Professional Documents
Culture Documents
& Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Geospatial Modeling of Site Suitability for Pond-Based Tilapia and Clarias Farming in Uganda
Herbert Ssegane , E. W. Tollner & Karen Veverica
a a a b
Department of Biological and Agricultural Engineering, University of Georgia, Athens, Georgia, United States
b
Department of Fisheries and Allied Aquacultures, International Center for Aquaculture and Aquatic Environments, Auburn University, Alabama, United States
To cite this article: Herbert Ssegane , E. W. Tollner & Karen Veverica (2012): Geospatial Modeling of Site Suitability for Pond-Based Tilapia and Clarias Farming in Uganda, Journal of Applied Aquaculture, 24:2, 147-169 To link to this article: http://dx.doi.org/10.1080/10454438.2012.663695
PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.
Journal of Applied Aquaculture, 24:147169, 2012 Copyright Taylor & Francis Group, LLC ISSN: 1045-4438 print/1545-0805 online DOI: 10.1080/10454438.2012.663695
Geospatial Modeling of Site Suitability for Pond-Based Tilapia and Clarias Farming in Uganda
HERBERT SSEGANE1 , E. W. TOLLNER1 , and KAREN VEVERICA2
Downloaded by [Florida State University] at 10:41 27 April 2013
Department of Biological and Agricultural Engineering, University of Georgia, Athens, Georgia, United States 2 Department of Fisheries and Allied Aquacultures, International Center for Aquaculture and Aquatic Environments, Auburn University, Alabama, United States
1
Seven criteria (water requirement, water temperature, soil texture, terrain slope, potential farm gate sales, availability of farm inputs, and access to local and regional markets) were analyzed to determine site suitability for tilapia and clarias farming in Uganda. Crisp and fuzzy approaches of criterion classication were implemented using GIS, and the results were compared. There was a statistically signicant difference between maps generated by crisp and fuzzy approaches. For both the crisp and the fuzzy approaches, over 98% of the land was classied as moderately suitable or suitable. Overall, the crisp method classied 16,322 hectares (0.09%) as very suitable compared to zero hectares (0%) by the fuzzy method. Simultaneously, the crisp method gave 297,344 hectares (1.96%) as unsuitable compared to 168,592 hectares (0.96%) by the fuzzy method. Of the 138 surveyed shponds that were operational, the crisp method classied 71% as suitable and 29% as moderately suitable, while the fuzzy method classied 71.7% as suitable and 28.3% as moderately suitable.
We thank the AquaFish CRSP for funding the project. The AquaFish CRSP is funded in part by United States Agency for International Development (USAID) Cooperative Agreement No. EPP-A-00-06-00012-00 and by United States and host country partners. We also wish to thank Freddy Nachtergaele and the FAO Land and Water Division for the digital soil data, and Dr. Nelly Isyagi, Maurice Ssebisubi, and John Walakira for the shpond survey data. Address correspondence to E. W. Tollner, Biology & Agricultural Engineering Department and Faculty of Engineering, University of Georgia, 0117 Driftmier Eng. Ctr., 597 D. W. Brooks DR., Athens, GA 30602 USA. E-mail: btollner@engr.uga.edu
147
148
H. Ssegane et al.
KEYWORDS GIS, sh farming, fuzzy logic, crisp sets, multicriterion evaluation, and pairwise comparisons
149
information service (GIS) data (e.g., 1 Km digital elevation models [DEM]) in contrast to the current readily available high-resolution (e.g., 90 m DEM) data. Also, the water balance model used to estimate the water requirement used a constant seepage rate of 8.0 cm per day without consideration for the effect of soil texture on seepage rates. Aguilar-Manjarrez and Nath (1998) use crisp or sharp boundaries to dene classes of suitability for different sh farming suitability criteria that do not account for fuzziness of the human cognitive processes. Therefore, the objective of this study was to develop a high-resolution sh farming suitability map of Uganda by comparing maps generated by the crisp and fuzzy methods using geospatial modeling techniques. This study took into account the pond size and drainage area required to provide the pond volume when computing for water requirement criterion. The study also presents a method of constructing and GIS implementation of fuzzy membership functions from criterion crisp limits based on expert knowledge in literature. Also, the above high suitability rating of sh farming in most of Uganda has not translated to the number of successful operations that one might expect, suggesting that analysis with improved tools might be useful.
150
TOP LEVEL [ Objective]
H. Ssegane et al.
INTERMEDIATE LEVEL [ Criteria] BOTTOM LEVEL [ Basic themes ] Rainfall (mm) Water requirement Evapotranspiration (mm)
Seepage (mm) Water temperature Soil texture Slope Farm gate sales Air temperature (C) Percent clay (%) Digital elevation model Population density Time to Kampala (hrs) Time to populated places (hrs) Time to regional markets (hrs) Total number of poultry Farm inputs Distance to feed agents (km)
Access to markets
FIGURE 1 Flow chart of the geospatial modeling process (color gure available online).
of the required drainage area. The higher the required drainage area, the less suitable the site is for sh farming unless groundwater or spring water is an alternative source. There are about 20,000 sh ponds in Uganda, and 90% have surface area of 100 to 500 m2 (Aganyira 2005; Mwanja 2005). Therefore, the choice of pond volume of 500 m3 is based on average pond surface area of 500 m2 (Mwanja 2005) and an assumed average depth of 1 m. Work by Veverica et al. (2009) on catsh cash ow analysis for Uganda shows that seven ponds of 250 m2 each and one nursery pond of 63 m2 can provide a steady net income of $160 per month plus 10 kg of sh per week for family consumption. The use of watershed drainage area in contrast to only net annual water requirement seeks to account for both the net monthly water requirement and the limitation of land in Uganda where the average farm size is 2 hectares (Jagger & Pender 2001). Equation (1) expresses the pond water balance (Figure 2); Equation (2) was used to estimate the required watershed drainage area. Annual water balance model for a pre-determined pond volume. Model : RO DA + RF A ET A Seep A = A d , [1]
151
RO
Seep
FIGURE 2 Water balance model showing inputs and outputs to the pond.
where RO = runoff (mm); DA = watershed drainage area (m2 ); RF = rainfall (mm); A = pond surface area (m2 ); ET = evapotranspiration (mm); seep = seepage (mm); and d = pond depth (mm). From Equation (1), the required drainage area was estimated by Equation (2). A d + ET + Seep RF , [2] RO Seepage. The seepage rate at each site was estimated from the soil textural classes dened in Table 1. DA = Runoff. The curve number method was used to estimate annual runoff. Equation 3 expresses the curve number model where Q is the runoff (mm),
TABLE 1 Daily Seepage Rates Based on Soil Textural Class ID 1 2 3 4 5 6 7 8 9 Soil texture Clay loam Sandy claya Sandy clay loama Sandy loam Loamy sanda Sand Clay Loam Silty clay loam Seepage rate (mm day1 ) 2.50 11.25 12.50 13.25 18.75 25.00 0.25 8.00 6.35M
The seepage values were estimated based on lower limit of seepage rates in Egna and Boyd (1997). a Seepage rates of these textural classes were estimated based on lower limit of percent sand (seepage [% sand/100] 25).
152
H. Ssegane et al.
P is the rainfall (mm), S is potential storage (mm), and CN is the curve number (dimensionless). Q= (P 0.2S )2 ; P 0.2S (P + 0.8S ) 0; P 0.2S 25400 254. CN [3]
where S =
The curve number was determined using the soil hydrologic group, land use, and land cover data. The soil hydrologic groups were estimated using the soil textural classes depicted in Table 2. The land use and land cover map of Uganda was reclassied into six classes (Table 3). Table 3 is a curve number lookup table based on land use, land cover, and the soil hydrologic group. MATLAB functions and scripts were developed to reclassify the soil textural classes into soil hydrologic groups based on data in Table 2. A union operation in arcMap geographic information system (GIS) was implemented to determine the respective land use, land cover, and hydrologic soil group for each map grid cell. Additional MATLAB codes were developed and implemented to generate a curve number map of Uganda based on data in Table 3. The soils textural data was extracted from the Harmonized World Series Database (FAO et al. 2008), while the land cover and land use data came from the Food and Agriculture Organization (FAO) Africover multipurpose land cover database (Africover 2002).
TABLE 2 Hydrologic Soil Groups Based on Soil Textural Class Group A Sand, loamy sand, and sandy loam Group C Sandy clay loam Group B Silt loam and loam Group D Clay loam, silty clay loam, sandy clay, silty clay, and clay
TABLE 3 Curve Numbers Based on Land Use, Land Cover, and the Hydrologic Soil Group Hydrologic soil group Land use and land cover Agriculture Forest Bare areas Rangeland Urban Water A 67 30 77 49 77 100 B 77 58 86 69 85 100 C 83 71 91 79 90 100 D 87 78 94 84 92 100
153
Evapotranspiration. Evapotranspiration was estimated by The BlaneyCriddle model (Equation (4)), which uses mean monthly temperatures. The variable p (Equation (4)) depends on the month and the latitude of a location. A value of p =0.27 was deemed appropriate for all months for Uganda because it lies at the equator. The minimum and maximum temperatures were extracted from the United Nations Environment Programme (UNEP) data portal (UNEP 2010). ETO = p(0.46Tmean + 8.0) Where Tmean =
Downloaded by [Florida State University] at 10:41 27 April 2013
[4]
Tmax + Tmin , 2
where Tmin , Tmax , and Tmean are monthly minimum, maximum, and mean temperatures respectively, while ETo is the potential evapotranspiration. Rainfall. Annual rainfall map of Uganda was generated by interpolating (inverse distance weighing) long-term average rainfall data (19611990) of various rain gauge stations in Uganda (347 stations), Kenya (706 stations), Tanzania (556 stations), Rwanda (40 stations), Sudan (171 stations), and the Democratic Republic of Congo (187 stations). The rainfall gauge data was extracted from the FAO Agroclimatic database (FAO 2010). ArcMap GIS was used to interpolate the above gauge data and to extract the Uganda annual rainfall grid.
WATER TEMPERATURE Boyd and Tucker (1998) give emphasis to water temperature as one of the most critical variables for sh growth. Most sh have a limited optimum temperature range for growth, a range at which feeding and oxygen consumption are at their highest. According to Balarin and Haller (1982), the feeding rate of tilapia decreased below 20 C and stopped at about 16 C. For java tilapia, temperatures of 18 to 34 C were considered optimum (Boyd & Tucker 1998), while 28 to 32 C showed the fastest growth. A study by Azaza, Dhraief, and Kraiem (2008) of juvenile tilapia in geothermal waters of southern Tunisia gave optimum growth and feed utilization at temperatures of 26 C and 30 C. El-Sherif and El-Feky (2009) in Egypt obtained similar results for Nile tilapia ngerling with optimum growth and survival observed at 25 C and 30 C. However, El-Sayed and Kawanna (2008) did not observe a significant difference in weights of Nile tilapia reared at 24 C and 32 C, while a doubling of weight was observed at 28 C compared to weights at 25 C and 30 C. Likongwe et al.s research (1996) of juvenile tilapia showed the impact of salinity on optimum temperature. The optimum temperatures for
154
H. Ssegane et al.
growth were 28 C and 32 C in waters of 0 and 8 gL1 salinity, respectively. Additional reported critical temperatures for Nile tilapia include a lethal temperature of 15 C (El-Sherif & El-Feky 2009). Based on the above values, the water temperature suitability was classied into four classes (Table 4). Water temperatures were estimated using a mean monthly water temperature model (Equation (5)) by Kapetsky (1994) that relates water temperatures to air temperatures of 51 data points at the Henderson Research Station near Harare, Zimbabwe, with a coefcient of determination of 0.82. The recommended mean monthly air temperature range for the model is 14 C to 26 C.
[5]
SOIL
AND
TERRAIN PROPERTIES
Soil texture and slope were the only soil and terrain properties used for this study because of their relevance in contrast to other soil properties. Other soil properties relevant for detailed site-specic analysis include soil pH, organic matter, and sulde content (Yoo & Boyd 1993). The impact of slope is related to construction, drainage, and potential erosion of upland soils. A level to gently undulating terrain is preferred due to the ease of machinery movements during pond excavation and construction, and ease of pond
TABLE 4 Summary of Crisp Values for Each Suitability Group Across the Seven Criteria. The Groups Include Very Suitable (VS), Suitable (S), Moderate Suitability (MS), and not Suitable or Unsuitable (NS) Criterion thresholds Criterion 1 Water requirement (ha required drainage area) 2 Water Temperature ( C) 3 Soil texture (% clay) 4 Slope (%) 5 Farm gate sales (people/km2 ) 6 Access to local and regional markets (travel hours) 7 Farm inputs Total number of poultry Distance to feed agents (km) Very Suitable <5 2832 1530 <2 2001,000 <1 Suitable 520 2428 1015 or 3040 25 30200 13 Moderate Suitability 20100 2024 510 or 4050 515 530 36 Not Suitable >100 <20 or >32 <5 or >50 >15 <5 or >1,000 >6
>100,000 <30
40,000100,000 3050
15,00040,000 50100
<15,000 >100
155
drainage. The International Center for Living Aquatic Resources Management (ICLARM) and Deutsche Geseuschaft fur Technische Zusammenarbeit (GTZ) (1991) consider slopes of 1% to 2% suitable for construction of 1 to 5 hectare ponds, while slopes of up to 5% are suitable for 0.01 to 0.05 hectare ponds. Soil texture is considered the most important soil property because of the need for some clay content (Yoo & Boyd 1993). The clayey soils are required for pond bottom sealing to control seepage and construction of pond embankments, however, higher clay contents are more susceptible to cracking, and thus require additional management. The slope and soil texture suitability (Table 4) were classied by modifying ratings by Yoo and Boyd (1993); Hossain et al. (2007); Hossain et al. (2009); and Hossain and Das (2010).
Downloaded by [Florida State University] at 10:41 27 April 2013
FARM INPUTS Manure is used as a supplement for tilapia and catsh ponds in the rst month of the growout cycle and then is used in reduced amounts by more commercial farmers who purchase feeds. Subsistence farmers will often use manure for the entire grow-out period and hence harvest 1 to 3 tons of sh per hectare. In a summary of demonstration pond results presented by Auburn University (2009), those who use feeds can obtain 5 to 8 tons per hectare with tilapia in static water ponds and 18 to 24 tons per hectare with clarias catsh. Manure availability therefore demonstrates the potential for subsistence sh farming for food security purposes; however, for commercial sh farming (which also concerns food security), access to sh feed is the most important. Therefore, access to farm inputs was estimated using two factors. The rst factor quantied the total number of poultry as an indicator of access to feeds. The total numbers of poultry were calculated at a sub-county level, and the geo-referenced data were based on the Uganda National Household Survey for Agriculture and Livestock data (IGAD-LPI 2010). The second factor was the distance to the location of major feed agents. ArcGIS was used to compute the Euclidean distance to the major feed agents. Computing for the maximum suitability between the two factors generated the overall access to farm inputs.
POTENTIAL
FOR
Farm gate sales are important for rural and small-scale sh farmers because the road network in Uganda is not fully developed and the transportation costs for fresh sh are high. Kapestry (1994), Kapetsky and Nath (1997), and Aguilar-Manjarrez and Nath (1998) use population density as a metric for potential farm gate sales. Pozzi, Robinson, and Nelson (2010) found a strong correlation between expenditure and population density in the Horn
156
H. Ssegane et al.
of Africa (Uganda, Kenya, Sudan, Ethiopia, Eritrea, Somalia, and Djibouti), while Jagger and Pender (2009) showed that the percentage mean household expenditures on sh were not signicantly different between urban (3.33%) and rural (3.29%) communities in Uganda. Therefore, this study used population density as the metric for potential farm gate sales. The same thresholds used by Kapetsky and Nath (1997) and Aguilar-Manjarrez and Nath (1998) were used. Urban areas (where population density exceeds 1,000/km2 ) were classied as low suitability because the cost of land is considered to be too expensive for pond-based sh farming. High land cost has a similar effect to high slopes or sandy soils: it requires the owner to switch to smaller tanks or lined ponds that, by the nature of their greater expense, require a more intensive form of management based upon complete, high-quality sh feeds and water exchange or recirculating systems. ACCESS LOCAL REGIONAL MARKETS
TO
AND
The demand for sh is increasing in the peri-urban and urban markets of Kampala and Jinja as supply from catch sh is decreasing, thus making these markets potential sale points for sh by small-scale sh farmers (Jagger & Pender 2001). Therefore, access to these cities is essential because of developed markets for catch sh due to their proximity to Lake Victoria and the River Nile. In addition to access to these cities, Jagger and Pender (2001) suggest that proximity to main roads is critical for access to potential traders and markets. For this study, accessibility to Kampala and accessibility to populated places with more than 50,000 people were used. The accessibility metric used was the travel time in hours in contrast to distance because it is a more realistic metric for areas where the quality of the transport network is variable (Pozzi et al. 2010). The data used were grid-based GIS layers generated by the IGAD Livestock Policy Initiative. Each grid cell represented the time required to reach the nearest point of interest (Kampala or populated areas) along the least cost route (Pozzi et al. 2010). The data assumes on-road travel by motorized vehicle and off-road travel by pedestrian movement. Using the two layers of travel time to Kampala and populated areas, a single layer was generated using the minimum value of the two layers for each grid cell.
To enable geographic information systems (GIS) operations during the multicriterion evaluation (MCE) process, the criteria thresholds were transformed into crisp and fuzzy set values for comparison. The crisp values used the class limits based on expert knowledge, while the fuzzy values allowed
157
for a degree of membership between classes. The crisp values were based on criterion thresholds in Table 4 such that values of one, two, three, and four represented not suitable or unsuitable (NS), moderate suitability (MS), suitable (S), and very suitable (VS), respectively. The fuzzy values (ranging from one to four) were generated by using a combination of triangular and trapezoidal membership functions. To make certain that the generated fuzzy values ranged from one to four, two rules were followed during the constructing of fuzzy membership functions. The rst rule was to ensure that the intersection of any two membership functions (from different classes, say, suitable and very suitable) is at the midpoint (0.5) of the degree of membership (01). The second rule ensured that the change point in the direction of a membership function of one class was the start point of another membership function for a different class. Figure 3 depicts the constructed fuzzy membership functions for each criterion. For every criterion, two GIS raster layers were generated. One layer was generated using the crisp values, while the second layer was generated using fuzzy membership functions. The crisp layer gave values of one, two, three, or four, while the fuzzy layer gave continuous values linearly varying from one to four.
SAMPLE GENERATION
OF
The water requirement criterion was used to illustrate the approach for generating fuzzy GIS layers using the respective fuzzy membership functions. The relevant membership function is showed in Figure 3. Four GIS raster layers were generated representing the degree of membership of each raster cell to the four classes of not suitable (NS), moderate suitability (MS), suitable (S), and very suitable (VS). The degree of membership () of each raster cell to each class was computed using equation 6. VS = 10 x ; 10 x /10; 0 x 10 0 x 10
S =
MS =
158
H. Ssegane et al.
NS =
x 50 ; 100 1;
[6] [7]
NS
MS
VS
MS
NS
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 5 10 15 20 25 30 35 Percent clay 40 45 50 55
VS
MS
NS
10
20
30
40
50 60 70 80 90 100 110 120 130 140 150 Required drainage area (hectares) MS S VS NS
Degree of membership
NS
0 14.0
16.0
18.0
20.0 22.0 24.0 26.0 28.0 30.0 Water temperature (degrees celsius)
32.0
34.0
FIGURE 3 Sample fuzzy membership functions based on crisp values in table 4 (continued on next page) (color gure available online).
159
VS
NS
Degree of membership
1 V
Degree of membership
FIGURE 3 (Continued)
where VS , VS , VS , VS are the respective degrees of membership of each raster cell corresponding to the classes of VS, S, MS, and NS. The degrees of membership vary from zero to 1, while WRfuzzy is the water requirement fuzzy raster layer generated by weighted aggregation of the degrees of membership (Equation (7)). The values of the fuzzy layer range from one to four.
160
H. Ssegane et al.
n1
[8]
TABLE 5 Pairwise Comparison of the Criteria. The Matrix Values Depict the Relevance of Each Row Factor Relative to the Column Factor on a Scale of 1 to 9
Water requirement Water requirement Water Temperature Soil texture Slope Farm gate sales Access to markets Farm inputs 1 1/2 1/4 1/5 1/6 1/7 1/3 Water Temperature 2 1 1/3 1/4 1/5 1/6 2/3 Soil texture 4 3 1 1/2 1/3 1/4 3/2 Farm gate sales 6 5 3 2 1 2/3 2 Access to markets 7 6 4 3 3/2 1 4 Farm inputs 3 3/2 2/3 1/3 1/2 1/4 1
161
27.6 10,869,061 33.5 37.4 60.2 40.5 18.9 1,495,910 4,667,588 3,378,842 6,348,820 8,594,357
49.5 5,889,347 27.6 6,582,582 17.9 10,595,008 8.4 11.6 7,131,198 3,316,793
162
H. Ssegane et al.
soil texture suitability results support ndings by Rcker (2005), who classied the average soil texture of Uganda as sandy clay loam (about 27% clay) with lower clay content (5.9%17.4%) located in the central region and in the west, while the highest clay (52%57.7%) content can be found in the highlands. The dominant slope suitability class was suitable (37.4%), while 27.6% were classied as very suitable and 8.5% as unsuitable. The dominance of topography with slopes of 2%5% was attributed to the majority of Uganda consisting of a plateau with 84% of the land area between elevations of 900 m and 1,500 m. The potential for farm gate sales was classied as 17.9% very suitable, 60.2% suitable, 19.2% moderate, and 2.7% unsuitable. Modications of metric thresholds used by Aguilar-Manjarrez and Nath (1998) were used in this study, and the results differ due to the change in population dynamics. The increase was attributed to increased average population density from 85 people per square kilometer in 1991 to 123 people per square kilometer in 2002. At least 15.1% of the country was classied as unsuitable for access to local and regional markets because the time of travel to a populated area (50,000 people), to Kampala, or to a regional market was greater than six hours. The dominant suitability class for farm inputs was moderate suitability with 48.8% of the areal coverage. About 20.7% of the area was classied as unsuitable, while 11.6% was classied as very suitable for access to farm inputs.
163
method. For both the crisp and the fuzzy approaches, more than 98% of the area was classied as suitable or moderately suitable; however, the distributions of the suitability values on each map varied. The very suitable locations by both maps are areas near Lake Victoria, while the unsuitable locations are areas in the northeast and southwest of the country. The northeast areas are characterized by dry periods, while the southwest areas experience low temperatures. There are two major areas that depict differences between the two maps: The areas include the northeastern part of the country and the areas near Lake Victoria and Lake Kyoga. The crisp method
Water availability crisp (left image) and fuzzy (right image) suitability.
FIGURE 4 Geo-spatial distribution of Tilapia and Clarias farming suitability generated by Crisp (left) and fuzzy (right) methods (continued on next page) (color gure available online).
164
H. Ssegane et al.
Water temperature crisp (left image) and fuzzy (right image) suitability.
Soil texture crisp (left image) and fuzzy (right image) suitability.
classied more areas as unsuitable in the northeast of the country and more areas surrounding Lake Kyoga and Lake Victoria as very suitable compared to the fuzzy approach. However, the fuzzy method classied more areas around the shores of Lake Victoria and Lake Kyoga as suitable compared to the crisp method. The differences are due to the intrinsic characteristics of the fuzzy method compared to crisp approach. Suitability values based on the fuzzy method are recommended because the fuzzy method models the fuzziness of the human cognitive processes such that suitability classes do not have sharp (crisp) boundaries but degrees of membership between
165
Percent slope crisp (left image) and fuzzy (right image) suitability.
Farm gate sales crisp (left image) and fuzzy (right image) suitability.
classes. Weighted linear aggregation of varying degrees of membership gave different results for the fuzzy-based map compared to the crisp-based map. To examine whether there was a statistical difference between crisp and fuzzy based suitability values, 400 points (same points on each map) were randomly selected and a two-sample paired t-test was implemented. The t-test results showed that there was a statistically signicant difference between suitability values generated by the crisp and fuzzy methods at 5% level of signicance. Further examination of the data points from the two maps showed that the fuzzy approach slightly overestimated suitability at
166
H. Ssegane et al.
Access to markets crisp (left image) and fuzzy (right image) suitability.
Access to farm inputs crisp (left image) and fuzzy (right image) suitability.
FIGURE 4 (Continued)
low values (1 to 2) compared to the crisp approach, while it dampened the suitability at high values (3 to 4). For values between 2 to 3, both methods gave comparable values. Therefore, the fuzzy method dampened occurrences of values at both extremes (unsuitable and very suitable). Comparison of generated suitability values with existing sh ponds showed that of the 138 shponds with operational pond status, the crisp method classied 71% as suitable sites and 29% as moderately suitable, while the fuzzy method classied 71.7% as suitable sites and 28.3% as moderately
167
suitable sites. For the 30 shponds with nonoperational status, both the crisp and fuzzy methods classied 83.3% as being in suitable sites and 16.7% as being in moderately suitable sites. The authors attribute management issues to nonoperational status of the 30 ponds rather than site suitability. One prevalent management issue is levee compaction such that even at sites with high clay content for pond sealing, inadequate levee compaction leads to pond leaking. However, as in many commercial sectors, sh farming success depends on more than just site suitability.
CONCLUSIONS
The GIS implementation of crisp and fuzzy methods to model site suitability for tilapia and clarias farming generated different maps. The difference in suitability values was more dominant in the northeastern part of the country and areas around the shores of Lakes Victoria and Kyoga. For both the crisp and the fuzzy approaches, more than 98% of the land was classied as suitable or moderately suitable. However, the spatial distribution of suitable and moderate suitability by the two classication methods varied. The crisp approach classied more land area as unsuitable (297,344 hectares) compared to 168,592 hectares by the fuzzy approach. Also, the crisp method classied 0.09% (16,322 hectares) of the land area as very suitable compared to 0% (0 hectares) by the fuzzy method. The fuzzy approach estimated slightly higher suitability values at the lower extreme (unsuitable) compared to the crisp approach, while it dampened the suitability values at the upper extreme (very suitable). Suitability values based on the fuzzy method are recommended because the fuzzy method models the fuzziness of the human cognitive processes such that suitability classes do not have sharp (crisp) boundaries but degrees of membership between classes. Unsuitable areas can still be used for aquaculture, but the farmer/investor must be aware that higher costs of production may result from selecting a less-suitable site. For example, where water is abundant but soil is too sandy, or slopes are too steep, small ponds or tanks can be used with accompanying water exchange. The lower suitability values reected with the fuzzy approach suggest that grower success may be enhanced by focusing on management strategies that address the physical issues that reduce suitability in the fuzzy versus the crisp approaches.
Downloaded by [Florida State University] at 10:41 27 April 2013
REFERENCES
Africover. 2002. AfricoverEastern Africa module. Land cover mapping based on satellite remote sensing. Rome: Food and Agriculture Organization of the United Nations (FAO).
168
H. Ssegane et al.
Aganyira, K. 2005. Aquaculture: A tool for sustainable development in Uganda. Trondheim, Norway: Department of Geography, Norwegian University of Science and Technology. Aguilar-Manjarrez, J., and S. S. Nath. 1998. A strategic reassessment of sh farming potential in Africa, CIFA technical paper . Rome: FAO. Auburn University. 2009. Fisheries investment for sustainable harvest . Cooperative agreement 617-A-00-05-00003-00, presented to the U.S. Agency for International Development USAID Uganda. Auburn, AL: Auburn University. http://www.ag.auburn.edu/sh/international/uganda/docs/nal-report.pdf Azaza, M. S., M. N. Dhraief, and M. M. Kraiem. 2008. Effects of water temperature on growth and sex ratio of juvenile Nile tilapia Oreochromis niloticus (Linnaeus) reared in geothermal waters in southern Tunisia. Journal of Thermal Biology 33(2): 98105. Boyd, C. E., and C. S. Tucker. 1998. Pond aquaculture water quality management . Boston: Kluwer Academic. Eastman, J. R., W. G. Jin, P. A. K. Kyem, and J. Toledano. 1995. Raster procedures for multicriteria multiobjective decisions. Photogrammetric Engineering and Remote Sensing 61(5): 539547. Egna, H. S., and C. E. Boyd. 1997. Dynamics of pond aquaculture. New York: CRC press. El-Sayed, A. F. M., and M. Kawanna. 2008. Optimum water temperature boosts the growth performance of Nile tilapia (Oreochromis niloticus) fry reared in a recycling system. Aquaculture Research 39(6): 670672. El-Sherif, M. S., and A. M. I. El-Feky. 2009. Performance of Nile tilapia (Oreochromis niloticus) ngerlings. II. Inuence of different water temperatures. International Journal of Agriculture and Biology 11(3): 301305. Food and Agriculture Organization (FAO). 2010. FAOClim-NET: Agroclimatic database management system. Rome: FAO. FAO, International Institute for Applied Systems Analysis (IIASA), International Soil Reference and Information Centre (ISRIC), Institute of Soil Science Chinese Academy of Sciences (ISS-CAS), and the Joint Research Center (JRC). 2008. Harmonized world soil database (version 1.0). Rome: FAO. Hossain, M. S., S. R. Chowdhury, N. G. Das, and M. M. Rahaman. 2007. Multicriteria evaluation approach to GIS-based land-suitability classication for tilapia farming in Bangladesh. Aquaculture International 15(6): 425443. Hossain, M. S., S. R. Chowdhury, N. G. Das, S. M. Sharifuzzaman, and A. Sultana. 2009. Integration of GIS and multicriteria decision analysis for urban aquaculture development in Bangladesh. Landscape and Urban Planning 90(34): 119133. Hossain, M. S., and N. G. Das. 2010. GIS-based multi-criteria evaluation to land suitability modelling for giant prawn (Macrobrachium rosenbergii) farming in Companigonj Upazila of Noakhali, Bangladesh. Computers and Electronics in Agriculture 70(1): 172186. International Center for Living Aquatic Resources Management (ICLARM) and Deutsche Geseuschaft fur Technische Zusammenarbeit (GTZ). 1991. The context of small-scale integrated agriculture-aquaculture systems in Africa: A case study of Malawi. Manilla, Phillipines/Eschborn, Germany: ICLARM/GTZ.
169
InterGovernmental Authority on Development Livestock Policy Initiative (IGAD LPI). 2010. GIS data for Uganda food and agriculture organization. IGAD Spatial Data Archive. Available online at: http://ergodata.zoo.ox.acouk/ (Accessed on April 06, 2012) Jagger, P., and J. Pender. 2001. Markets, marketing, and production issues for aquaculture in East Africa: The case of Uganda. Naga, The ICLARM Quarterly 24(1&2): 4251. Kapetsky, J. M. 1994. A strategic assessment of warm-water sh farming potential in Africa, CIFA technical paper . Rome: FAO. Kapetsky, J. M., and S. S. Nath. 1997. A strategic assessment of the potential for freshwater sh farming in Latin America, COPESCAL technical paper . Rome: FAO. Klasen, S. 2004. Population growth, (per capita) economic growth, and poverty reduction in Uganda: A brief summary of theory and evidence. Gttingen, Germany: Georg-August Universitt.. http://www.iai.wiwi.uni-goettingen.de/ klasen/uganda.pdf. Likongwe, J. S., T. D. Stecko, J. R. Stauffer, and R. F. Carline. 1996. Combined effects of water temperature and salinity on growth and feed utilization of juvenile Nile tilapia Oreochromis niloticus (Linneaus). Aquaculture 146(12): 3746. Moehl, J. F., M. Halwart, and R. E. Brummett. 2005. Report of the FAO-Worldsh Center Workshop on small-scale aquaculture in sub-Saharan Africa: Revisiting the aquaculture target group paradigm, Limb Cameroon, 23-26 March 2004, CIFA occasional paper . Rome: FAO. Mwanja, W. W. 2005. National aquaculture sector overview. Uganda. Rome: FAO. http://www.fao.org/shery/countrysector/naso_uganda/en National Agricultural Advisory Services (NAADS). 2005. NAADS aquaculture technical manual . Kampala, Uganda: NAADS. Nyombi, K., and S. Bolwig. 2004. A qualitative evaluation of alternative development strategies for Uganda sheries. Washington, DC: International Food Policy Research Institute. Pozzi, F., T. Robinson, and A. Nelson. 2010. Accessibility mapping and rural poverty in the Horn of Africa. IGAD LPI working paper no. 210. PPLPI working paper no. 47 . Rome: FAO. Saaty, T. L. 1977. Scaling method for priorities in hierarchical structures. Journal of Mathematical Psychology 15(3): 234281. United Nations Environmental Programme (UNEP). 2010. The GEO data portal, version 1.4. Nairobi, Kenya: UNEP. Veverica, K. L., N. Isyagi, W. Daniels, and R. Asiimwe. 2009. Planning a small aquaculture business based upon rural markets in Uganda. Paper read at World Aquauclture Society, September 2529, Vera Cruz, Mexico. Yoo, K. H., and C. E. Boyd. 1993. Hydrology and water supply for pond aquaculture. New York: Van Nostrand Reinhold.