribution to the field verification exercise of the following officers of the Sri Lanka National Aquatic Resources Agency - NARA - is gratefully acknowledged: Mr A. Gunaratne, Mr A. Athukoria and Dr S. Jayamanne.

The study has been conducted in the framework of the European Space Agency - ESA - Third Announcement of Opportunity for the Exploitation of ERS Data. The provision by ESA of the necessary ERS-SAR satellite data greatly facilitated its implementation. Special thanks are directed to Mr Gunther Kohlhammer, ERS Mission Manager, and to Ms Fabrizia Cattaneo and her colleagues of the ESA-ERS Help/Order Desk, for their constant support throughout the project.


1. Introduction

1.1 Perspective on shrimp farming

Global farmed shrimp and prawn production amounted to 932 000 t in 1995 compared with some 170 000 t in 1984. Production has plateaued since 1991. Eighty-seven percent originated in Asia.

Constraints on the further development of shrimp farming

Despite the rapid growth of the industry, there have been setbacks due to diseases and due to the growing awareness of the environmental and social impacts of shrimp farming. Much of the debate has focused on the sustainability of shrimp farming. There is a trend for discussion on principles, the development of guidelines and the need for better management practices (FAO, 1998). It has been recognized that shrimp farming can be made more sustainable. Impacts can be reduced in a number of ways including through better regulatory and planning processes at State level. Key considerations are the siting of shrimp farms and monitoring their development.

1.2 State of shrimp farming in Sri Lanka (Northwestern coast)

Shrimp farming began in Sri Lanka in the early 1980s. Farming of the black tiger prawn, Penaeus monodon, was a successful and lucrative venture until major disease outbreaks occurred in the late 1980s (Wijepoonawardena and Siriwardena, 1995). Although the main cause for these outbreaks was thought to be the introduction of an exotic viral pathogen, uncontrolled proliferation of farm operations and related aquatic environmental implications appear to have made a direct contribution. Similarly, lack of planned development was identified as one of six constraints on shrimp farming and suitable locations in the NW were said to be almost saturated (Piyasena, 1996).

Currently, the shrimp culture sector in Sri Lanka is facing many of the problems previously encountered in other countries. The technical knowledge base of the majority of the shrimp farmers is very low and becoming increasingly so, as more small-scale farms are developed. Shrimp farming is still relatively small scale in Sri Lanka with a total area of approximately 2600 ha according to Funge-Smith (1998). These farms can be broken down by surface area as shown in Table 1.

The majority of the unregistered farms have encroached into reserved areas and are small farm operations - their size generally being below 2 -3 ha. Farms over 4 ha are required to fulfil Initial Impact Investigation or Environmental Impact Assessment; small farms are exempted from this and this has contributed to the proliferation of small illegal operations.

Table 1: Size and relative occurrence of shrimp farms in Sri Lanka (Funge-Smith, 1998)
Area (ha) % of total area

> 20

32

10 - 20

9

4.5 - 10

15

2 - 4.5

10

< 2

6

Unregistered farms

28

Small farms are usually owner operated and do not have a high level of technical input. There appears to be some form of technical service available whereby farmer groups are visited by local consultants. Large farms have well trained managers - often with overseas experience.

It is the lack of accurate information available to the farmers that results in inappropriate farming techniques, disease and production losses.

Plans are currently underway to develop shrimp farms in other parts of the country, utilizing seawater abstracted from the sea (full salinity) and not the brackishwater usually found in the Northwest Province lagoon systems. Culture in full strength seawater is possible providing a regular water exchange regime is practiced. Alternatively culture should only take place during the monsoon or rainy season, to prevent excessive salinity in the ponds. Before these developments proceed further, it is important to establish the principle factors underlying the disease problems in the Sri Lankan shrimp industry.

There is also the consideration that the Northwest Province may provide much of the country’s broodstock, and the development of shrimp farms in this area would certainly increase the risk of contamination of the broodstock supply

1.3 Objective of the study

The main objective of this study has been to demonstrate, under operative conditions, and in support of TCP/SRL/6712 "Revitalization and Acceleration of Aquaculture Development" the usefulness of high resolution SAR data for the inventory and monitoring of shrimp farms, in view of developing and field testing adequate methodologies for future use in similar environments elsewhere.

To achieve this objective in the Sri Lanka case study, ERS SAR satellite data, acquired in 1996, 1998 and 1999, have been processed for shrimp farms inventory, and the resulting information has been compared to substantiate changes and trends in the development of shrimp farms.

Basically, the Sri Lanka Government requires up to date information on the spatial distribution of shrimp farms in order to enforce development regulations and in order to ensure a productive environment for shrimp farming with the least impact on other uses of land and water resources.

This study is timely because it coincides with two related activities, one of which is the zoning of aquaculture in Sri Lanka, and the other on disease prevention and health management in shrimp farming, both of which are FAO Technical Cooperation Programme projects.

1.4 Study area

Fig. 1. Study area (geographic reference grid

The area under examination occurs along the western coast of Sri Lanka, north of Colombo (Fig. 1), approximately from 8°23' to 6°50' North Latitude (Table 2), covering a strip 30 km wide at its maximum and 172 km long. It is a coastal flatland, characterized by a series of inland lagoons and lakes (from North to South: Puttalam Lagoon, Mondal Lake and Chilaw Lake) connected by channels, meandering rivers and creeks.

The vegetation existing prior to the development of shrimp farms is reported in the 1984 topographic maps as being mainly composed of low forest, grassland and mangroves; the main agricultural crops were rice and coconuts.

Shrimp farms, which cover extensive portions of this area, can be subdivided into two major groups: industrial and artisanal.

Industrial shrimp farms (Fig. 3) cover usually large areas with individual shrimp ponds arranged in an orderly way, rectangular in shape and all of the same size, with average dimensions of 30 x 50 meters. Industrial shrimp farms are usually surrounded by high walls or fences, in consideration of the considerable value of their product. Conversely, artisanal shrimp farms cover relatively small areas, the shrimp ponds are of various sizes and, often, of various shapes, and dykes surrounding individual ponds are less prominent than that occurring in industrial shrimp farms. Further, the shape of the complex is somewhat irregular as it exploits natural areas along creeks and canals.

The study area is fully covered by two ERS SAR frames (Table 3). For monitoring purposes four different data sets were studied, acquired respectively by ERS 1 on 18 April 1996 and by ERS 2 on 3 July and 16 October 1998 and 5 March 1999.

The ERS SAR GEC (geocoded) data have a spatial resolution of 12.5x12.5 meters. Each scene covers an area of 100 x 100 km. Nominal frequency of data acquisition over a given area is of 35 days. The adopted projection is UTM (Zone Number 44N, spheroid and datum WGS84).

Table 2. Coordinates of the study area
UTM 44N, WGS84 Upper left corner Lower right corner

Northing (y)

928340

755690

Easting (x)

352219

382007

Table 3. Satellite data used in the study
ERS SAR GEC Orbit/Frame Acquisition date

ERS 1

24885/3465 and 3447

18/04/1996

ERS 2

16735/3465 and 3447

07/03/1998

ERS 2

18238/3465 and 3447

16/10/1998

ERS 2

20242/3447

05/03/1999


Fig. 2. False colour composite of three ERS SAR images of the study area.
Red: 18/04/96, green: 03/07/98, blue: 18/10/98


2. Methodology

2.1 Shrimp farm mapping by imaging radar

The identification of shrimp farms (Fig. 3) on SAR images is based on several elements: the signal received from the water surface of the ponds and from their surrounding dykes, the shape of the individual ponds, the pattern of groups of ponds and the relative direction of the dykes vis-à-vis the incoming radar beam. The location of shrimp farms is also typical, thus the analysis of their position and of the former land use of the area is necessary to verify the identification. All these elements are discussed in the following sections.


Fig. 3. Industrial shrimp farm (south of Chilaw Town)

SAR signal of ponds and dykes

Water-covered surfaces, such as shrimp ponds, are easily identifiable on SAR images due to their characteristically low response, resulting in a very dark gray level. The reason for the low response of water-covered surfaces, as well as of other very smooth surfaces, lies in the wavelength employed and in the peculiar acquisition geometry of SAR images.


Fig. 4. Interaction of a radar beam with a smooth surface


Fig. 5. ERS SAR, 18/04/96: shrimp ponds at Vidatamunai, Puttalam Lagoon

The ERS SAR system operates in the C-band (l @ 5.6 cm), and the angle between the perpendicular to the imaged surface and the direction of the incident beam is approximately 23°. According to the Rayleigh criterion, a surface is considered smooth if the mean height of its structure is smaller than the incident wavelength. Small water bodies generally satisfy this condition, and according to Fresnel's Law, they reflect all the incoming radiation at an angle equal to the incidence angle, thus away from the satellite antenna (Fig. 4). Consequently, the signal received from a small water body approaches zero, and the water body is visualized as a dark gray surface (Fig. 5).

The identification of low-reflecting surfaces in an image is a simple task. Unfortunately, a low response is obtained by all calm water bodies, such as small lakes, reservoirs and flooded areas. The problem in identifying shrimp ponds on SAR images is then focused on separating the various kinds of low-reflecting surfaces by means of their peculiar characteristics of shape and pattern, and the characteristics of their neighborhood.

Shrimp ponds, unlike other water bodies, are surrounded by dykes. A dyke is an earthen wall whose thickness ranges approximately from half a meter to several meters, and whose elevation from the water surface is at the most a meter. The surface of a dyke can be considered smooth; the radar beams interaction with the dyke depends on their relative position.

If the long axis of a dyke is perpendicular to the cross-track direction, the resulting radar beam paths are shown in Fig. 6. The radar beams that reach the upper part of the dyke are reflected in the direction opposite to the sensor, giving no return signal.

The beams which reach the vertical part of the dyke are reflected towards the water surface of the pond first, then back to the sensor, generating a very strong return signal.

The same happens to the beams which hit the water surface near the dyke: they are first reflected towards it, then back to the sensor. If part of the dyke is inclined, the beams, which hit it, are reflected directly into the sensor, generating an even stronger return signal.


Fig. 6. Interaction of radar beams with a dyke

Fig. 7. Return signal as a function of the angle between a dyke and the cross-track direction

Another consequence of this peculiar imaging geometry is that the position of the dyke in the image appears to be shifted towards the sensor (layover effect: Fig. 6).

The value associated with an ERS SAR image pixel (digital number, DN) is given by the average signal received from the corresponding surface on the ground. When a portion of the ground contains a small but highly reflective feature, the average signal is almost equal to the imaged feature, which appears then to be as big as the entire pixel. The object has "saturated" the pixel's value. This effect is caused by the dykes as well. The dykes are long but narrow structures, thus the DN value of a pixel is made up by the return signals from both a portion of the dyke and its surroundings. However, a dyke's return signal is so strong in comparison with the low signal of the water-covered ponds that the resulting value of the pixel is very high.

The multiple reflection effect spreads the high return signal to the surrounding pixels, increasing their values as well. Therefore the dyke, when perpendicular to the incoming radar beam, is easily identified in the SAR image as a white stripe (Fig. 5) composed of several adjacent pixels, thus thicker than in reality.

Paddy fields are surrounded by smaller and lower dykes, which do not generate the same effect and consequently are not identifiable in SAR images.

The multiple reflection effect discussed above and its consequences take place on the portion of the dyke that directly faces the incoming radar beam. Consequently, the intensity of the return signal depends on the angle between the cross-track direction and the direction of the longer side of the dyke (Fig. 7).

When a dyke is perpendicular to the cross-track direction, all the corresponding pixels have a very high return signal, as discussed previously.

When the angle between the longer side of the dyke and the cross-track direction decreases, the multiple reflection effect decreases as well. The resulting signal is weaker, and the dyke's pixels are displayed as darker gray in the SAR image.

Finally, when a dyke is parallel to the cross-track direction, it barely interacts with the radar beams, and is thus characterized by very dark gray tones in the image.

It has been observed previously that the dykes perpendicular to the cross-track direction appear to be thicker than they really are, due to the saturation effect. Dykes positioned at smaller angles from the cross-track direction have lower return signals; the saturation effect decreases and their apparent thickness decreases as well. When the angle between a dyke and the cross-track direction is zero, i.e. the dyke is parallel to the cross-track direction, it is barely visible in the image.

Both of these effects are evident in the ERS 1 SAR image of 18 April 1996 (Fig. 5).

The average DN values and thickness of 35 dykes in this image have been plotted in Fig. 8 and Fig. 9. The two graphs show the effects explained above.

The along-track direction of the satellite during the acquisition was 192.5 deg. North.

Dykes parallel to the along-track direction (i.e.bearing 12.5 deg. North) look very bright and large in the image. At increasing angles between the dykes and the along-track direction, the dykes appear grayer and thinner. Finally, dykes perpendicular to the along-track direction (i.e. bearing 102.5 deg. North) are not visible in the image.


Fig. 8. Dykes'average values (DN) vs. bearing (deg.)

Fig. 9. Dykes'average thickness (m) vs. bearing (deg.)

To: Inventory and monitoring of shrimp farms in Sri Lanka by ERS SAR data, Part 2



Back to Top FAO Homepage