...........2..........
Global mapping
   

   
 

<< Previous    Next >>
    2.2. Defining livestock production systems
 
Map 3. Global livestock production systems
Map 3a. Livestock only, rangeland-based (LG) production systems
Map 3b. Mixed rainfed (MR) production systems
Map 3c. Mixed irrigated (MI) production systems
Map 3d. Temperate/tropical highland (LT, MT) production systems
Map 3e. Possible industrialised landless (LMS, LS) production systems
Table 1. Livestock production systems—description and examples (Seré and Steinfeld, 1996)
Table 2. Total land area (km2): region/country by production system
Table 3a.  Total human population 2000: region/country by production system
Table 3b. Total human population 2050: region/country by Livestock production system 

Existing maps of agricultural production systems typically use crops, level of technology and intensity of management to differentiate between different system categories (e.g. FAO, 2001b). There is no global production system map that adequately reflects the importance of livestock in agriculture. However, Seré and Steinfeld (1996) produced a global livestock production classification system which has since been widely used, for example in ILRI’s recent priority-setting exercise (Thornton et al., 2000), and a study on livestock–environment interactions (Steinfeld et al., 1997). Seré and Steinfeld built on FAO’s agro-ecological zone work and included detailed country tables with disaggregated data by area, population, livestock numbers and livestock outputs for each production system category.

     Brief descriptions of the various livestock systems are given in Table 1. Examples of these systems in terms of a set of household characteristics for a ‘typical’ poor household are given in Appendix 2. These include a rating of five types of livelihood assets (human, social, natural, physical and financial), and an indication of any trend associated with these (increasing, decreasing), the main causes of food insecurity and poverty, the degree of crop–livestock intensification, and various other household characteristics.

     To show the livestock production systems on a global map, we tried to replicate Seré and Steinfeld’s classification as closely as possible, but with some translations of their definitions to make use of existing global data sets. Our system is conceptually identical, but has slightly modified descriptors for the four categories: landless systems, livestock only, rangeland-based systems (areas with minimal cropping), mixed rainfed systems (mostly rainfed cropping combined with livestock), and mixed irrigated systems (where a significant proportion of cropping uses irrigation and is interspersed with livestock). All but the landless systems were further dissagregated by agro-ecological potential as defined by the length of growing period (LGP). We define three different agro-ecological zones: temperate/tropical highland, arid/semi-arid, and humid/subhumid. The following describes in more detail the steps taken to produce Map 3a, Map 3b, Map 3c, Map 3d and Map 3e (illustrated in Figure 2).     

     We obtained digital maps showing LGP from FAO (Fischer et al., 2000). These data come as a global grid consisting of 0.5 x 0.5 degree cells and were used to define all climatic zones except the highland temperate category. For the latter, we used two climatic databases from the CIAT and the International Water Management Institute (IWMI) (Jones and Thornton, 1999; IWMI, 2001). Our final map layer showing the three agro-ecological zones is based strictly on the Seré and Steinfeld definition.     

     The second step, to differentiate ‘mixed systems’ (areas that combine cropping and livestock) and ‘livestock only, rangeland-based systems’ (areas with minimal or no cropping), required an evaluation of existing global maps showing the extent of croplands. The only global map showing cropped areas and not potentially cropped areas uses coarse-resolution satellite imagery and is based on a common global land-cover characteristics database. It was initiated by the International Geosphere Biosphere Program (IGBP) and produced by the US Geological Survey (USGS) and the University of Nebraska–Lincoln. The land-cover regions in the database are based on interpretation of advanced very-high-resolution radiometer (AVHRR) satellite imagery consolidated into monthly global composites for the period April 1992 to March 1993 (Loveland et al., 2000; USGS/EDC, 1999).     

     The global land-cover characteristics database identifies between approximately 130 and 260 seasonal land-cover regions (SLCRs) per continent (e.g. 167 for South America and 205 for North America). Each SLCR represents an area with similar land-cover associations and distinctive patterns of biomass production, such as the onset, peak and duration of greenness. As part of a broader global land-cover characterization process, seven different thematic global maps, each with a separate land-cover legend, were produced by aggregating these detailed SLCR map units. These seven maps have corresponding customized legends to make them useful for specific global applications such as environmental modelling, land management and monitoring.
 Figure 2.  Decision tree for mapping the livestock systems classification of Seré and Steinfeld (1996).

 

1. As defined by the global landcover characteristics database
2. ppsk = persons km-2; LGP = length of growing period (days); LMS = landless metropolitan systems, high population density areas; LS = landless systems in high population density areas 
For data sources, see text

     We evaluated three different global maps showing the extent of cropped areas based on this global land-cover characteristics database: 1. One of the seven global maps based on the USGS Land Use/Land Cover System legend (Anderson et al., 1976); 2. A map based to a large extent on the same land-cover characteristics database but with a different legend developed directly from the seasonal land-cover regions (Wood et al., 2000); and 3. A map that combines the global land-cover characteristics database with sub-national data on areas cropped (the 1992 croplands data set, described in Ramankutty and Foley, 1998).
     Our analysis of these maps found that all of them underestimated cultivation level and extent. This analysis relied heavily on expert knowledge at ILRI and CIAT, and higher-resolution resource inventory monitoring (RIM) data for Nigeria (ERGO, 1992) and JICA (Japanese International Co-operation Agency) data for Kenya (Holmgren, 1992), for example. These underestimates resulted from the loss of detail for aggregated global map legends, the coarse resolution of the imagery and the unreliability of the satellite interpretation. All available global legends classified each 1-km x 1-km cell with less than 30% of its area cropped as having no cropland at all. The reliability of the image interpretations varied from region to region, reflecting differences in the structure of land cover and in the availability of reliable ground-verification data. Specific agricultural land-cover types whose interpretation was considered to be problematic included irrigated areas, permanently cropped areas (especially tree crops on forest margins) and extensive pastureland. For a comparison of maps based on the global land-cover characteristics database and higher-resolution land-cover and use data for Central America, see Wood et al. (2000). It is important to keep these limitations in mind, especially when it comes to interpretation of area estimates for each production system. The global maps and tables showing the extent of different production systems are most useful for depicting the approximate location of these production systems, and they indicate the relative size of broad production system categories.
     Our conclusion from the second research step was that we should combine a map based on the USGS Land Use/Land Cover System legend (Anderson et al., 1976) with maps of population density and LGP to capture areas that had a significant proportion but less than 30% of each 1-km x 1-km cell under crops.
     We first produced a map that aggregated the 24 separate USGS legend items into three broad categories: croplands, rangelands and other.
      Croplands is the aggregate of dryland cropland and pasture, irrigated cropland and pasture, mixed dryland/irrigated cropland and pasture, cropland/grassland mosaic and cropland/woodland mosaic. Rangelands consist of areas classified as grassland, shrubland, mixed shrubland/grassland, and savanna. The remaining 15 land-cover classes, including forest, wetlands, water bodies, barren or sparsely vegetated, snow or ice and urban and built-up land, defined the other category.
     With the help of an LGP map, the rangelands category was then further divided into cultivatable (LGP >60 days) and not cultivatable (LGP <60 days) rangelands. A third map, human population density in 2000 (same sources as Map 1), helped to identify additional cropping within this rangelands category. All cultivable rangelands that had a population density of greater than 20 persons km-2 were then added to the cropland category and defined the mixed production system category for our global map. The remaining area under the rangelands category defined the rangeland/livestock-based category on our global map. The threshold density of 20 persons km-2 was based on comparisons of maps depicting different thresholds with higher resolution land-cover data for Latin America, West and East Africa, and on expert opinion at CIAT and ILRI.2
     Combining all this information with the maps of arid/semi-arid, humid/subhumid, and highland/temperate agro-ecological zones allowed us to define the following six systems:

     LGA    Livestock only, rangeland-based arid/semi-arid
     LGH    Livestock only, rangeland-based humid/subhumid
     LGT    Livestock only, rangeland-based temperate/tropical highland
     MRA   Mixed rainfed arid/semi-arid
     MRH   Mixed rainfed humid/subhumid
     MRT    Mixed rainfed temperate/tropical highland

     To identify all mixed systems that had a significant proportion of land under irrigation, we added the only global data set of irrigated areas that is available at 0.5o resolution   (Döll and Siebert, 1999; Siebert and Döll, 2001). While Seré and Steinfeld defined mixed irrigated systems as areas having more than 25% irrigated, we used a threshold of 10% for each grid cell. This modification was necessary because of the large cell size of the original data. Thus, all areas that were classified as mixed systems in the previous research step and had more then 10% of each grid cell under irrigation became the following three categories:

     MIA  Mixed irrigated arid/semi-arid
     MIH  Mixed irrigated humid/subhumid
     MIT   Mixed irrigated temperate/tropical highland

     Finally, to define ‘landless’ or industrialised livestock production areas, we used a combination of population density and the Nighttime Lights of the World database. The population density data are based on the same source as Map 1. The Nighttime Lights of the World database, a 1-km resolution map, is derived from nighttime imagery from the Defense Meteorological Satellite Program (DMSP) Operational Linescan System (OLS) of the United States (NOAA, 1998). We used a threshold of 450 persons km-2 to identify areas that would have sufficient local demand for investments in ‘landless’ or industrialized livestock production areas. This map of high population density areas was then combined with the data on density of city lights to delineate areas with significant urban infrastructure, representing generally more diversified economies and higher incomes. Our global map depicts the following two landless categories (note that these are different from the original landless categories of Seré and Steinfeld [1996] shown in Table 1, because of the difficulty in defining ruminant versus monogastric landless systems):

LS Landless systems in high population density areas
LMS Landless metropolitan systems, high population density areas with  significant     urban infrastructure

     To calculate the statistics for Table 2 , Table 3a, Table 3b the map of the livestock production systems was scaled to match the resolution of the population data, i.e. a generalization from 1-km x 1-km to 4.6-km x 4.6-km cells. To calculate population by production system category, population was assigned to the majority production system type. For example, population in cells that had 51% LGA and 49% LGH were allocated to the LGA category. We used the country boundaries from the Environmental Systems Research Institute (ESRI) Digital Chart of the World CD-ROM (ESRI, 1993) to calculate country statistics.
     A brief critical assessment of the livestock systems map (Map 3) can be found in Appendix 3.

2. 

Human population density has been shown to be strongly correlated to the amount of land cultivated  (Reid  et al., 2000). Data from Zambia, Mali and Burkina Faso (Reid and Ellis, 1995; Reid et al., 1995) show that  with densities below 15 people per km2, an average of 12% of the land is cultivated in these three   countries. Above 39 people per km2, more than 66% of the land is cultivated and above 78 people per  km2, over 95% of the land is cultivated. In semi-arid West Africa, the practice of fallowing disappears when human populations reach 50–85 km-2 (Goddard et al., 1975). Synthesis of additional data from McIntire et al. (1992) shows that fallowing disappears at about 85 people km-2 in 33 sites spread from the semi-arid to humid zones across Africa. Once the practice of fallowing ends, agricultural fields coalesce. We estimate that the threshold of 20 people km-2 will generally be equivalent to 15–25% of the land being cultivated.

<< Previous    Next >>