Posted December 1999

Inventory and monitoring of shrimp farms in Sri Lanka by ERS SAR data, Part 4

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

3. Results

The methodology described in the previous chapter, with interpretation keys refined after the field verification of the preliminary results, was applied to the ERS SAR data of the study area acquired in 1996, 1998 and 1999. As inventory and monitoring of shrimp farms were the objectives of the study, the maps produced show only four classes, namely: 1. water bodies (lagoons, canals, creeks); 2. shrimp farms occurring up to 18 April 1996; 3. expansion of shrimp farms up to 16 October 1998; and 4. expansion of shrimp farms up to 5 March 1999.

To facilitate field use, seven maps at 1:50 000 scale with UTM grid have been prepared. A union sheet at 1:250 000 scale shows the entire study area and the relative position of the seven larger scale maps.

Fig. 22. Shrimp farms map of the Seguwantiyu test site

Fig. 23. Shrimp farms map of the Uppadaluwa test site

Figures 22, 23 and 24 show the inventory and monitoring of the expansion of shrimp farms at the three test sites. Tables 6, 7 and 8 quantify the results for each test site. Finally, Table 9 shows the comprehensive results of the mapping of the shrimp farms in the three test sites and in other portions of the study area.

Fig. 24. Shrimp farms map of the Dutch Canal test site

For the immediate use of the results by the FAO project TCP/SRL/6712, maps have been converted to IDRISI files and additional information such as roads, railroads and other reference points has been added.

Figures 22 and 24 clearly define the enormous expansion of shrimp farms in the Seguwantiyu and Dutch Canal test sites. Conversely, Fig. 23 shows a static situation in the Uppadaluwa test site, probably because opportunities for the expansion of shrimp farms were few because of their already high density in this area on or before April 1996.

The cumulative results in Table 9 indicate the rapid expansion of the shrimp farm industry in Northwestern Sri Lanka, which has increased its area coverage by 44.08 percent in less than three years.

4. Discussion

The methodology developed in support of TCP/SRL/6712 and field tested in the study area in Sri Lanka has proven to be reliable and very accurate. As far as we know, this is the first time that SAR imagery has been employed in this way. As indicated in the previous chapter, the field verification of location and occurrence of shrimp farms at 32 sites identified through ERS SAR images showed an 86 percent positive identifications. The calibration of the interpretation keys resulting from this field verification definitely increased the accuracy of the approach, as it has been possible to eliminate some potential misinterpretations of SAR data. It is thus estimated that the final accuracy of the methodology described in this report is more than 90 percent. Thus, the most recent estimate (1999) of shrimp pond surface area in Northwestern Sri Lanka is 8846.05 ha ± 885 ha.

Inventory and monitoring of shrimp farms are essential tools for decision-making on aquaculture development, including regulatory laws, environmental protection and revenue collection.

There are two main advantages to employing SAR for shrimp farms inventory and monitoring. The first is timeliness. Our results indicate that shrimp farming is growing at a very rapid rate in north-western Sri Lanka and that the surface is much more extensive than reported by Funge-Smith (1998). The second, an important advantage over traditional surveys, is that the resulting digital radar maps can be incorporated into an existing GIS. Once incorporated into the GIS, the shrimp pond locations can be evaluated in terms of a number of characteristics of site suitability and also with regard to prior uses of the land. In this way the development of shrimp farming can be planned and regulated in a more rational way than is possible without such information. In this regard, it is important to note that such information is of use not only to government, but valuable also to associations of commercial shrimp farmers whose underlying purpose is to maintain a dependable supply of good quality products at competitive prices.

The need for shrimp farm mapping is both qualitative and quantitative. In this regard, the results of this pilot study, reviewed in the preceding chapter, show that the location of commercial shrimp farms can be accurately obtained, and their collective size estimated with satisfactory results. It is sometimes difficult to estimate the area coverage of individual, small sized shrimp farms, but it is generally possible to estimate with good approximation the area coverage of a cluster of shrimp farms.

Once the potential of ERS SAR data for shrimp farms mapping has been tested and verified, it is necessary to perform a cost/benefit analysis of the entire procedure to assess its practical applicability. In this particular case, Table 10 indicates costs and time associated with SAR mapping of shrimp farms, obtained from this study.

As indicated, a ERS SAR scene covers 100 x 100 km; its cost is of Euro 1 400 (approx. US$1 530). This cost is independent of the size of the study area, as no subscenes can be acquired.

Once the first SAR inventory of shrimp farms in a given area is completed, its update on a routine basis (i.e.: once a year) is an easy task. SAR provides both timeliness and flexibility because of its independence from weather conditions on the ground. Thus, in theory, an update can be obtained by ordering the acquisition of an image on a month’s notice.

In fact, the most time- and money-consuming task, i.e. the calibration and validation of the methodology, is performed once and for all in the inventory phase. Thus, ground checking can be reduced to a bare minimum, and only changes in land use should be assessed and quantified. Table 11 shows costs and time needed for monitoring the expansion of shrimp farms.

The image processing and interpretation times described in Table 10 and 11 have been obtained by a trained remote sensing professional with experience in radar image analysis.

Although hardware (PC-based digital imagery analysis systems) and software (ERDAS 8.3 or equivalent) are now usually available in remote sensing agencies and laboratories, the methodology used in this study implies a good background in imaging radar theory and a considerable practice in handling and processing SAR data; both requirements are not common knowledge at present. However, the report provides detailed examples of SAR imagery interpretation and a clear sequence of actions, thus it can be considered as a case of technology transfer as well.

Possible improvements and present constraints

All ERS SAR data used in the present study were acquired in descending orbit, thus the SAR cross-track direction always had the same relative direction vis-à-vis the longer axis of dykes bordering shrimp ponds. Thus, the 1996, 1998 and 1999 images always show in particular evidence the same group of dykes.

Conversely, using two sets of SAR data, one from a descending and one from an ascending orbit, the shrimp farms would be "illuminated" from two different directions: each image would show a different set of dykes, complementing each other's information. Applying the same methodology to such a data set would certainly greatly increase the dyke’s discernibility and consequently improve mapping of shrimp ponds.

Unfortunately, at least over Sri Lanka, the number of SAR acquisitions during ascending orbits is very limited, as other ERS sensors are generally active during these orbits; it was thus impossible to study our area with data from both ascending and descending orbits.

Further, as data acquisition from non-ESA receiving stations is based on various types of agreements, we discovered that the recording of a particular SAR scene, indicated as possible in the ESA listing, does not necessarily take place. Long processing time, usually a month or more, from data acquisition to delivery to user in georeferenced format (GEC) and the impossibility to have an indication of data quality if not after the processing of a particular scene has been requested (a Russian roulette scenario), are the main constraints of working with ERS SAR data.

On the other hand, we believe that SAR data are unique for mapping shrimp farms, not only for their inherent all-weather capabilities, but mainly because the backscatter from surrounding dykes allows for recognition and separation of shrimp ponds from all other water-covered surfaces. Sensors working in the visible and near-to-mid infrared portions of the electromagnetic spectrum, such as Landsat TM, SPOT, IRS, permit clear identification of industrial shrimp farms only. Artisanal shrimp farms, with their small size and irregular shape, may be easily confused with other water covered surfaces such as flooded rice paddies, etc. In addition, the main limitation of these sensors is that the study area is clearly visible only in cloud-free days; a serious problem, as shrimp farms are located in tropical and sub-tropical areas.

In the context of government aquaculture development policy, much attention needs to be focused on the identification and monitoring of the expansion of shrimp farms. Thus, the availability of an accurate, fast and, mainly, objective methodology that also allows the observation of remote areas, assumes a great value. The methodology is also economically viable, as the value of shrimps more than justifies an accurate inventory and monitoring of the development of the farms.

As indicated, some constraints occur at present, such as the scarcity of SAR data over some areas. However, this difficulty could be overcome by utilising SAR data acquired from other satellite systems (JERS, RADARSAT).

Finally, a sound technology transfer programme on SAR data handling is recommended to acquaint potential users in Fisheries Departments and Remote Sensing Agencies of concerned countries on the routine use of the methodology and its associated tools, such as Geographic Information Systems (GIS) and Global Positioning Systems (GPS).

References

Barber D.G, Hochheim K.P., Dixon R., Mosscrop D.R., McMullan M.J.,1996, The role of earth observation technologies in flood mapping: a Manitoba case study, Canadian Journal of Remote Sensing, Vol. 22, No. 1

Beaulieu N. et al., 1998, The contribution of RADARSAT - 1 SAR imagery to monitor land use in coastal areas of Costa Rica and Nicaragua. Proceedings of ADRO Final Symposium, Montreal

Dallemand J.F., Lichtenegger J., Raney R.K., Schumann R. 1993. Radar Imagery: Theory and Interpretation. Lecture Notes. RSC Series n. 67, FAO

FAO, 1998. Report of the Bangkok FAO Technical Consultation on Policies for Sustainable Shrimp Culture. Bangkok, Thailand, 8-11 December 1997. FAO Fisheries Report No. 572

Frost V.S., Stiles J.A., Shanmugan K.S. , Holtzmann J.C., 1982, A model for radar images and its application to adaptive digital filtering of multiplicative noise. IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI-4

Funge-Smith S. J., 1998. Disease prevention and health management in coastal shrimp culture. TCP/SRL/6614(A). Consultancy mission report. FAO

Gilabert M.A., Melia J.,1990, Usefulness of the temporal analysis and the normalized difference in the study of rice by means of Landsat-5 TM images: identification and inventory of rice fields. Geocarto International, Vol. 4

Hen L.L., Melak J.M., Filoso S., Wang Y.,1995, Delineation of inundated area and vegetation along the Amazon floodplain with the SIR-C synthetic aperture radar.

IEEE Transactions on Geoscience and Remote Sensing, Vol. 33 No. 4

Lee J.S., 1980, Digital image enhancement and noise filtering by use of local statistics. IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI-2

Lee J.S., 1981, Refined filtering of images noise using local statistics, Computer Graphics and Image Processing, No. 15

Li C., 1988, Two adaptive filters for speckle reduction in SAR images by using the variance ratio, International Journal of Remote Sensing, Vol. 9, No. 9

Lillesand T.M., Kiefer R.W. ,1987, Remote Sensing and Image Interpretation. 2^nd ed. John Wiley & Sons Ed., New York.

Lopes A., Nezry E., Touzi R., Laur H., 1993, Structure detection and statistical adaptive speckle filtering in SAR images, International Journal of Remote Sensing, Vol. 14, No. 9.

Nagao M., Matsuyama T., 1979, Edge preserving smoothing, Computer Graphics and Image Processing, No. 9

Panigrahy S., Chakraborty M., Sharma S.A., Kundu N., Ghose S.C, Pal M.,1997, Early estimation of rice area using temporal ERS-1 synthetic aperture radar data, a case study for the Howzah and Hughly districts of West Bengal, India. International Journal of Remote Sensing, Vol. 18 No. 8

Piyasena, G., 1996, Brackishwater aquaculture and management. In: Morris, M.J., Masammichi-Hotta and Atapattu, A.R. (eds.). Report and Proceedings of the Sri Lanka-FAO

National Workshop on Development of Community Based Fishery Management. Colombo, 3-5 October, 1994

Pouncey, R., Schrader S. , 1996, ERDAS IMAGINE Radar Manual, Version 8.2, ERDAS Inc. USA

Ramsey E.W. III.,1995, Monitoring flooding in coastal wetlands by using radar imagery and ground based measurements, International Journal of Remote Sensing, Vol. 16 No. 3

Richards, J.A., 1993, Remote Sensing Digital Image Analysis: An Introduction. 2^nd ed. Springer-Verlag, Berlin

Soo Chin Liew, Suan-Pheng Kam, To-Phuc Tuong, Ping Chen, Vo Quang Minh and Hock Lim., 1998, Application of multitemporal ERS-2 synthetic aperture radar in delineating rice cropping systems in the Mekong River Delta, Vietnam.

IEEE Transactions on Geoscience and Remote Sensing, Vol. 36 No. 5

Survey Department of Sri Lanka, 1987, Topographic maps: scale 1:50 000, Series A.B.M.P., Edition 1, Sheets 29 Kalpitiya, 34 Puttalam, 40 Battulu Oya, 46 Chilaw, 52 Kochchikade

Touzi R., Lopes A., Bousquet P., 1988. A statistical and geometrical edge detector for SAR images. IEEE Transactions on Geoscience and Remote Sensing, No. 26

Wang Y, Koopmans B.N., Pohl C.,1995. The 1995 flood in the Netherlands, monitored from space, a multisensor approach. International Journal of Remote Sensing, Vol. 13, No. 3

Wijepoonawardena, P.K. and Siriwardena, P.P. , 1996. Shrimp farming in Sri Lanka: Health management and environmental considerations. In: Subasihghe, R., Arthur, J. and Shariff, M. (eds.) Health management in Asian aquaculture. Proceedings of the Regional Expert Consultation on Aquaculture Health Management in Asia and the Pacific. Serdang, Malaysia, 22-24 May 1995. FAO