The results are quite positive for the development of inland fish farming in Latin America. They show that both commercial and small-scale fish farming is possible over vast areas without serious constraints, either from the lack of suitability of basic factors that are important for development and operation of fish farms, or from constraints of temperature on fish growth. One of the most significant findings of this study is that, from a commercial fish farming viewpoint, combinations of relatively high population urban centres and nearby road infrastructure place large areas that are suitable for fish farming within easy reach of urban markets. Another significant finding is that large areas of Latin America are suitable for the farming of a variety of species. The numbers of crops per year can be maximized by a combination of relatively high feeding rates and harvesting at moderate weights.
From a country-by-country viewpoint, the results are also generally positive. For small-scale farming of Nile tilapia, 19 of the 21 continental countries have 6 to 100% of their national areas that rate at least suitable and from which at least 0.9 crops/year can be obtained. The corresponding results for small-scale farming of carp are that 20 of 21 countries have some areas that are at least suitable and these range from 24 to 99% of national surface areas. From these areas, at least 0.9 crops/y are possible.
Commercial fish farming of Nile tilapia with the 75% feeding rate × low harvest weight combination, is possible in 18 of 21 of the countries. The surface area of these countries that rates at least suitable ranges from 14 to 95%, and at least 1.2 crops/y can be obtained. The corresponding results for carp indicate that fish farming is possible in all 21 countries, with 4–99% of their national surface areas rated at least as suitable. Further, at least 1.2 crops/y can be raised. Of the 21 countries, 19 and 18 can support tambaqui and pacu culture respectively, with the corresponding surface areas ranging from 8 to 99% and 18 to 98%. At least 0.7 and 1 crops/y are possible for tambaqui and pacu, respectively.
Latin America has a larger relative area than Africa with potential for inland fish farming. That is, about 31% of Africa encompasses potential for small-scale farming of Nile tilapia and about 13% encompasses potential for commercial farming of the same species from which 1 to 2 crops/y can be obtained (Kapetsky, 1994). In comparison, 38% to 60% of Latin America scores suitable to very suitable for small-scale farming of Nile tilapia and carp, respectively. In the same areas, 0.9 to 1.8 crops/y can be realized. For the same two species, 19% to 44% of Latin America scores suitable to very suitable for commercial farming, and 1.2 to 2.4 crops/y can be realized on the same areas. In absolute terms, the situation is reversed. Continental Latin America is about 60% of the area of Africa. In these terms, Latin America has about ⅔ as much area as Africa that is appropriate for farming of Nile tilapia at the small-scale level and about 80% as much of Africa that is suited for commercial farming.
The estimates of area with potential for fish farming development (or lack of it) are influenced by many factors. Among them are the use of surrogate factors, the selection of thresholds that describe various classes of suitability, and the use of nominal area estimates.
Surrogate factors had to be used where data on actual factors were not available. For example, in order to estimate the availability of agricultural by-products as feed and fertilizer inputs for fish farming, it was necessary to use estimates of land suitability for agriculture because there are no comprehensive maps that show the actual spatial distribution of agriculture at a resolution that would have been useful in this study. Other examples are the use of population density as a surrogate for potential for farm-gate-sales and population size for market demand at urban centres.
The results were affected by the selection of thresholds for each of the factors and for the commercial and small-scale models, and its equivalent in the fish yield model, namely the specification of feeding rate × harvest weight combinations. In each case, the thresholds were based on values or indications found in the literature, or in the instance of the determination of the relative importance of the factors for the models, expert assistance was brought in. However, in some cases it was necessary to use data that were already classified, such as the ranges of population sizes of urban centres that were used to make estimates of the potential demand for farmed fish.
In order to keep the results manageable and for ease of interpretation, four classes of suitability were employed throughout the study, from very suitable to unsuitable. In order to minimize subjectivity in the scoring process, the selection of thresholds was always based on an examination of frequency plots of grid cells for each factor and model.
In order to avoid a serious misinterpretation of our results, it is important to keep in mind that our estimates of areas that have potential for fish farming development have to be thought of as encompassing or containing such potential. Not all of the area that we have identified as having potential can be given over to fish farming. In fact, in practical terms, only a small part of it is likely to be available in some places. For example, in some of the areas, especially those with relatively high population densities, the land already may be in other uses that provide returns greater than those that can be realized from fish farming.
Another situation makes the area estimates appear to be greater than they are actually likely to be. This is related to the unavailability of digital maps of the protected and reserved areas of several of the countries, including Argentina, Chile, French Guiana, Guyana, Mexico, Paraguay and Uruguay. Thus, in these countries, some of the area that has been identified as having potential for fish farming is undoubtedly reserved for other uses.
Results obtained for tambaqui and pacu are somewhat unexpected for the warmer waters of Latin America because the former species is generally considered to grow more rapidly in waters in the range 25–35°C (e.g., Saint-Paul and Werder, 1980; Saint-Paul, 1989). Although some experiments (e.g., Miyasaka and Castagnolli, 1992) have shown that pacu may grow better than tambaqui in mean water temperatures ranging from 27–30°C, we suspect that the crops/y results obtained for these species (Figures 3.19 and 3.24) may in part be due to the lack of sufficient grow-out data to enable better growth model parameterization for tambaqui (see also Annex 3). Pacu does, however, show better tolerance to colder temperatures than tambaqui (Saint-Paul, 1989) and therefore the higher crop/y output registered from Central Brazil southwards for this species (Figure 3.24) compared to tambaqui (Figure 3.19) is perhaps not surprising.
The present study is both more refined and sophisticated than the strategic assessment of aquaculture potential carried out for Africa by Kapetsky (1994). The most important refinement was that new data allowed for a fourfold increase in resolution over that used in Africa (i.e., to 5 are minutes), thereby making the results much more usable in order to assess fish farming potential at the national level. Sophistication was added by making direct estimates of yield potential as the number of crops per year possible for four species by integrating elements of the models from POND©. In Africa, only one species was used and only indirect estimates of potential were made.
Other refinements include:
A pond water balance model that was that used to estimate water loss from seepage and evaporation on a monthly basis in place of water supply via surface runoff from annual precipitation.
A wide variety of gridded soil data attributes that allowed for better prediction of the feasibility of pond construction and terrain suitability for ponds compared with only two attributes - texture and slope - that were used in Africa.
A road infrastructure × urban population size combination that made it possible to estimate urban market potential separate from potential for farm-gate sales, whereas only different thresholds of population density were used as surrogates for market demand in the Africa study.
Modelling of factors important for the development and operation of commercial and small-scale fish farming that was an advance over the assumption of equal importance among factors in Africa.
Nevertheless, additional improvements can be made, some of which could be carried out with existing data. For example, stochastic weather data could be used to look at best and worst cases for the operation of ponds and for the production of fish. Such analysis could include inter-annual temperature variations as they affect fish growth and production, and precipitation and evaporation as they affect water availability and pond operation costs. Another improvement would be the overlay of the results for commercial farming in order to establish where the best yields for all four species coincide.
Other improvements would require new data or new approaches. Among the most important are:
Land cover at 1 km resolution:
to enable better estimates of land actually available for aquaculture; and
to discern the actual kind and distribution of crop types that, in turn, would allow a better inference of availability of agricultural by-products as inputs.
A complete set of polygon data for protected areas, to include all of the countries of Latin America.
A set of 5-arc-minute precipitation data covering the Caribbean to allow inclusion of the Caribbean island states in the analyses of fish farming potential.
Extending the study to:
include more warm-and temperate-water species. However, in practice, even for those species that are most widely cultured and for which the data are most complete, there are still great gaps in data from operational fish farms. Such data should be included in the growth model. A further extension of this idea would be to improve the growth model to investigate polyculture opportunities;
include more intensive fish culture systems, such as flowing water and re-circulating systems; and
include cold-water species.
However, at a point not too far beyond that to which this study has gone, the utility of a continental-level GIS is lost, and it is much better to proceed to a national-level GIS in order to bring more factors into the analysis, and to take advantage of data at resolutions higher than are appropriate at a continental scale.
The results of the verification exercise suggested that reasonable confidence can be placed on the predictions of the suitability of grid cell “sites” for commercial fish farming derived from the commercial model. However, it has to be kept in mind that the commercial model is based only on five, very general, factors in weighted combination. As is well known, the weights of the factors can change in relation to the species under culture, the intensity of farming adopted and the location of the fish farming enterprise. The commercial and small-scale models address the relative importance of the factors in a very general way at present and there is room for refinement should there be interest in using different species and different culture systems.
At least three possible reasons for the lack of correspondence between predicted Nile tilapia yields for grid cell sites and the presence of operational farms have been explored. Of these, there is evidence to suggest that low crops/y output may be related to predicted water temperatures being less than observed values. It is also likely that the other two factors (i.e., differences in growth potential among tilapia strains and the lower temperature threshold used for Nile tilapia) may have contributed to the low crops/y output. In practical terms, the verification exercise suggests that the yield model is likely to have predicted a lower growth potential for tilapia at altitudes in excess of about 600 m than is actually the case.
Potential for inland fish farming is high in continental Latin America. The most important factor - urban market potential - scores high across much of the continent. Therefore, the low level of development is not one of lack of physical access to markets that are within easy reach of suitable fish farming areas.
Latin America appears to have a larger relative area potential for inland fish farming than Africa.
Overall, the results suggest that, in Latin America, there may be only marginal possibilities for commercial aquaculture operations that target a large harvest size of fish, but use low feeding rates.
In general, the results suggest a high potential for the high feeding rate × low harvest weight type of intensive commercial aquaculture operations for all species. It is also evident that growth models which account for the temperature preferences of these species can be used to gauge the suitability of their culture in different regions. Availability of fingerlings, suitable feeds, local markets and export potential - among other factors - will, of course, dictate the type of species that should be chosen for a given region.
From 38% to 60% of Latin America scores suitable to very suitable for small-scale farming of Nile tilapia and carp, respectively. In the same areas, from 0.9 to 1.7 crops/y for tilapia and from 0.9 to 1.8 crops/y for carp can be realized by harvesting at a modest weight.
For the same two species, from 19% to 44% of Latin America rates from suitable to very suitable for commercial farming, and from 1.2 to 2.4 crops/y for Nile tilapia and from 1.2 to 2.3 crops/y for carp can be realized on the same areas by feeding at 75% satiation and harvesting at a moderate weight. Tambaqui and pacu occupy an intermediate position in terms of the surface area that is suitable or very suitable for commercial farming. From 0.7 to 1.4 crops/y for tambaqui and from 1.0 to 2 crops/y for pacu can be achieved from areas that are suitable or very suitable by feeding at 75% satiation and harvesting at a moderate weight.
Although the verification exercise was carried out only in one country and involved information on farms culturing predominately tilapia, useful insight was gained into the utility of the commercial model and yield model to predict fish farming potential. The results suggest that the commercial model gives quite reasonable predictions for the weighted combination of the five factors that it incorporates. The yield model to a large extent is dependent on good predictions of water temperature. The lack of a complete gridded data set for weather variables may have been a primary reason for predicted temperatures being lower than values reported for actual farm locations. Lower predictions for yield potential may also be due to different strains of tilapia (e.g., with differing degrees of cold tolerance) being cultured. This problem can potentially be circumvented by re-parameterization of the growth model for cold-tolerant varieties of species such as Nile tilapia.
Nevertheless, the most important conclusion here is that the commercial model and yield model appear to work well together to give reasonable indications of fish farming potential. Therefore, the main task of improving the predictions is one of refinement, not reconstruction.
Based on the experience gained in this study a number of future applications can be foreseen. Some are considered below.
In areas where community water bodies - small reservoirs - are common there are good opportunities to increase the fish supply through various kinds of enhancements. Fundamental questions are:
The growth model provides a way to look at temperature effects on fish yield potential in both unenhanced and enhanced situations. For example, it should be possible to match candidate species for stocking to a temperature regime that favours growth.
In terms of water temperature prediction, a small reservoir is a large pond. Therefore no great difficulties are foreseen in estimating small-reservoir water temperatures
From the viewpoint of water availability, the water balance approach can be used, modified for reservoir size.
From the viewpoint of productive potential, apart from water temperature, soil characteristics that affect water fertility will be important. Generally, these can be inferred from gridded data sets of soils attributes.
