Previous Page Table of Contents Next Page


Consumption in the Lower Mekong Basin as a Measure of Fish Yield


"The fishery is undervalued and so cannot fairly compete with other sectors."

Kent G. HORTLE
Mekong River Commission
Assessment of Mekong Fisheries
Component
Vientiane, Lao PDR

Simon R. BUSH
Australian Mekong Resource Center
University of Sydney
Sydney, Australia

Wild capture fisheries in river systems are usually under-recognised and undervalued in official statistics (Coates, 2002). Official estimates of the yield from inland fisheries in developing countries are generally based on larger-scale aquaculture figures and in some cases the catches from commercial fisheries from lakes or very large rivers. Trade figures are also used where available in calculating the food balances produced by the FAO. The small-scale river (including floodplain) fisheries that predominate in the Lower Mekong Basin (LMB) and in many other large river systems are not included in official statistics. The resulting official estimates of yield from the basin are often grossly underestimated.

This lack of recognition of the importance and size of the wild capture fishery has several serious consequences for sustainable development in the Lower Mekong Basin.

The fishery is undervalued and so cannot fairly compete with other sectors (e.g. hydroelectricity, irrigation, flood control) which, by changing flow patterns or water quantity or quality, may impact the wild fishery. Decisions on water management planning or projects require assessment of true costs and benefits. For example, gains in GDP through increased rice production or hydroelectricity should be evaluated against losses to the fishery. Similarly, less tangible impacts on livelihoods, nutrition and culture are not fairly accounted when considering development proposals.

Investments in inland fisheries are usually directed to aquaculture or to production of juvenile fish for stocking of dams or reservoirs even though the effectiveness of such strategies is not well known (De Silva, 2001). Investment in the wild fishery, by managing the environment to increase yield, are likely to be highly cost-effective in comparison (Coates, 2002). Similarly, minimal investments to improve the efficiency of processing, storage and transport of wild fish may show greater returns. But wild capture fisheries cannot compete for investment within the fisheries sector if they are invisible in official statistics.

Small-scale fisheries themselves may be impacting the fishery and may be competing with larger-scale river fisheries, which may potentially be licensed and taxed and managed efficiently. Lack of information and the absence of management increase the risk to investors in government-managed fisheries such as fishing lots and so may reduce their value to government. Moreover, governments cannot fulfil their obligations under international treaties if the impact of small-scale fisheries on aquatic biodiversity and endangered species are not known.

Terminology and scope

A river fishery includes the fish, fishers, traders and consumers as well as the environment in which the fish live. In general, the term also includes other aquatic animals (OAAs), both vertebrates (aquatic reptiles, amphibians, water birds and aquatic mammals) and invertebrates (including molluscs, insects and crustaceans).

In biological terms, production refers to the total biomass of fish or OAAs produced (e.g. as kg/ha/year). Production includes yield, which is the portion removed by fishers, plus the produced biomass, which is not caught. In national economic statistical tables, e.g. those compiled by the FAO, 'yield' is usually called 'production', a statistic of primary interest in gauging the relative size and value of any food industry.

The yield of a river fishery may be estimated in three main ways:

1) By directly assessing the fishery. For example by counting the number of fishers and multiplying by their individual catches, or by counting gear and multiplying by catch rates per item of gear. In large single-species fisheries (particularly marine) this method may provide accurate estimates. In river fisheries very large errors result from the difficulty of surveying all types of fishers and from the variability in space and time of the fishery.

2) By estimating the total area of habitat and then multiplying by a yield per unit area (if known from detailed study of sub-samples of the habitat). Although flooded area is one useful parameter that generally correlates with yield, it is difficult to precisely define or measure, varies greatly from year-to-year and includes a large range of habitats each with differing productivity. Moreover, yield per habitat also varies depending on the number of fishers and fishing intensity.

3) By multiplying the population of consumers by their per capita consumption to provide an estimate of total consumption. From this subtract the volume of imports and add exports, wastage and fish used as feed. The origin of aquaculture-derived fish must also be known, particularly whether such fish represent additional (autotrophic) production, or whether they represent converted wild-caught fish or feed from other sources. The main advantage of using consumption to estimate yield is that population censuses and estimates of individual consumption rates are generally easier to measure or estimate and have lower errors (or absolute ranges) than the measures used to estimate catch or the habitat/yield measures (Bayley and Petrere, 1989). In an environment where all the fish is from wild-capture and all the fish is caught and eaten locally, consumption data provide a very good basis for estimating yield.

Lack of information and the absence of management increase the risk to investors in government-managed fisheries such as fishing lots and so may reduce their value to government.

This paper briefly describes the method used to derive an estimate of consumption for the entire Mekong Basin, which is the unit of most interest for planning purposes for the MRC. A full description is being prepared for publication by the MRC. We then describe the range of accuracy of this figure and the extent to which it can be used to estimate the yield of the basin. Finally we describe the basic scope of household surveys, which are needed to collect information on consumption and associated fishery statistics and the studies needed to improve the methods used.

Estimating consumption

The first synthesis of LMB consumption was published in Sverdrup-Jensen (2002), together with a map, which shows the basin-wide distribution of consumption. The synthesis was based on multiplying per capita consumption by population for each province. The total estimated consumption (of inland fish and OAAs) for the LMB was 2.033 million tonnes per year by a population of 56.3 million people. This provides a mean per capita estimate of 36 kg/person/year. The range of consumption was quoted at 10-89 kg/person/year.

The synthesis that we are preparing is based on more studies and also corrects a number of minor errors in Sverdrup-Jensen (2002). Some errors were caused by using totals from studies which did not include one or two of the main components of consumption: inland fresh fish, inland processed fish, or OAAs, or by using data which included marine products for the inland fish and OAA total. Further errors arise from not including some provinces, which overlap the LMB catchment boundary, or using population figures from different years. Our final consumption figure is likely to be somewhat higher than that of Sverdrup-Jensen (2002), at about 3 million tonnes per year as whole-animal equivalent weight. For our review, population figures for each LMB province were obtained from national censuses for mid-2000, or in one case from 1998 and adjusted to equivalent mid-2000 figures. Consumption estimates were obtained from 17 studies which had been carried out in the lower Mekong Basin, and which in total covered 32 provinces.

The provincial per capita estimates were then used to estimate consumption in adjacent provinces, judged to be similar geographically and socially. Of the studies reviewed, two were by direct measurement of foods eaten, and two were by logbook, i.e. the respondents recorded their own meals. The remaining 13 studies were based on the recall of respondents of foods eaten over a typical year. Consumption was in all cases only one part of the surveys, which usually attempted to describe many aspects of the small-scale fishery and, in some cases, included data on catches and aquaculture.

Quality of the consumption figures

A notable feature of the survey reports that we reviewed was the lack of systematic attention to the quality of the data. It is therefore worth emphasising that environmental surveys should include a Quality Assurance Plan (QAP), the objective of which is to ensure that surveys provide data, which has known quality, consistent with the objectives and budget of the study. A QAP should include five data quality indicators (DQIs): bias, precision, representativeness, completeness and comparability (as defined in Appendix 1, derived from Keith (1988). None of the studies included any explicit information on bias or precision. Most included information on representativeness (e.g. the extent to which samples were stratified to represent the target population). Some provided estimates of the variability of statistics, but these are of little value unless it can be assumed that precision is high and bias is low. Therefore, it is not possible to provide any statistical measures of the likely variance of the consumption estimates. Moreover, in many cases the data are not directly comparable, for example OAAs were excluded from some surveys or only partly reported.

Accuracy of the MRC consumption estimates.

Although it is not possible to provide conventional error estimates based on the data for the reasons mentioned above, the range of values possible for LMB consumption can be estimated assuming that errors in the population figures are insignificant compared with errors in the consumption estimates. A reasonable estimate of the possible range in per capita consumption is 30-90 kg/person/year, based on FAO world figures (see Point 3 below). The resulting range, 1.7 to 5.0 tonnes/year (30 x 56.3 to 90 x 56.3), provides an indication of the maximum possible range for the final mean estimate of about 3 million tonnes/year.

Four approaches used to judge the accuracy of final consumption figures:

1) Comparison with fishing activity and catches. The small-scale surveys all showed fishing activity to be important throughout the basin, including highland areas. Most people fish at some time and many people culture fish. There are few people who never eat fish or OAAs (in contrast to some developed countries). In some studies estimates of consumption from reported catches and small-scale aquaculture production were compared with consumption, and showed reasonable agreement.

2) Yield calculations based on floodplain area x production/ha. This comparison is presented in Sverdrup-Jensen (2002) where a yield of 230 kg/ha of floodplain is multiplied by a floodplain area of the LMB of 9.69 Mha of wetlands to derive a yield of 2.23 million tonnes for the LMB. The similarity of this gross estimate to that from the consumption figures is clearly accidental given the errors involved, but does show that the consumption estimate is of the correct scale.

3) World figures for consumption. The FAO provides estimates of per capita consumption of 'seafood', from official national figures on catches, imports and exports, and sales for animal feed, to derive a 'whole animal' figure. Seafood includes all fish and OAAs from all sources. The FAO figures from developed countries may be considered accurate, but are subject to two significant sources of error, which may balance each other to some extent. First, wastage is not subtracted from the whole-animal figures. Second, consumption from recreational fisheries is underestimated or not included. The FAO figures range from 14.5-90.7 kg/capita/year (Figure 1) and among these the official figures for the LMB countries can be seen to be among the lower figures. This is in comparison to countries where the general populace does not eat much fish or OAAs compared to people in the LMB. Therefore, the FAO ranking of the four LMB country figures seems far too low.

Figure 1 FAO estimates for consumption of all seafood (fish and aquatic animals) for some developed countries, compared with FAO data for LMB countries (in black). Based on FAO website data for 2000

A comparison may also be made to studies of similar environments. For example Bayley and Petrere (1989) summarised results from consumption studies of inland fish from the Amazon Basin. In lowland areas consumption varied from 27-101 kg/person/year. In highland areas where cheap beef is available the lowest consumption was 4 kg/person/year. Hence the LMB average is also at the lower end of the lowland Amazon range. Assuming some similarity between these two floodplain systems, the LMB figure appears to be conservative.

4) Consumption by expatriate LMB country people. Sechena et al. (1999) used quality-assured standardised survey protocols among expatriate Asians in Washington State (USA) and found high annual seafood consumption among people from LMB countries (Table 1).

Seaweed/kelp was subtracted from totals and an average body weight of 60.2 kg (excluding Samoans) was used for converting these figures from g/kg/day.

In terms of whole animal weights, these figures should be approximately doubled. Comparing with the figures in Sverdrup-Jensen (2002, Table 1) for three countries they are between 1.3 and 3.5 times higher than the LMB estimates, suggesting that the LMB figures are conservative. Consumption of course depends upon food availability and income in the new country, but it is also clear that Asians in the USA eat much more seafood than the general population. Moreover, older respondents ate more seafood than younger respondents, perhaps indicating retention of original eating habits. The value of this comparison is that we may have had cause to doubt the LMB figures if the USA figures had been found to be lower than the LMB figures.

Table 1 Consumption among LMB country expatriates in the USA

From Sechena et al. (2002)

Ethnicity

Consumption (kg/person/yr) as edible portions

Total

Shellfish

Fish

Cambodian

31.2

20.2

11.0

Lao (lowland)

42.2

19.7

22.4

Hmong (highland Lao)

12.9

5.5

7.4

Vietnamese

57.4

34.7

22.7

Estimated yield from consumption

The total yield from the LMB can be estimated as:

Yield = consumption - imports + exports + animal feeds + waste + fish fed to aquaculture - aquaculture products - marine products

The values for the terms in this equation cannot be entered with certainty but the following points are worth noting:

Imports of marine products are excluded from the consumption studies. Imports from adjacent basins would be very minor. Therefore, there is no need to subtract imports from the figures.

Exports from the LMB are significant. These include exports of Pangasius catfish from the Delta to other parts of Viet Nam and to other countries and fresh and processed fish from the Great Lake in Cambodia to Thailand and other countries. It is likely that exports exceed imports. Animal feed and waste are unknown, but are certainly at least an additional 10% per year.

Fish feed to aquaculture would be mainly for snakehead (Channa) and catfish (Pangasius) cage culture. For snakehead, the conversion factor of wild fish to aquaculture fish is approximately 4:1 (i.e. 4 kg of wild fish produces 1 kg of snakehead). For catfish the conversion factor is approximately 1, because the fish are fed about 50% rice bran, about 25% marine fish and 25% small freshwater fish. Some other fish used in aquaculture (carp and tilapia) provide additional 'new' autotrophic production but these would be minor at present in the LMB.

Overall it would appear from considerations of the other terms in the equation that yield could be much higher than the figure indicated by consumption alone, perhaps by as much as 50%. This would imply a yield higher than 3 million tonnes for the LMB, well within the range of possible values mentioned in Section 4.

Conclusions regarding the methodology for small-scale fishery surveys including consumption

Consumption is of interest basin-wide, as are other indicators of the scale of the fishery, and it is clear even from the rather scattered data which we have reviewed that the fishery is very large and critical for food security, nutrition and livelihoods. Further work to define the scale of the fishery would be of value, but given the threats to the fishery it is more urgent to establish indicators of trends. Monitoring some small-scale fisheries in each country, including their catches and consumption is feasible and would be cost-effective to provide indications of changes. At the same time long-term intensive studies on relatively few villages would provide the basis for a validation of the large-scale survey methods.

Standardizing methods basin-wide, (and indeed in developing countries generally), would enable valid comparisons of data. In this respect the lessons from studies in the LMB to date are clear, but are perhaps not new. A QAP is essential (Appendix 1), and the following points apply particularly to small-scale fishery surveys.

1) Fewer data of good quality are preferable. Most surveys collected many types of data, many of which were not used, wasting time and resources. For example, catches usually are dominated by five species at any location (50-80% of total weights), so records of their catches and weights suffice. Data on the other species (50+ in some studies) are likely to suffer from poor recall or not be representative.

2) Simple questions, which have categorical answers, are less prone to error than questions that require estimation. It is critical to define participation, either as a catcher/collector, seller or buyer, as an owner of equipment or as a consumer. Such questions can be applied at village, household and individual level. Similarly, frequency data are likely to be more reliable than estimates of quantity.

3) The surveys should be designed with similar formats and questions to allow crosschecking between village, household and individual catches and consumption.

4) The categories of catches or foods should be standardized as far as possible, with particular attention to OAAs for which there is great inconsistency between studies. Many studies did not include some or all OAAs, and in some studies eels were even categorised as non-fish. OAAs can be best categorised taxonomically (separated into vertebrates/invertebrates and then classes) and by habitat: aquatic, amphibious, terrestrial. Similarly, it is important to separate aquaculture from wild fish. Further work is needed to standardise conversion factors used for converting processed fish products to fresh fish equivalent.

5) Visual aids should be used in any interviews. In particular, a comprehensive set of photos is needed for OAAs where there is much confusion over terminology and definition.

6) Quantitative annual data including consumption and catches may be obtained by interviews, but their accuracy has never been validated for the LMB. Intensive studies to compare actual catches and consumption against recall would be invaluable for calibrating results from other broader surveys. If recall is used, accuracy can be improved by using portion size estimation aids (PSEAs), about which there is an extensive literature (e.g. Mitchell et al. 1996; Sechena et al. 1999; Shimizu et al. 1999; and Swindale and Ohri-Vachaspati, 1999).

7) Recall (24-hour) is commonly used in developed countries because the target population can be reached by telephone and because field surveyors are expensive. In the LMB, actual measurement of consumption (and catches) supervised by surveyors is more feasible and cost-effective and would produce some reliable figures against which to compare recall.

8) Consumption on certain days may be high for cultural reasons. Consumption may also be clumped seasonally and over short time periods because high fish or OAA catches are obtained under particular environmental conditions (e.g. relating to flows or lunar phase). Any longitudinal study of consumption would require minimum time blocks of 2 weeks. Ideally some studies would track families daily for a year so that the results could be used to examine variance and define the optimal sampling frequency and duration.

9) A final point relates to publication. Many of the studies reviewed for this report were difficult to obtain as most are 'grey literature' and some were incomplete drafts. Some other studies in LMB languages could not be found. Others no doubt exist, but are poorly known or not referenced in information systems. To enable a basin-wide perspective and to improve the efficiency of further work, we need to improve peer-review and the publication of documents, translation into and from the national languages and to consolidate LMB references in a single location with an accessible referencing system.

Conclusion

Consumption of fish and OAAs in the LMB is conservatively estimated at about three million tonnes per year, or 56 kg/person/yr, with a possible range of 1.7 to 5.0 tonnes/yr as whole-animal equivalents. Various other data suggest that this mean figure is conservative and will be revised upward. Yield from the LMB includes exports, waste and animal feed, and is probably greater than 3 million tonnes/year, a figure that requires refinement based on analysis of existing studies and collection of further data.

Further studies of consumption in each country are of particular value for monitoring trends and should be carried out in a comparable manner using standardised and quality-assured protocols, which still need to be developed and validated.

References

Bayley, P. B. and M. Petrere (1989). Amazon Fisheries: Assessment Methods, Current Status and Management Options. Can. Spec. Publ. Fish. Aquat. Sci.106: 385-398.

Coates, D. (2002). Inland capture fishery statistics of Southeast Asia: Current status and information needs. Asia-Pacific Commission Bangkok, RAP Publication No. 2002/11, 113 pp.

DeSilva S.S. (2001). Reservoir fisheries: broad strategies for enhancing yields. Pages 7-15 in DeSilva S.S. (Ed.) Reservoir and Culture-based Fisheries: Biology and Management. ACIAR Proceedings No. 98. 384 pages. Australian Centre for International Agricultural Research, Canberra, Australia.

Keith L.H. (Ed) (1988). Principles of environmental sampling. American Chemical Society Professional Reference Books, Washington.

Mitchell D.C., Jonnalagadda S.S., Smiciklas-Wright H., Meaker K.B. and P.M. Kris-Etherton. (1996). Accuracy of portion size estimation using the 2-dimensional food portion visual. FASEB Journal, 10(3), abstracted in www.ncc.umn.edu/abstract/abs138.htm.

Sechena R., Nakano C., Liao S., Polissar N., Lorenzana R., Truong S. and R. Fenske (1999). Asian and Pacific Islander Seafood Consumption Study in King County, WA. USEPA 910/R-99-003. 169 pp.

Shimizu H. Ohwaki A. Kurisu, Y. Takatsuka, N. Ido, M. Kawakami N. Nagata C. and Inaba S. (1999). Validity and reproducibility of a quantitative food frequency questionnaire for a cohort study in Japan. Japanese Journal of Clinical Oncology 29(1): 38-44. Available on http\\www.ncc.go.jp/en/jjco.

Sverdrup-Jensen S. (2002). Fisheries in the Lower Mekong Basin: Status and Perspectives. Mekong River Commission Phnom Penh, MRC Technical Paper No. 6, 103 pp.

Swindale A. and P. Ohri-Vachaspati (1999). Measuring Food Consumption: a Technical Guide. FANTA Project, Academy for Educational Development, Washington DC. 14 pp. Available on http\\www.fantaproject.org.

Appendix 1

Important elements of a Quality Assurance Plan for environmental data. The plan should define the acceptable values for the five main data quality indicators (DQIs).

Bias = Accuracy: the difference between measured and accepted, reference or true values.

Precision: a measure of variability of measurements of the same property by the same method.

Representativeness: the degree to which data accurately and precisely represent a characteristic of the population.

Completeness: the amount of valid data obtained compared to the amount that was expected under normal conditions.

Comparability: expresses the confidence with which one data set can be compared to another. Covers: sampling networks, analytes and units, methods, QA, accuracy and precision.


Previous Page Top of Page Next Page