The LEWS (http://cnrit.tamu.edu/lews) was designed to provide an early warning system for monitoring rangeland forage conditions, livestock nutrition, and health for maintaining food security of pastoralists. The program is an integral part of the existing framework of early warning systems for drought and famine in five countries (Tanzania, Kenya, Uganda, Ethiopia, and Eritrea) in pastoral areas of eastern Africa (figure 22.1). The development and implementation of LEWS include spatial characterization, establishment of monitoring sites, biophysical modeling, model analysis and verification, and automation of information dissemination.
Spatial characterization first required stratification of long-term historical weather patterns in each of the five countries. Subsequently, a map showing zones of similar climate was created from a 5 x 5-km gridded weather surface of the entire continent of Africa developed by Corbett (1995) using the AUSPLINE algorithm of Hutchinson (1991). To conduct the climate clustering, the grid cells of the region were identified on the continental climate surface, and the primary weather attributes were queried. These attributes included maximum and minimum temperatures, annual rainfall, potential evapotranspiration (PET), and accumulated rainfall corresponding with the onset of a growing season. The attributes of the weather grid subset were then subjected to a Ward's minimum variance clustering algorithm (SAS, 1999). The resulting climatic clusters provided a mechanism to help define the boundaries for mapping the point-model output and allowed an objective mechanism to ensure that monitoring sites were located in a manner that optimized the subsequent geostatistical analysis of the model output.
The LEWS monitoring technology toolkit was built to serve both relief agencies and pastoral communities in the Greater Horn of Africa (Eritrea, Ethiopia, Djibouti, Somalia, Sudan, Kenya, Uganda, Tanzania, Rwanda,
and Burundi). Emphasis was placed on assisting institutions to use the system and to create educational programs to help pastoral communities better use the information to cope with drought. Keeping this in mind, the selection of the monitoring sites depended on the inherent infrastructure (roads, markets, towns, water) of the region, location of rural populations, livestock density, location of the dominant ecological sites, traditional animal movement networks, disease incidence, and conflict.
Each monitoring site was so located that the trained site monitor could easily report on-site conditions or provide the analytical team feedback on how the models were performing for that site. On average, one to three sites were selected for the site monitor to visit each month. Because the LEWS system seeks to model the forage production and availability, each site had to be surveyed to help characterize the available grazing area, water resources, animal numbers, livestock movement rules (e.g., where the animals can move during the grazing year), herd density, and estimates of other households who share the same grazing resource. This information was needed to parameterize the forage production model for those points. The households in each site were asked to designate the most probable area of migration should drought set in and force them to move. Locating both the primary and likely migration points allowed the forage conditions in both locations to be captured for the system. Both the household and the migration areas were geo-referenced using inexpensive, handheld global positioning system (GPS) units. The number of points was dictated by the analytical technique used. The minimum number of points per zone within the region was set at 30. As of this writing, there are eight 30,000-50,000-km2 zones activated in East Africa, with 30 or more monitoring points located in each zone (figure 22.1). Five additional zones are at various stages of development.
Once a monitoring site was selected, the sampling protocol was largely driven by the input needs of the biophysical model used for the analysis. Point-based biophysical modeling is a mechanism to capture complex relationships at key localities (e.g., Bouman, 1995). In the case of LEWS, the Phytomass Growth Simulator model (PHYGROW; http://cnrit.tamu.edu/ phygrow/; Rowan, 1995) was selected as the primary analytical engine for the monitoring program and was designed by our research team. PHY-GROW is a hydrologic-based plant growth model that uses multiple plant species, soil, weather, and multiple grazer parameters to simulate daily plant production and water dynamics under grazing pressure for a specified ecological site or a designated plant community. Parameters required by the PHYGROW model include the physical and hydrologic characteristics of soils, physical response and competitive potential of plant species, and forage preferences and forage demand of grazers.
PHYGROW requires that the following major attributes be measured for each site: (1) grass basal cover by species or species group using a modified point method, (2) forb frequency, by species or species group, (3) woody plant effective canopy cover, by species or species group, (4) the name of soil and the depth and texture of each soil layer, slope, degree of rockiness, and (5) the estimates of the temporal density for each kind of livestock (this can be derived from discussions with pastoralists or individual landholders).
Results obtained from the PHYGROW simulations include quantity and relative quality of daily forage production, changes in stocking rates, and a daily water balance (comprising runoff, drainage, transpiration, evaporation, interception, and soil moisture). At each monitoring site, PHYGROW simulates differences in forage production resulting from the varying plant communities, soils, grazers, and weather parameters. Each plant community is composed of its major plant species, with each plant characterized by physiological response to weather, soil conditions, and grazing pressure. Biomass production and water balance are calculated daily for each site using loops to depict natural feedback mechanisms throughout the ecosystem, as a function of the intercepted radiation, precipitation, and temperature. Plant community dynamics (growth, turnover, consumption, decay, and competition) progress with each simulated day, influencing forage production and water balance for the site.
Given the dearth of information on growth characteristics of native species, considerable effort has been made to catalog plant growth attributes (e.g., leaf area index, base/ceiling temperatures for plant growth, turnover rates for leaves, stems, litter, day length sensitivity). The Food and Agriculture Organization ECOCROP (FAO, 1998; http://pppis.fao.org) database was an excellent starting point, and our team has cataloged growth attributes of several hundred species in East Africa. A great need now exists for the ecological scientific community to design algorithms to estimate growth parameters from the known growth habits, taxonomy, and morphology of herbaceous and woody plants to accelerate the parameterization process.
Because the PHYGROW model is used in grazed environments, it was critical to have proper information on temporal changes in the animal population densities and their dietary preferences for plant species. We have collected, by interviewing experienced pastoralists, sufficient information to classify major species into the preference categories (i.e., preferred, desirable, undesirable, toxic, nonconsumed, or only used as an emergency forage by the grazing animals).
Excellent soil parameter estimators are available, which use basic information on texture and soil family class to estimate parameters such as wet bulk density, saturated hydraulic conductivity, and water holding capacity. The most useful estimators for this exercise were the Map Unit Use File (MUUF) soil attribute estimator (Soil Conservation Service, 1997) and the Washington State University hydraulic properties calculator (Saxton et al., 1986; http://www.bsyse.wsu.edu/saxton/soilwater/).
Other biophysical models that would be suitable for capturing rangeland response include the SPUR model (Wight and Skiles, 2000), the SAVANNA landscape and regional ecosystem model (Coughenour, 1992), the Erosion Productivity Impact Calculator (EPIC; Williams et al., 1984; 1997), the USDA Water Erosion Prediction Project model (WEPP; Flanagan and Nearing, 1995), and the GRASP rangeland model (Littleboy and McKeon, 1997). GRASP has been used as the basis of a prototype drought alert information system for Australia (Brook and Carter, 1996). These models have not been applied in an early warning context.
The LEWS toolkit has focused on developing fully automated computing environments that capture geo-referenced weather data, link the weather data with pre-parameterized PHYGROW files (soils, plant community, and stocking rules), and generate the necessary data files and graphics files for each monitoring site. The automation system constructs the graphs and texts and updates a Web site where the information is fully accessible to the outside world. Additional files are generated for distribution to the Arid Lands Information Network (ALIN) (http://www.alin.or.ke/), which places the data on the FTP site of the African Leaning Channel. These data are then uploaded and broadcasted as HTML files (containers) to laptop/desktop computers linked to WorldSpace satellite radios (http://www.worldspace.com/; chapter 21).
Was this article helpful?