Protocol for the integrated decision support system methodology

Impact Assessment Definition

• Methods for acquisition of primary and secondary data to support biophysical and economic analysis
• Databases acquired, developed, and/or modified as national and regional resources

Spatial Characterization of Landscapes and Lifescapes (Livelihoods)

• The Almanac Characterization Tool (ACT): Capturing spatial and tabular data for non-GIS users
• Geo-corrected weather data for systems analysis
• Methods for establishing the spatial framework for integrated analysis of impact

Surveys and Data Gathering

• Rapid Appraisal and Intensive Surveys

Biophysical Analysis of Croplands, Rangelands and Livestock

• Estimating yields as a function of simulation zones in regional analysis
• Linkages between individual elements of the decision support system —biophysical, environmental and watershed models

Farm Level Economic Analysis

• Farm Level Analysis Model (FLAM)

National/Regional Level Economic Analysis

• The Agriculture Sector Model

- Linking food security analysis with economic sector analysis

• Methods for scaling economic and biophysical products from farm to national levels
• Methods for scaling economic and biophysical products from farm to national levels

Environmental Impact Analysis

• Coupling crop mix projections and population driven spatial models to assess water yield and soil erosion in the Sondu River basin

Geographic Synthesis

• Use of analysis of geographic equivalence for extrapolating the application of technology or policy to non-contiguous areas with similar geographic features

Capacity Building and Institutionalization in Collaborating Institutions

• Application of modern communications technology and innovations for distance education
• User-friendly access to the DSS through the use of spreadsheet interfaces or meta models

• Impact of small holder dairy technology — example of USAID research by an IARC
• Economic and Environmental Impacts of Sorghum-Millet and Peanut technology — example of AID research by U.S. universities (CRSP)
• Impact of risk aversion on economic well being and food security of consumers and producers
• The impact of future population projections on food demand and technology requirements to maintain or enhance food availability
• Alternative crop production strategies for averting the effects of and managing drought among smallholders in West Africa
• Economic and environmental consequences of intensification and extensification of agricultural production to meet future food demands
• Economic and environmental consequences crop and livestock interactions in mixed farming systems
• Influence of climate change and alternative production systems on future risk of hunger
• Assessment of factors affecting adoption of technology in the Sikasso Region of Mali
• Global Decision Support System — incorporation of national and subregional data and models into the FAO-WAICENT system
• Application of the Decision System for assessment of the impact of the FAO Special Programme on Food Security in Mali
• Cross cutting assessment of the application of the DSS in East and West Africa — geographic, socio-cultural, and economic comparisons and contrasts

A suite of geo-referenced economic, environmental and biophysical models was developed and linked to holistically assess the impact of changes in technology or policy on food, agriculture and use of natural resources in developing countries. The resulting Decision Support System (DSS) also includes critical foundation data for spatially explicit analyses. DSS was developed and adapted for use at levels of scale from farm to global. Methods were developed and refined in collaboration with ultimate decision-makers and relevant real-world assessments were used as development-demonstration platforms. This approach produces methods tailored to needs of specific users who participated in their development. The applications also produced early useful products from the analysis. The DSS involved a set of cooperative efforts with FAO and national and regional policy decision-makers in developing countries to develop and use DSS at national, regional and global levels in evaluating the consequences of policy options aimed at enhancing food security, reducing poverty, and making prudent and environmentally sound use of natural resources.

The development and application of the DSS employed an overall systems approach wherein various disciplinary and methodological approaches were integrated into a total analytic product. A set of activities was defined to capture the individual approaches in an overall management framework. While this served its management purpose, the actual product of the development was an integration of the parts into an overall analysis system. The integrated product provides new methods for evaluating policy and technology options in Mali and Kenya and, to a limited extent, surrounding countries. As part of the total development, case studies were undertaken involving application of the DSS to priority areas of concern to national decision makers. These studies provided both a framework for developing the methods and immediately relevant and credible products for national decision makers.

This section of the report provides a description of the most important products of the Decision Support System Project. The section is divided into two major parts. The first deals with methodological accomplishments. The second deals with the applications of these methods to a series of specific issues and questions that were identified by senior decision makers and research collaborators in the host countries where we worked.

In this methodology sub-section, we focused on the accomplishments that have been made, and do not intend to provide a full rendition of the methods. We describe these in the context of the

new functional capacity they provide and illustrate how the various tools have been applied. We organized the presentation under the framework of our overall development protocol (see below).

In the applications sub-section of accomplishments, we summarized the results of a series of studies that address key issues identified by senior decision makers in Mali and Kenya. These studies were used as a framework for developing the methods. However, they also provided relevant inputs to the issues of highest priority in the minds of the senior decision makers with whom we were engaged and, by illustrating the utility of the DSS, added to probability of its future use. In this sub-section, we have also provided only a summary to illustrate what has been accomplished.

• Protocol for the integrated decision support system methodology

The integrated and interactive suite of models and related databases embodied in the DSS methodology required that special attention be devoted to the initial design of the protocol to have a fully functional final product. Special communication needs were identified as scientists from different disciplinary backgrounds came together in a common modeling environment. Collaborators with widely different cultural and educational backgrounds had to be brought together into a functional and interactive team. The fundamental precepts of the scientific method, which are broadly understood, formed the basis for establishing the general protocol under which this team developed its capacity to effectively interact. The principles are straightforward and include defining the problem carefully, determining what data and background are required, developing the models and analytic procedures needed, managing data effectively by using a common spatial framework, and putting the pieces together in an integrated and iterative way. While the principles are perhaps obvious, we found that a relatively rigorous protocol was not so easily achieved. It required careful development through an ongoing interactive basis. It provided improved understanding at disciplinary interfaces and sharply increased the efficiency of the overall process of data acquisition and application. Ultimately, the protocol was instrumental in the development of the integrated suite of models. The approach has allowed us to extend the use of GIS methods coupled with a variety of analytical methods and has increased the utility of resulting models in conducting assessment of policy and technology options. We include the details of the protocol as an attachment to this report, because of its length. However, we have used the basic building blocks in the protocol as the organizational framework for the part of this report dealing with methodology as it appears to provide a logical framework and flow. Figure 1 provides a flow chart showing the interactions of the major components of the protocol and our experimental design for development of the DSS.

• Methods for acquisition of primary and secondary data to support biophysical and economic analysis

Data necessary to support an integrated impact analysis is often scattered among a large number of sources, in terms of both publications and non-published formats. In many cases the data required for the biophysical and economic models do not exist or cannot adequately address the spatial and temporal dimension of a more comprehensive impact analysis that spans geography and time (past, present and future). Initially, the boundary of the analysis sets the extent of natural resources, socio-political and economic data that will be required. Secondary data must be acquired by establishing good working relationships with individuals in the institutions relevant to the impact analysis. Much of these data are internal documents prepared in-house or by consultants for specific purposes and not published. In the case of biophysical data such as soils and weather, the information is site specific and in many instances has gaps that create problems for use in the models. There is an array of attribute simulator programs for weather (e.g., geo-corrected generators), plant growth (e.g., base/suppression temperatures), soils (e.g., saturated hydraulic conductivity), and animal attributes (e.g., milk yield). However, in some environments these tools fail or there simply are no tools available, requiring the use of specialists, expert panels, Delphi methods or local indigenous knowledge to approximate critical model parameters. Where funds and time allow, carefully constructed survey instruments applied in a spatially coherent manner were used to generate primary biophysical and economic data where data did not exist. Throughout this study, innovations in rapid field assessments were pursued to make the impact assessment process more time efficient. Many of these methods are provided throughout this document.

• Databases acquired, developed, and/or modified as national and regional resources

One of the most important products being delivered as a result of this project is the set of calibrated, current, and organized information and data needed to use the DSS or to apply to many other analytic methods. The wide array of model runs and associated support data created during the analysis of cropping systems in the INSORMIL CRSP, PEANUT CRSP, small holder dairy technology, Sikasso intensification/extensification, and Rift Valley agricultural intensification technologies has allowed delivery of a comprehensive set of analytical tools and data. These are organized in a manner that allows analysis to be pursued by analysts interested in assessing the impact of agricultural and natural resource technologies and policies in the future. Data were also repackaged in spreadsheets interfaces where the large number of model runs were condensed into mathematical meta-models which were linked to pre-parameterized agricultural sector model (ASM) and farm level analysis model (FLAM) for Mali and Kenya reflecting the analysis for the intensification studies in Sikasso and Rift Valley, respectively.

• The Almanac Characterization Tool (ACT): Capturing spatial and tabular data for non-GIS users

The SANREM CRSP contributed to the continued development and deployment of the Almanac Characterization Tool (ACT) in partnership with USAID Office of Foreign Disaster Assistance, CIMMYT and the Global Livestock CRSP. This collaboration resulted in a comprehensive suite of GIS shapefiles of spatial and tabular data used in the impact analyses for Mali and the surrounding West Africa Sahelian countries as well as Kenya, Tanzania and Uganda and other countries in the Greater Horn of Africa. The software, data and training were provided to our partners in each country where impact analysis was conducted to allow them to explore the wide array of spatial relationships between biophysical and socio-economic data provided in each country’s ACT. Design plans were implemented to determine how best to integrate the spatial nature of ACT with the model analytical capacity via the Common Modeling Environment.

• Geo-corrected weather data for systems analysis

Within the spatial sampling frame, there is a need for using geo-corrected or spatially explicit weather data to insure that the climatology within the simulation zones is properly accounted for in the biophysical modeling, especially where weather station coverage is sparse. Two tools have been created to assist with this in the DSS. The first is a tool to generate spatially explicit weather profiles for the long-term climate profiles needed in generating historical and future yields from the biophysical models. This tool does a geo-correction of the WXGEN weather generator coefficients for user-defined points. The WXGEN model has generator coefficients for the majority of WMO weather stations in Africa. The geo-correction tool extracts a subset of climate parameters from the climate surfaces developed for ACT at the user-defined point, and replaces these same parameters in the weather generator file for the nearest WMO weather station having similar climatology. The weather generator can then be run for a pre-defined set of years and the output used in simulation modeling.

The second tool developed is a near real-time weather data website that enables DSS modelers and other users to download near real-time weather data for the continent of Africa (http://cnrit.tamu.edu/rsg/rainfall/rainfall.cgi). This product can be used to fill a variety of needs. This climate can be used as driving variables in biophysical models to provide spatially explicit estimates and analyses. The data also provide a higher resolution estimate of climate in areas where conventional weather stations are sparse.

A spatially coherent sampling frame is needed in the DSS to allow for a representative sampling of villages and/or farms across a region, and to allow scaling to the appropriate levels. Sampling frames were developed that emphasized biophysical traits of the region such as soils, climate, and human population density rather than arbitrary political boundaries. This insured that the range of spatial differences that might influence farm yields and natural resource management across these regions was captured for both the economic and environmental components of the impact assessment. The primary steps in developing the spatial sampling frames are as follows:

1. Unique zones of similar climate and soils are defined for biophysical model simulations (simulation zones)
2. Simulation zones are weighted according to human population density with those having the highest population density receiving the greater weights.
3. Representative farms or villages are selected based on the proportional weights of population within the simulation zones.
4. Surveys and data gathering occur at each of the selected villages and/or farms.
5. Climate, soils, crop and livestock management data are catalogued and entered into the biophysical models for simulations

Simulation zones are developed using interpolated climate surface layer for the region of interest and the available soil data layer. A cluster analysis is performed on the climate grid to determine areas of similar climate (climatic clusters). A spatial cross-tabulation of the climatic cluster layer and the soil layer allows the identification of spatially explicit zones of similar soils and climatology (simulation zones).

Simulation zones can then be spatially cross-tabulated with population densities for the region to assist in selection of representative farms. This insures that selection is weighted more to areas having greater population density. Other spatial layers such as distance to primary roads, infrastructure, and land use also can be added into the cross-tabulation to refine farm selection. For example, in the impact assessment conducted in the Sikasso region of Mali, 56 simulation zones were identified. This spatial layer was then merged with the 1990 population density grid for the region (Figure 2). This allowed the delineation of 10 unique (or "best") simulation zones that represented 76% of the total land area and 80% of the population in the Sikasso region (Figure 2). These 10 simulations zones then served as the primary areas for selection of villages and farms to collect data for economic and biophysical modeling. Other spatial information was then introduced into the GIS to aid in village/farm selection to insure that villages selected were representative of rural agriculture villages. Characteristics such as infrastructure (schools, hospitals, markets) distance to roads, and distance to primary markets were included in the analysis. Villages having characteristics of a rural agriculture village were clustered, and two villages for each of the 10 simulation zones were randomly selected from the subset. These villages served as the areas where rapid appraisal surveys were conducted.

This method provides an integrated and spatially referenced approach for a spatial framework to analyze new technology introductions and other agricultural policies. It spans multiple dimensions of natural resource management including watersheds, soils, and rangelands. It also has a varied scope that ranges from firm level analyses (farms) to higher levels such as the watershed, national, and global levels.

As part of the impact assessments conducted in both Mali and Kenya, a series of rapid appraisals and farm surveys were conducted to gather data for the DSS. Data obtained from these surveys provided baseline information for use in the subsequent case studies and provided the basis for further development and integration of the suite of models in the DSS. Rapid appraisal surveys were used to gather data from a large enough sample size to determine farm characteristics and modality (or representativeness). For example, in both the Sikasso region of Mali and the central Rift Valley in Kenya, approximately 100 farms were surveyed.

After rapid appraisal surveys are completed, a cluster analysis is conducted on a set of summarized variables to define farm types and locations of representative farms. For example, in the Sikasso region of Mali, cropland hectares, cotton hectares, cereal grain hectares, and number of bullocks were used in the cluster analysis. This resulted in the delineation of four farm types (clusters) (Figure 3). Within each farm type, farms having variables closest to the cluster mean were selected for intensive surveys.

Intensive surveys are conducted at representative farms to collect information on crop management, livestock management and economic information. This information is then used in parameterization of biophysical and farm-level economic models, as well as for budget information to supplement the agriculture sector analysis.

Modern analytic methods and models often have substantial data requirements to run. In many instances, data for biophysical models from secondary sources is either incomplete or missing. They may be either unavailable or prohibitively expensive to acquire. Methods were developed to produce credible estimates for these missing data. In the case of incomplete or missing weather data, the WXGEN weather generator program was used. WXGEN can be used to fill incomplete weather data sets and can also be used to generate weather based on statistics of a weather station. For the majority of impact assessment studies, the weather generator is used for the latter.

Soil is another example of data that many times has missing parameters or incomplete datasets. Soil parameter estimators are used to fill in these gaps. Since soil texture is usually collected on a regular basis, the soil hydraulic properties calculator (http://www.bsyse.wsu.edu/saxton/soilwater/) is particularly useful. The EPIC model also has built in algorithms for estimating several soil properties.

The simulation zones are used as the basic unit for determining yields within the spatial sampling frame. Information on soil properties within each zone are catalogued and input into the biophysical models. Information on various crop mixes, rangeland plant communities, and future crop and rangeland technology options are gathered from rapid appraisal surveys, intensive farm surveys, secondary data sources, and expert opinion. This information is entered into the biophysical models to reflect specific scenarios to be modeled. Crop and rangeland management information (e.g., planting dates, fertilization methods, livestock movement) are gathered during surveys and interviews or from secondary data sources so that biophysical models reflect the farmer’s decision-making process.

Once specific biophysical modeling scenarios are determined, the biophysical models are run for a series of weather conditions. The weather can be historical data or generated data. In the case of the studies conducted in the Sikasso region of Mali and the central Rift Valley of Kenya, generated weather data was used in order to capture more robust states of nature. Once model runs are completed, yield information is summarized into formats for use in both farm level and sector level economic analysis. Yield information transferred to farm level economic models (FLAM, FLIPSIM) is generally simulation zone specific. However, sector level economic models (ASM) generally require that yield information be area weighted to a political or administrative boundary (Figure 4).

Area weighting methods provide a framework for scaling yield, runoff, erosions estimates and other biophysical model output to watershed, regional or sector level analyses. Because the simulation zone is the basic aggregation unit, differences in the biophysical model output at each of the varying soil types, climates, and land uses can be properly accounted for in the aggregate yields for the region of interest

Another technique that has been used to assess yields over large landscapes is to use cokriging to create surface maps. Cokriging is a geostatistical interpolation technique that takes advantage of the cross-correlation between a spatially sparse and a spatially rich dataset. In the DSS, biophysical model simulations can be done for a minimal number of points, and then, if a correlation exists between the output and NDVI satellite imagery (a spatially rich greenness index), a surface map of forage production can be created for a region (Figure 5). These surface maps can then be used in any type of analysis that requires spatially explicit vegetation production for a region.

• Linkages between individual elements of the decision support system —biophysical, environmental and watershed models

The methodology utilized in the various studies focused on establishing a spatial framework for the problem area of analysis and targeting the appropriate tools that could be supported with the available data. We chose to use a "loose coupling" paradigm to pursue impact assessment for the current methodology. However, we recognized the need for tighter coupling via the use of embedded meta-models and the creation of the common modeling environment in the later stages of this project to allow greater transferability of analytical tools to our partners in East and West Africa.

The use of spatial characterization of the target impact domain establishes bounds of analysis, relationships of biophysical data among the crop, rangeland and livestock models and linkages of biophysical responses to economic data that has both a temporal and spatial context. The primary tool used was ARCVIEW to set the spatial analytical frame while four biophysical models were employed to capture the interaction of natural resources, management practices and climate in terms of yields and environmental effects. These biophysical models included the SWAN crop model, PHYGROW grazingland model, SWAT basin scale hydrology model and the NUTBAL PRO livestock production model. Soils and weather formed the foundation datasets for SWAN, PHYGROW and SWAT, requiring careful analysis of data quality and resolution desired relative to the impact issues being pursued. As expected, weather data is generally too sparse and soils data is too coarse in spatial resolution for any fine-grain analysis of biophysical phenomena.

A major product that emerged from these analyses was a coherent set of biophysical data on weather, soils, plant growth attributes, cropping practices, forage quality profiles and animal production characteristics that have been reviewed, and missing data elements subjected to techniques to "fill" datasets capable of being used by these daily time-step models. Organization of biophysical data for future modeling efforts will save our partner countries many years of down time and pointed to data deficiencies that need to be addressed in the future.

• Farm Level Analysis Model (FLAM)

The farm level analysis model (FLAM) went through several stages of evolution during the five-year research period. In addition to modeling the impact of technology and policy options at the farm level, an overarching objective of the DSS development was to provide the ability to scale up or scale down the results of these options and to link their economic and biophysical-environmental consequences. To accommodate the spatial aspect and the linkages to ARCVIEW, a new version of FLAM was developed to accept input directly from ARCVIEW. This was achieved through the use of spatially explicit analysis which, in turn, drove several innovations to the economic models. This creates consistency between FLAM and the spatial framework, as it allows FLAM to tap into the biophysical and spatial databases created during analysis and housed in GIS. To estimate environmental impacts of alternative farming techniques, the structure of FLAM was generalized to include the time varying aspects of changes in soil physical and chemical properties. This required integration with the biophysical models PHYGROW and SWAN, as well as integration with the environmental model SWAT. In both cases, the integration was achieved through meta-functions, which create analytical representations of the complex and computationally exhaustive biophysical and environmental simulations. Since scaling has been a central theme in the DSS methodology, a version of FLAM was developed that allows it be directly embedded into the ASM. This provides the utmost consistency between farm level decision making and national level market outcomes since ASM prices are directly incorporated in the embedded FLAM model.

The Agriculture Sector Model (ASM) is an economic model that simulates the behavior of consumers, producers, and traders involved in, respectively, consumption, production, and trading of agricultural commodities. The model has regional and national level spatial characterization of the agriculture sector. At regional level, it considers production and consumption of the major agricultural crop and livestock commodities, while at national level it considers imports and exports of these commodities. The model is used to assess the impact of various policy or technology alternatives on prices, production, consumption, inter-regional transportation, and trade of agricultural products. The impacts are also assessed in terms of relative economic benefits to producers and consumers. The impact results are computed at the regional level with a reflection on the overall national level impact. Inter alia, the model has been used for the climate change impact, the impact of carbon sequestration program, and food security analysis. The utility of the model was substantially extended in these studies by the addition of stochastic elements that can be related to risk assessment and aversion. Recently a new feature was added to ASM to further augment its capabilities for food security analysis focusing the African countries. The model was extended to include the FAO methodology for quantifying food security in terms of an index called ‘Prevalence of Hunger’ or ‘Risk of Hunger.’

A prime objective of agricultural polices in the developing countries is to achieve food security, yet the concept has different connotations in the community of development researchers. Hunger, malnutrition, caloric deficit, insufficient access to food, low body mass index (BMI), stunting, and wasting are some of the terms used in describing the state of food insecurity. Use of these terms in the food policy literature is replete, however, without much substantive policy relevance. An extensive review was made on approaches for quantitatively measuring food security with a focus on their relevance to policy. The extensive review showed that the FAO approach of computing risk of hunger not only quantifies food insecurity, but also had strong policy relevance.

Accordingly, an extension of the sector level economic models was done to include the FAO methodology in the Mali Agriculture Sector Model (MASM). The model was used in a study on the impact of population, resource degradation, and new technologies on food security in Mali by year 2015. The results show that even with the wider adoption of existing improved technologies in Mali, the risk of hunger in Mali may not decrease by year 2015 from its current level of 32 percent. Hence, achieving the World Food Summit objective of reducing the world hunger by half by year 2015 would require substantially more efforts. These efforts may include a more concentrated effort on developing new cultivars, improved natural resource management, and an extension of cropland area.

The incorporation of the FAO methodology in the DSS has added a new dimension to DSS’s capability for food security analysis. Also, in the process of integrating the FAO methodology with the DSS, the utility of the methodology was also enhanced by transforming it from a merely accounting procedure to the one tied to a behavioral and decision making framework.

• Methods for scaling economic and biophysical products from farm to national levels

The key to scaling outputs of models, either economic or biophysical, is to establish a spatial sampling frame that provides a mechanism to capture the diversity of production environments across a region and then scale those responses proportionally within recognized geographical areas. Several methods were devised that allowed area weighting of biophysical responses and have been described in the section on spatial analysis. Scaling up or scaling down biophysical and environmental variables has been done in a geographic framework. Levels range from farm to simulation zone to watersheds to national levels. At the current level of our research, scaling these parts of the DSS has been more achievable than scaling economic dimensions.

It has been important, as noted elsewhere, to provide linkages between farm and sector level economic models since the outputs of each of these models is influenced by inputs from the other. However, the variability in economic preferences among farmers and consumers is quite immense and many of the complexities that dictate market (and non-market) outcomes are not completely contained within current DSS modeling parameters. As this research continues, these important factors will be captured through entropy techniques allowing aggregation to varying scales of interest to better reflect economic consequences of technology and policy.

The Sondu River, a major drainage area of Lake Victoria, was chosen to explore the environmental impact of land use change and associated evolution of small holder dairy technology from 1978 to 1997. The basin possesses a diversity of environmental types constituting four of the seven agro-ecological zones identified for dairy production in Kenya. These characteristics were identified when the watershed boundaries theme layer in ACT was overlaid with an agro-ecological based map of the dairy production zones (Figure 6). Land use change projected by the Kenya ASM economic analysis was coupled with population density projections and expert opinion on maximum adoption rates to project changes in crop mixes within the watershed. The proportion of different land uses were then applied to each grid cell within the basin to provide a calculation of land area per grid cell that is occupied by the various land uses for each Smallholder Dairy technology scenario. The grid cell then acted as sub-basins in the SWAT basin hydrology model with water and erosion loads routed through the watershed based on the designated land use and area within the grid.

The results of this characterization indicate that increasing population pressure most likely influences land use changes greater than the adoption of smallholder dairy technology. At the outlet of the Sondu River basin, there was a simulated increased water flow of 23% from the traditional smallholder dairy to the current land use composition reflecting adoption patterns to date. However, given the increased use of Napiergrass and crop shifts, there is only an anticipated 2% increase when full adoption is attained. The sediment load was projected to around 7 million tonnes at the end of the 20-year period for all three technology scenarios. At full adoption, sediment load only increases 0.93% unless trespass farming on forest reserves occurs in the basin.

The ability to extrapolate results of the DSS is important when considering the application of technology or policy to non-contiguous areas. This provided a basis for evaluating the impact of technology developed at one location to other geographically similar locations in that country or adjacent countries. One way of doing this is to use geographic equivalence. Using the spatial characteristics of the sampling frame, ranges of characteristics can defined. These characteristics can then be entered into a regional GIS system and areas having similar geographic properties can be identified. The ability to do this is limited only by the amount, quality, scale and resolution of the spatial data available for the region

The technique of geographic equivalence was developed and tested in the impact assessment of small holder dairy technology in Kenya. The critical first step for the regionalization analysis was to use the description of the Kenya smallholder dairy environments and extrapolate those conditions over Uganda and Tanzania. Using the description of the resulting ‘types’ of smallholder dairies from the spatially defined areas in Kenya, areas having similar biophysical traits (soils, temperature, precipitation) in Uganda and Tanzania were identified (Figure 7). The horticultural zone (HORT) was by far the most dominant environmental zone to be projected into both Tanzania and Uganda.

Once the extrapolations were done, other environmental constraints that could limit dairy operations were evaluated. This assisted in determining the locations of sites suitable for small holder dairy in Uganda and Tanzania and areas where small holder diary technology would have the greatest impact. Geographic layers of livestock disease pressure and human population density were cross-tabulated with the extrapolated geographic equivalence layers allowing stratification of these zones. For example, areas in Uganda were identified as suitable for the introduction of small holder dairy technology that is similar to the Horticultural (HORT) small holder dairy zone in Kenya (Figure 8).

The use of the geographic equivalence allows first order approximations to be made of where technology developed and evaluated in Kenya might be applied to Uganda and Tanzania. The utility can be improved by including in the assessment other relevant variables, such as disease and population that were not modeled in the first level extrapolation from Kenya.

• Application of modern communications technology and innovations for distance education

In year 5 of the DSS project, the ability to employ modern distance education methods for use in capacity building in Mali and Kenya are being actively explored. Texas A&M is a recognized leader in the application of technology for real-time internet linked two-way video and related capacity in managing other teaching and data materials. Feasibility of its use in the target countries is paced by regulations, licensing, and availability of what is expected to be relatively low cost equipment. If this capability is not achieved under SANREM II, it may be developed in follow-on capacity building projects with CILSS and LEWS.

• User-friendly access to the DSS through the use of spreadsheet interfaces or meta models

DSS has aimed at building models that are not only simple to use but also have a greater relevance to real world issues. However, the sheer nature of the spatial and economic diversity that underlies the types of processes that the DSS was tasked to replicate resulted in the construction of some fairly complex models. For instance, the Agriculture Sector Model (ASM) was tasked to simulate how a host of agricultural commodities are produced, consumed, and traded in various regions of a country. The model’s use requires such programming skills that are not often readily available in the developing countries. To overcome this limitation, the DSS was focused on developing a user friendly approach for easing access to DSS tools. As a first attempt, a spreadsheet based interface was developed for ASM.

This ASM spreadsheet interface does not require specialized programming skills beyond navigating through an EXCEL spreadsheet, resulting in minimal training costs for host country institutions. The user can provide input through spreadsheet on items such as crop yields, available cropland, and adoption rates for improved cultivars. The output items include regional and national level production and prices, and producer/consumer surplus. The interface works in four simple steps. The user enters input in spreadsheet, the interface sends this information to ASM code, runs the model, and brings results back into spreadsheet without having the user to ever see or change GAMS code. The first application of ASM interface is made for Mali ASM. Similar spreadsheet approaches will be developed for other models in the DSS during year 5.

Small holder dairy technology has progressed from a primary native zebu-based production system to greater concentrations of exotic dairy breeds and their crosses with zebu in peri-urban environments. Ex poste analysis estimated the magnitude of the positive impact that improved dairy technology has had on the Kenyan economy and social welfare. Adoption studies for the future predict further positive impacts, in the context of the predicted growth in population and related demand for milk. Population expansion will require future improvement in dairy production to meet this growing demand. With the adoption of the improved dairy technologies, total social welfare to date increased an additional 705 million Ksh annually. These results indicate that the improved dairy technologies have substantially benefited producers and their families through expanded supplies and lower prices for milk and other commodities and through reduced milk imports. When the dairy technology improvements are fully adopted under demand growth rates associated only with rising population for the next 15 years, as contrasted to current adoption rates and demand levels, both consumers and producers benefit. Regional consumers in towns and cities nationally are projected to gain 181.54 billion Ksh (113.2%) annually, while home consumption expenditure by farmers and their families is increased 58.3 billion Ksh (107.0%) annually. Producers return to land and labor would increase 11.8 billion Ksh each year. The increase in home consumption expenditure for food substantially outweighs the increase in producer’s return to land and labor. Foreign surplus increases only slightly, up 274 million Ksh annually, or about 0.3%. Total social welfare in Kenya is increased 135.31 billion Ksh (67.0%) annually under the demand growth scenario.

These results indicate that even under demand growth conditions, domestic consumers in towns and cities are likely to be the major beneficiaries of the smallholder dairy research and technology transfer relative to rural producers and those families that adopt the new technologies.

Results from the deterministic and stochastic simulations of the representative farms indicate that the horticulture, peri-urban and coastal agro-ecological zones generally benefited most from the adoption of the improved dairy technologies. Net cash farm income was positive and increased as the dairy technologies are adopted on these farms under deterministic conditions. When price and yield variability are taken into account, only farms in the coast region and horticultural agro-ecological zone experienced slight increases in net present value, net cash farm income, and real net worth from adoption of the improved dairy technologies. Other farms exhibit a mixed pattern of income and net worth mean values as a result of the dairy technologies.

The environmental impact of smallholder dairy technologies has been relatively neutral when averaged across administrative districts. However, the evolution of these technologies from traditional zebu dairying on common grazing lands to the current mix of farms and technologies has resulted in an increased streamflow of approximately 23% while sediment loading has risen by 5% using the Sondu River basin as a point of reference in the Highlands of Kenya. Most of these effects have been realized over the past 20 years. Given the saturation of land use in the Sondu basin, only minor increases in erosion and increased runoff are expected unless government policy does not curtail illegal farming of the national forest reserves in the watershed.

In this study, the short and long term impacts of a sorghum production system consisting of improved germplasm, higher fertilization rates, and water conservation through ridge tilling were evaluated. The product was developed, in part, by the USAID INTSORMIL CRSP.

This study underscores the critical link between the adoption of new technology and market demand conditions that is nearly universally associated with the basic food commodities. Under current demand conditions, for instance, nearly all of the benefits from new technology introduction would accrue to consumers. In this study, the economic impacts from a 40 percent adoption of the new sorghum production system was estimated to lead to an increase in consumer surplus of about 2.58 percent, whereas producers would see their surplus fall by 30 percent. This transfer of economic surplus from producers to consumers stems in large part from the low price elasticity of demand for the basic foods. In aggregate, after accounting for consumers, producers, and foreign trade, the resultant effect of new technology is a modest increase of 0.76 percent in societal benefits. Although this figure is low, the analysis was only for a limited number of commodities intended to showcase the DSS methodology, and moreover did not account for the introduction of new technology outside of the Sikasso region.

However, estimates of impact when new sorghum and pearl millet technologies are adopted under future demand conditions (year 2015) suggest that both domestic consumers and producers will be beneficiaries. Urban consumers nationally gain 33.02 billion fcfa (4.25%) annually while home consumption expenditures by farmers and their families is increased by 55.76 billion fcfa (43.04%) annually. Producers’ returns to land and labor are increased 112.37 billion fcfa (82.68%) resulting in a net welfare gain of 56.61 billion fcfa annually when combined with home consumption expenditures. In contrast foreign surpluses are eliminated as cotton exports are reduced by some 27.2 thousand tons. Total social welfare in Mali is increased 94.63 billion fcfa (12.04%) annually under the demand growth scenario. These results emphasize the importance of assumptions about demand growth when economic impacts of new technologies are assessed in developing economies where agriculture is a dominant source of gross domestic product and employment.

The environmental impacts of new technologies were found to stem from two factors. One is the use of improved tillage practices that utilize animal traction to build soil ridges. The flow of water along field surfaces is greatly reduced using ridging, and leads to much lower soil erosion rates. The second positive impact of the new technologies is on improved seedling and young plant growth that increases vegetation early on in the growing season, and provides increased water absorption and increased protection from water runoff. As with ridging, soil erosion rates are reduced. The biophysical models estimated that soil erosion rates would fall in nearly all of the regions in Mali. Specifically, the change in soil erosion rates by the year 2015 through the introduction of new technology would fall by 13 percent in Koulikoro, 43 percent in Kayes, 11 percent in Sikasso, 41 percent in Mopti. In the three remaining regions no changes in soil erosion were estimated (Segou, Tombouctou, and Gao).

Risk is a key feature of agricultural enterprises; farmers' decisions are influenced by changes in weather, market conditions, and biotic factors that lie outside their control (pests and weeds). Despite the obvious role of risk in agriculture, it has often been ignored in agricultural policy synthesis. The DSS employs a systematic approach that utilizes weather patterns to improve the information provided to decision makers by incorporating the behavior of farmers towards the spectrum of weather outcomes that they potentially encounter. One of the more important contributions to the ASM approach was the addition of a stochastic component that allowed decision makers to have better predictive tools.

The impacts of farmers’ risk aversion on food prices, food production, and welfare measures were estimated by the ASM. When relative risk aversion was increased from 0.0005 (near risk neutral) to 4 (highly risk averse), the results revealed that farmers would shift to a less aggressive crop portfolio and a subsequent decline in food production. As a result, ASM estimated higher food prices, lower farm profits, and an overall decline in societal welfare. In particular, Malian farmers' welfare decreased by 40.53 percent, consumer surplus increased by 5.92 percent, and societal welfare fell by 2.8 percent.

This establishes that while the socio-economic realities and high levels of uncertainty lead smallholder farmers to being rather risk averse in their decision making and have only mild consequences to farmers on an individual basis, the aggregate effect can be quite significant on both farmers as a group and society at large. For larger and more sweeping technology introductions, where the stakes would be larger, the effects of high levels of risk aversion will likely be much larger than that found in this study. As risk aversion generally decreases with wealth and access to improved physical and financial infrastructure, policies

to promote these measures would have the co-benefits of reduced risk aversion.  

• The impact of future population projections on food demand and technology requirements to maintain or enhance food availability

Given the concerns that exist over sub-Saharan’s ability to feed itself over the coming decades, a cornerstone of the DSS research agenda was the use of the DSS to estimate impacts that alternative agricultural policies would have on enhancing future food security. One of the first applications of the DSS in this area was in Mali, where an agricultural sector model (ASM) was used to: (1) determine what level of productivity gains in the food crops would be required to simultaneously hold food prices at current prices (in year 2000) and feed the projected year 2015 population at current demand levels; and (2) determine what the effects of a "do-nothing" approach to agricultural technology would be on future food availability and its resultant effect on food prices.

In the "do-nothing" alternative, food prices were found to rise to alarmingly high levels compared to current prices. The average food price in the year 2015 would be nearly 280 fcfa/kg, nearly four times as large as they are today, 76 fcfa/kg. Specifically, maize would rise to 295 fcfa/kg, sorghum to 276 fcfa/kg, millet to 269 fcfa/kg, and rice to 302 fcfa/kg. The results clearly indicate the inability of existing technology, deeply rooted in traditional practices and local germplasm, to satisfy the food requirements of a society striving towards modernization.

With intensification in the food crops, the study found that food crop yields would need to increase by about 3.5 percent per annum in order to keep food prices at current levels (Year 2000) while maintaining existing demand levels. The crop specific yield increases required per annum would be: 2.8 percent for maize, 3.6 percent for sorghum, 2.8 percent for millet, and 5 percent for rice. These annual yield increases are likely to be at the outer reaches of agricultural research in the developing country context, as yield growth of around 2 to 2.5 percent has been the norm for developed countries.

With this in mind, two other alternatives were considered: (1) a 20 percent expansion of crop area, and (2) the import of food from world markets. The largest effect on reducing the required yield increases to maintain current food prices was from food imports. With food imports and crop intensification, required yield increases would fall from the above listed values to 0.6 percent for maize, 0.6 percent for sorghum, 0.6 percent for millet, and 1.1 percent for rice. Land extensification would also relieve technology’s burden on increasing food production, although not to the extent of food imports. With land extensification and food imports, required yield increases would fall from the above listed values to 1.1 percent for maize, 1.9 percent for sorghum, 1.2 percent for millet, and 3.2 percent for rice.

The results suggest that new technology alone is not sufficient to meet future needs without some type of mixed strategy that combines new technology with food imports and land extensification.

• Alternative crop production strategies for averting the effects of and managing drought among smallholders in West Africa

In many regions of Mali, the primary constraint to agricultural production is insufficient soil moisture that results from scanty and irregular rainfall. The risk of years of low rainfall preoccupy much of farmer’s decision making in these areas, and influences their choice of cropping and livelihood strategies. Based on field surveys and interviews in collaboration with our Malian counterparts, it was decided that the most relevant risk management strategy to study was the adoption of early season cultivars. These cultivars are more tolerant to drought and drought like conditions and reduce the effect of drought and low rainfall on crop yields.

Subsequent FLAM analysis of a typical short season sorghum cultivar (Malisor 84-7) using the farm stratification developed in the rapid and intensive surveys indicated the short season varieties increased farmers’ income by 35 percent (Large farmers), 28 percent (Medium farmers), 22 percent (Small farmers), and 18 percent (Subsistence farmers). More importantly, the distribution of income was shown to conform to the risk preferences of the smallholder farmer, with the majority of the yield and economic gains accruing in the average and below average rainfall years. In years of low rainfall (one standard deviation below average), for instance, food production with short season adoption would be 31 percent higher than with local cultivars (Subsistence farmers), with a corresponding increase of 18 percent in farm income (Subsistence farmers). Adoption of the short season variety was also found to enhance household food security: in the low rainfall years the increased food production brought farm households from a predicament where they were required to purchase 57 percent of their basic foods with local cultivars (Subsistence farmers), to a situation with short season cultivars where only 18 percent of the food would be purchased (Subsistence farmers).

While this study estimated significant economic impacts of short season varieties on farm income of as much as 35 percent, the subsequent impacts of farmer’s risk aversion on regional and national market outcomes could be negative (see separate study).

In the Sahel of West Africa, population pressure is expected to aggravate the problems associated with soil degradation as it places additional strain on already fragile agro-ecological systems. This will particularly be the case in areas where farmers continue land-clearing practices where the subsequent environmental consequences accelerate over time. The environmental impacts associated with the clearing of marginal lands and the removal of the protective vegetative covering were estimated in PHYGROW and SWAN and were integrated into FLAM to investigate the extent to which farmer’s economic preferences and rationale for long term planning would affect the observed propensity to clear new land in lieu of adopting more intensive production techniques.

Figure 9 illustrates the link between a typical Malian farmer’s rationale for either short or long term planning and its subsequent environmental consequences. For farmers that would rationalize short-term planning and discount future profits, land clearing provides about 20 percent higher food production and 27 percent higher incomes over the initial 15 years. Farmers that would adopt a long-term planning horizon (20 years or longer) would adopt intensive technology instead of clearing new land. Their food production would remain around 8 tons/year and their farm income would remain $4,000 per year for the initial 15 years. The environmental consequences become significant to farmers that plan over the short-term after about 25 years, where food production peaks at about 11.5 tones per year, and then falls sharply at a rate of about 700 kg/year. Rather quickly food consumption needs are not met and food purchases grow as much as 3.2 tons/year.

The findings provide quantitative estimates of outcome that are likely to be sobering for agricultural planners in both Mali as well as regions with geographic equivalence. In these areas with mildly sloping and undulating geography, the time constants associated with land degradation are so small (yield declines < .5 percent per annum) that farmers have difficulty accounting for them in their planning. Although subtle, the accumulated effect of such degradation over the course of about 15 years is large. Unfortunately, by the time the environmental consequences become significant (after about 25 years) the effects are largely irreversible and farmers’ belated attempts to increase productivity through adoption of more intensive techniques are seriously jeopardized. For planners this points to the need to diffuse new technology sooner rather than later to avoid the irreversible costs associated with environmental degradation.

• Economic and environmental consequences crop and livestock interactions in mixed farming systems

Given the competition for resources between expanding crop and associated livestock production systems in the Sikasso region, the economic and environmental consequences of rangeland conversion were estimated. FLAM was used to synthesize the mixed farming decisions of the Malian agro-pastoralist, simulating the effects that future changes in prices, production practices, and rangeland condition would have on relative profitability between crop and livestock activities. PHYGROW and NUTBAL were use to assess changes in forage, livestock and runoff associated with land clearing.

The economic calculus used by agro-pastoralists in crop/livestock decisions is critically linked to the physical condition of the rangelands. Well-maintained rangelands promotes increased livestock production and limited expansion of farming onto common grazing areas, whereas the increased costs of forage associated with degraded rangelands erodes livestock profitability and spur significant cropland conversion. For example, the village of Nadiasso, under well-maintained rangeland conditions expand their 600 ha of cropland by 5% by the year 2015 but expand cropping 200% to 1750 ha by the year 2015 if severe degradation is allowed.

Environmental effects were most notable where loss of basal area of the dominant and preferred perennial grasses occurred, especially Andropogon gayanus, allowing dominance of less desirable annual grasses. In degraded rangelands, 50% loss basal area resulted in a 9-22% reduction in aggregate forage yields (Figure 10). However, with a 95% reduction in perennial basal area and replacement by annual grasses, aggregate forage yields declined between 35-60%. Severely degraded zones with complete loss of perennial grasses and 50% reduction in annual grasses experienced forage yields reductions of 75-86% (Figure 10).

The economic effects of rangeland degradation and subsequent loss of forage were estimated by FLAM to be significant. In the study village of Nadiasso, agro-pastoral income for well-maintained rangeland conditions would be 1.6 million fcfa ($ 2,350 US) per year, nearly 3 times as large as the income would be under severely degraded rangeland conditions (0.561 million fcfa per year). Households’ livestock holdings, which are typically associated with family wealth could expand with good grazing management but decline to 30% of potential if severe degradation progressed.

The communal rangelands not only provide forage for livestock but serve as a protective barrier to the surrounding lands. The ecological benefits are severely compromised by degradation. In early stages degradation, nearly all of the rainfall is maintained on the rangelands (between 9-18% increase in runoff; Figure 10) However, when most of the perennial grasses and significant tree canopy cover are lost, runoff increases between 35-53%. (Figure 10). Lowland flooding of cropland increases and the water balance of the village ecosystem is compromised. Runoff was 4 to 13 times greater on the shallow uplands as compared to the bottomland or mid-slope sites.

Given Mali’s historical regional comparative advantage in agro-pastoral activities, agricultural polices in the future will need to be balanced and adequately consider the role of livestock in future development and perhaps, as this study suggests, in environmental management.

• Influence of climate change and alternative production systems on future risk of hunger

Studies conducted on the impact of climate change show that not only agriculture is likely to be adversely affected but it may also play an important role in the mitigation of this global phenomenon through carbon sequestration. Using its suite of economic and biophysical models, studies were done on climate change impact and carbon sequestration potential in the U.S and Africa. So far, Studies were completed on the impact of climate change studies for U.S and Mali, and a carbon sequestration study was completed for U.S. Work has commenced on carbon sequestration in East and West Africa by collecting information required for the study.

U.S. and African Climate Change Studies: Climate change studies were completed using five of the DSS models. The economic models used were Mali Agriculture Sector Model and Global Agriculture Sector Model, while the biophysical models used were PHYGROW, SWAN, and NUTBAL. The IAG members were part of the National Assessment Synthesis Team on assessment of climate change impact on U.S economy.

The results of the U.S study show that at the national level, the U.S agriculture sector may gain due to climate change leading to higher production and low prices. As a result, consumers may gain, while producer may lose. The results of the Malian study show that as a result of climate change, the risk of hunger in Mali may increase to a range of 40-49 percent of the population compared to 32 percent at present. The economic losses may range from 1 to 3 percent of Mali’s annual national income. These results are consistent with the general understanding that climate change may benefit cold climate areas, while it may adversely affect hot climate regions.

The U.S. Carbon Sequestration Study: The IAG study on carbon sequestration is the first major study that aimed at economic feasibility of carbon sequestration. The study used the DSS economic and biophysical models — respectively, GAMS and EPIC. The study focused on land-based sequestration and emission offsets. The land based sequestration included: afforestation, reforestation, and agro forestry; replacing the conventional/deep tillage with ridge tillage; and growing more perennial crops instead of short-term crops such as vegetables. The emission offsets included: concentrated livestock feed, fewer animals, and lower use of chemical fertilizer; substitution of fossil fuel use by bio-fuel such as ethanol replacing petroleum gas; and burning of biomass in the form of switch grass or short rotation woody crops instead of burning fossil fuel. The results of the study show that the U.S agriculture can potentially store about 400 million metric tons of carbon each year. Carbon sequestration offers a low cost alternative for climate change mitigation. The average cost of carbon sequestration in U.S is $10-25 per ton as against $200-250 in the non-agricultural sectors.

• Assessment of factors affecting adoption of technology in the Sikasso Region of Mali

Assumptions must be made about the adoption rate of new technology or policy to produce final estimates of impact of technology using the DSS. Estimates of adoption rates have been developed using expert opinion panels with extension and research workers and representatives of the Ministry of Agriculture. These estimates range from 20 to 40% within a five year time frame after introduction of new technology. To extend this methodology, a study is underway on the factors affecting adoption in the Sikasso region of Mali. Interviews with households and village leaders have been conducted in research led by our Malian collaborators. Field data are being analyzed using a Probit-Logit approach. Results from this study will improve the accuracy of estimation of adoption rates and provide insights into how adoption might be enhanced or facilitated.

• Global Decision Support System — incorporation of national and subregional data and models into the FAO-WAICENT system

As the databases and models of the DSS are completed over year five of the project, they are being organized and packaged for the FAO Worldwide Agriculture Information Center where they will be made available through their website to users in both developing and developed countries. The information and methodology will also be available to the several departments in FAO with whom we have collaborated over the duration of the project. The use of the DSS by national and regional decision makers is targeted to applications related to improving food security and natural resource management. These applications are intended for use in Mali and Kenya to develop plans and evaluate progress toward achieving the goals of the World Food Summit and the Convention to Combat Desertification. The Agriculture, Economics and Sustainable Development Departments in FAO have advised on the development of these models, provided data and expert opinion, and will now have ready access to the products for their further use in helping developing countries achieve the goals of these and related international agreements.

• Application of the Decision System for assessment of the impact of the FAO Special Programme on Food Security in Mali

This program, applied through South-to-South exchange of experience and practice is aimed at short term improvement in food security at the household and village level. The program has been in place for a number of years in many developing countries. The impact at local levels has been judged to be very good. There is now need to assess the potential of extending the principles of such program from local to broader areas of scale. The DSS is well suited to this kind of analysis. The data from the Mali Programme Special Securite Alimentaire (PSSA) were used as input to the updated Mali agricultural sector model to estimate the sub-national and national impacts of adopting the program. The output estimated the economic impact resulting from the PSSA as well as its potential for reducing the risk of hunger. This can be regarded as a pilot study to evaluate the use of the DSS for evaluating the SPFS in other countries and perhaps at a global level. The results are being finalized.

• Cross cutting assessment of the application of the DSS in East and West Africa — geographic, socio-cultural, and economic comparisons and contrasts

In keeping with the SANREM effort to assess the key lessons learned from specific regional studies that have transcending applications on a global basis, a key part of the DSS plan is to compare and contrast the application of the DSS in East and West Africa.

The choices of Mali and Kenya provide a rather dramatic comparison for sub-Saharan Africa since their geographies and climates are so different. The Rift Valley is a diverse region with a great deal of climatic and geographic richness, whereas Mali is much less diverse. The evaluation of USAID sponsored technology by CRSPs and IARCs and with crop and livestock systems provides further diversity to enrich the comparison. Very similar experimental designs in the two locations have been employed for development of the DSS and the specific applications studied were selected to permit a comparison of results. The results are intended to form the basis for application of the DSS as a global decision tool, showing how the models are made specific by their inputs — both quantitative and socio-cultural. This study will be completed when the ongoing Rift Valley study is finished.

SANREM Management defines "impact" as fundamental changes in development indicators that affect a large number of target population and/or environment as a result of the adoption or use of technology, methods, data or information produced by your research projects. Development indicators include improved standard of living, food security, improved health, natural resource conservation, institutional and human capacity building, and improved policy environment. The SANREM definition of impact requires that the research not only produce useful results, but that users put results into practice. Thus, it implies that technology transfer will be successfully accomplished. It is well known that this is often a daunting goal in the developing world as the factors which affect adoption of new knowledge and technology often lie outside the reach of the research effort which generates them.

Adoption is often a function of scale — highly localized experience and innovation can often be successfully applied at that specific location. The application of principles and policies that affect larger levels of scale and government are often not as easily achieved. Yet, this larger goal is necessary for there to be aggregate impact that affects outcomes of peoples’ lives on a scale that addresses the critical questions of food security, poverty reduction and sustainable NRM. A balanced development portfolio is obviously desirable. The DSS enhances the capacity of decision-makers in developing countries at multiple levels of scale as they consider impacts and trade-offs for policy options and the planning and use of technology to enhance the sustainable production of food. Their success will be measured by the ability to increase the quantity of food (availability) and generate new economic activity from agriculture (reduce poverty and create access to food). For enhanced production of food to be sustainable, those considering new policy and technology must deal with the combination of production of food and the sustainable use of natural resources. The DSS, as an integrated suite of economic, biophysical, and environmental models, provides decision-makers the capability to make such rationale and balanced decisions.

The successful development of the suite of models in the DSS will be achieved at the end of the five year effort, with all major goals completed. The databases to use these models have been developed and are being delivered to the intended users in national and regional programs. The DSS represents a fundamental enhancement of the methodology for impact assessment and decision making for both developing and developed countries. The DSS provides new capacity for analysis through (1) linking economic, biophysical, and environmental models to produce a more holistic output, (2) providing methods for scaling up or down between various levels of government and geography, and (3) using the concept of geographic equivalence to project the adaptability of new methods or technology from their origin to other regions or countries, (4) provision of methods that allow for assessing future outcomes of current decisions over time, (5) introducing risk assessment and the impact of risk aversion on outcomes, and (6) provision of well organized and validated natural resource and agricultural databases that provide very substantially increased access to critical information for decision making.

The usability and utility of the DSS has been assured by the development of user friendly interfaces that allow the models and databases to be applied with nominal background and experience. By the end of the project, a cadre of analysts and scientists with multidisciplinary backgrounds will be trained and able to use the DSS in ongoing applications in Kenya and Mali. Decision makers at levels of government from the offices of the president to first level administrators in government ministries and research institutes will understand the use of the DSS and have in hand multiple case studies that have demonstrated its utility by responding to tasks defined by these users. In West Africa, commitments have been made by the CILSS to develop the capacity to use the DSS in the member nations of that regional organization to plan and evaluate progress towards achieving the goals of their new four year strategy.

The DSS methodology and related databases will be delivered to the FAO-WAICENT for access and use by other developing and developed country users, including the several departments within FAO where collaboration has been ongoing over the duration of the project. The methods provide the ability to evaluate options to achieve the goals of several relevant international agreements such as the World Food Summit and the Convention to Combat Desertification. The methods have been presented and are available to the CGIAR Standing Committee on Impact Assessment and the Center staffs that conduct such analyses in the IARCs as well as to the USAID Global and Africa Bureau technical managers and to the USAID CRSP Directors for use in assessing the impact of the various research portfolios managed by these groups.

The methods are having immediate utility in the U.S. through their applications in related state and federal governmental planning and decision making as well as at local levels. For instance, the DSS is being used to develop criteria and management methods for administering a new federal program for forage insurance. Ongoing applications of elements of the method are used to advise livestock operators in 29 states on the status and availability of forage. These methods are also being adapted for use in developing planning and management models for addressing the very pressing issues related to agricultural bioterrorism. The methods for assessing the overall impact of climate change and of options to reduce the predicted impact of greenhouse gases have been jointly developed with this project and are receiving national and regional use in this country. The fundamental breakthroughs in analytic capacity to conduct integrated ex ante and ex poste impact assessments are being used to plan and evaluate research and extension activities in the Texas A&M University System.

Have fundamental changes in indicators identified by the SANREM ME been achieved? Would fundamental changes in these indicators be expected at this point, given that this project has been in the research and development mode for four of its five years and the fifth year is now being devoted to building the capacity to use the product? The impact of improved decision making at various levels of government can substantially influence outcomes of actions taken but may not uniquely contribute to changes in food security, poverty level, and natural resource management. The matter of attribution requires careful consideration in this type of assessment. In the case of the commodity oriented CRSPs or IARCs, it is relatively straightforward to quantitate impact — how many hectares are planted to the new crop, what is the quality and quantity of food produced and what has been its economic impact? Similar assessments of the impact of improved methods and decision making do not easily lend themselves to this kind of analysis. The DSS development has engaged decision makers from inception to application. The main focus has been on national level decision makers, but the methods apply to other levels of government and scale.

This project involved 13 specific applications of the DSS to quantitatively evaluate the impact of policy and technology options identified by decision makers. These studies were done to provide operational scenarios in which to develop the models. They also provided credible demonstration of utility to decision makers. They have direct implications on contemporary strategic issues for Mali and Kenya with direct application to decisions about food security, poverty reduction, and sustainable use of natural resources. The DSS methodology has been used to specifically evaluate both ex ante and ex poste impacts of technology and policy options and show the quantitative impact of these alternatives on food security, poverty, and NRM. In addition, further studies have been conducted by collaborators outside the SANREM project (see Training and Institutional Strengthening).

Have explicit decisions by policy makers emanated from these studies — and have they been put to practice in the field? Not yet. But, specific case studies have provided quantitative input for the decision making process, which is ongoing. The impact of this process has its ultimate effect on farmers and rural communities— with policies, regulations, investments and interventions that affect household level production of food and fiber. But they also have direct impact on the rapidly growing population of urban households in the developing world.

Absent support to continue these studies as proposed in SANREM III, Texas A&M is actively seeking alternative funding to continue the process of capacity building and mentoring for the use of the methods in both East and West Africa.

In summary, a highly innovative integrated set of models and related databases has been developed and its utility and usability demonstrated for application in Mali and Kenya. With additional support, the ability to use these methods will be extended to other countries in SSA. Specific applications of the method have been made for highest priority policy issues identified by senior decision makers in these countries. Results of these analyses have direct and meaningful relevance to the decision making process in these countries. The results obtained make direct contributions to the decision processes related to development of an improved standard of living, enhanced food security, improved health, natural resource conservation, institutional and human capacity building, and improved policy environment. Cadres of analysts and scientists in both Mali and Kenya are being trained to use and apply the DSS in government and national research institutes. All models and databases either have been or will be delivered in usable form by the end of the fifth year.

The process of dissemination or technology transfer was a fundamental part of the strategy and planning of this project. The priority of the commitment of the host countries to the World Food Summit and Convention to Combat Desertification and the linkages of these commitments to the activities in FAO provided an entre to key national decision makers to discuss methods to assess progress toward meeting their obligations as signatories to these and related international agreements. The Comite Permanent Interstats de Lutte Contre la Secheresse Dans le Sahel (CILSS) through the Institut du Sahel (INSAH) sought the development of the DSS for regional applications before SANREM II was initiated. Senior officials of the governments of Mali and Kenya were engaged to define their needs for this kind of analytic tool. They were involved in developing the strategy, participating in the ongoing workshops where methods and results were presented, in defining and evaluating the priority case studies done to demonstrate the utility of the model, the commitment of national resources to the collaboration and the subsequent actions toward capacity building and institutionalization of the methodology.

NGOs were also involved in the workshops and planning. For example, CMDT in Mali was instrumental in providing data from the Sikasso region and have stated the intent of applying the methods to their ongoing planning and advisory functions for farmers in this region. The Kenya Institute for Public Policy Research (KIPRA), the WFP and other players in Kenya are active participants in the development of the DSS in that country. Colleagues in the various ministries of government and in the national research institutes in both Kenya and Mali have been active collaborators in the planning, conduct, analysis, and interpretation of the results of the multiple studies that have been conducted to develop the DSS. Both long and short term training has been provided for host country colleagues at Texas A&M. In turn, these people are taking active leadership roles in subsequent capacity building at national and regional levels. A commitment has been made by CILSS to seek support to build capacity to use the DSS in the member nations of that organization for both national and regional applications. There is active interest in the application of the methods in Tanzania, Ethiopia, and Uganda.

The dissemination of results to "wider user communities" is intrinsic to this project and is described in detail in other parts of the report. Methods for scaling both economic and biophysical results are covered in detail. Methods for applying technology or policy options outside the immediate environs where they are developed are also described in detail elsewhere in the report. We believe the effective linkage from farm to national and multinational levels of scale is a major contribution to the goals of SANREM II that were stated by the sponsor. As a result of these efforts, we will have in place by the end of year 5 the capacity and know how to use the DSS at these multiple levels of scale by a variety of different operators.

In the appendix of this report there are three additional sections dealing with the dissemination and technology transfer process: (1) two matrices that summarize the various parts of the governments of Mali and Kenya that have been integral to the development of the DSS — showing the relevant aspects of their function and how the DSS contributes to that function, (2) a brief listing of the models and databases that have been or will be delivered to each country, and (3) a summary of the common modeling environment methodology which provides the user interface to the models and databases that have been developed.

The project approached human capacity development as a fundamental and integral part of the development of the DSS. National partners have been involved in developing the needs, conceptual framework, design, development, evaluation, and application of the DSS. Several levels of engagement have been involved. Senior decision makers in the governments of Mali and Kenya have actively participated in developing perspectives on needs and applications and have been interactive with developers as progress occurred. Collaborators that were actively involved in field studies and analysis have been actively engaged throughout the development and application of the DSS. Progressive workshops have been held in both Kenya and Mali that provided for awareness and general appreciation for the DSS by a larger number of potential participants from within the governments, universities, and national research institutes.

There have been three major workshops in Mali and three in Kenya that have provided opportunity for exposure to the DSS for both national and regional participants. Approximately 30-45 participants have been involved in each of the workshops. Regional participants have included scientists and analysts from adjacent countries as well as active participation by the INSAH and AGRHYMET as part of the CILSS organization. The CILSS NRM pole has been provided briefings on the DSS as part of the awareness program that has led to the decision to apply the DSS more broadly in West Africa. These workshops have also involved several NGOs that have been involved both as collaborators and as workshop participants. Consideration has been given to gender balance in the workshop participants with an average of 5-7 females actively participating in 30 person meetings.

One Kenyan scientist and two Malian scientists have received intensive short term training at Texas A&M in the use of the models. These are among our most involved collaborators. They are now actively engaged in planning and conducting capacity building programs in both Kenya and Mali that will last over a period of about six months during year five of the project. These national level programs will provide training for interdisciplinary teams comprised of economists, biologists, and GIS-natural resource scientists who will train together and later work together in the application of the DSS in both government and research institutions. We will train four teams of three persons in Mali during the period July 2002 to February 2003 with primarily in-country programs and active use of internet-based instructional materials. We will train a team of 4 Kenyan scientists and analysts with an intensive one-month program conducted at Texas A&M in the fall of 2002.

The DSS project has been engaged at the regional level in West Africa for more than five years through the Comite Permanent Interstats de Lutte Contre la Secheresse Dans le Sahel (CILSS), and its Institute du Sahel (INSAH) and, to a lesser degree thus far, the Regional Centre de Agrométéorologie, et Hydrologie (AGRHYMET). Following a series of meetings with the CILSS Secretariat over the last year, the methods were presented to representatives of the CILSS nations that are involved in planning the implementation of the CILSS strategy at regional and national levels. This meeting, held in Dakar, Senegal in June 2002, led participants to enthusiastically support use of the DSS in planning and evaluating the options involved in addressing the key food security, poverty, and NRM use issues at national and regional levels that are embodied in this strategy. They recognized the need for substantial capacity building at both national and regional levels to enable the use of the DSS. A proposal has been prepared for CILSS to request support to initiate a comprehensive training program on the use of the DSS. This program will build on the experience in developing and using the DSS in Mali, which has served as a pilot study for the broader regional application. Scientists from IER and their regional collaborators will be trainers in the program, along with a continuing active involvement of Texas A&M scientists that have been previously involved in SANREM II.

There follow two specific examples of involvement of our national partners who have been trained in the use of the DSS.

The Kenyan collaborators on the DSS, led by Dr. Robert Kaitho, now have full capability to use all models and related databases in the DSS. They have used the methods in conducting new studies with other collaborators. For example, a study has been published with the University of Nairobi, the International Livestock Research Institute, and the Southern Africa Centre for Cooperation in Agriculture Research and Training on an economic analysis of cross breeding programmes in cattle for SSA. Dr. Kaitho continues an active appointment in the Kenya Agricultural Research Institute and with the International Livestock Research Institute as well as an active collaboration on both the SANREM DSS project and the Global Livestock CRSP Project on Livestock Early Warning System. He is actively engaged with decision makers throughout the Government of Kenya in conducting research and planning the future use of the DSS in ongoing applications.

Mr. Alpha Kergna has been the national coordinator in Mali for the DSS development over the past five years. He is a master’s level graduate from Texas A&M and has received intensive short term training in the use of the models embodied in the DSS. He conducts independent research using the GAMS based ASM that is part of the DSS and is an active participant at the regional level in developing and using impact assessment methods for assessing both technology and policy options. He is a principle instructor in previous and current capacity building program for the DSS in Mali. He is an active participant in the planning of the CILSS capacity building effort and will be a key person in its implementation through the establishment of a regional training center at the Sotuba Station in the Mali IER. Through his efforts, a team of both national and regional (Sikasso) based scientists have been brought together for active collaboration on the most recent studies of food security and natural resources in the Sikasso region of Mali.

PhD degrees have been awarded to five international scientists who did their thesis work on the development of the decision support system. Three Master’s level persons did their research on the DSS. A number of other students are working on the DSS as they work on advanced degrees. Three of the PhD students have returned to their home country, one is now in a post-doctoral program and the third is a faculty member at Iowa State University where he has recently published an article in Science on Carbon Sequestration.

Scientists and Collaborators: The Texas A&M Center for Natural Resource Information Technology (CNRIT) is the institutional home for the Impact Assessment Group. This is the nexus of faculty, staff, and students that are involved in the DSS from the U.S. side. Over the duration of the project, an average of 7 senior faculty have been involved in the project. Most of these have been engaged in the ongoing work of the project, others have supervised graduate students and research associates that have been more directly involved. An average of about six graduate students- research associates have been involved part time in the project over the five years. There has been a substantial turn-over in both faculty and students during the project. CNRIT provides connections to the broader resources of the TAMUS, many of whom have contributed to the project. The number of host country scientists involved in the project has varied with time. A core group of two to four scientists were involved part time in Mali and the same number in Kenya on a more or less continuous. These colleagues were augmented with other collaborators from both the government and national research institutes as field studies and related analysis were performed. These amount to between six and eight colleagues. The IAG and the core host country collaborators were involved in all aspects of the project. The colleagues who participated in specific parts of the project were involved in planning, conducting, and analyzing results from these studies. All have been involved in the relevant workshops. Overall resource allocation has been mainly done by the IAG members. Host country collaborators were actively involved in planning and using the resources allocated for in-country research. There has been active collaboration with the Malian based participants in the INTSORMIL and Peanut CRSPs in conducting the case studies of the impact of their research in the early part of this project.

National Governments: National governments from the Offices of the President, through the Permanent Secretaries of the relevant Ministries, to the directors and staff involved in analysis and line functions in the Ministries have been actively involved in planning and evaluation of the project methods and applications. Analysts and line officers within the action agencies have participated in experimental design, conduct of research, evaluation of results and communication with senior officials on outcomes.

Non-Governmental Organizations: NGOs in Mali, such as the World Food Programme and the Sasakawa Global 2000 project were actively involved in the ongoing workshops and offered useful advice on approaches for methods and scope of application studies. The Kenya Institute for Public Policy Research and Analysis (KIPPRA) is a para-governmental organization that is related to the Ministry of Finance and Planning. Engagement with this group is providing a mechanism for institutionalizing the DSS in a stable environment for the future.

International Agricultural Research Centers: The International Livestock Research Institute was an active collaborator in an early application of the DSS to evaluate the impact of smallholder dairy technology in Kenya. An informal collaboration continues with ILRI and they provide substantial logistical support for the activities of this group. ICRISAT has an active involvement with the Mali IER in the area of spatial analysis and GIS research. This project has collaborated with ICRISAT at the Sotuba location on the development of the models and acquisition of several important databases for the DSS.

U.S. Based Public and Private Sector Partners: The general area of research embodied in the DSS project is supported substantially by federal and state funds in the U.S. These are noted under the section of this report on leveraging. There has not been direct involvement with private sector partners from the U.S.

Host Country Private and Public Sector Partners: The Malian CMDT, a para-governmental organization that functions as a cooperative for cotton farmers in the Sikasso region has been an active collaborator in the Sikasso study, providing village and farm level data on production yields, costs and prices and expert opinion of factors affecting adoption. They anticipate using the DSS methods in their business.

Texas Agricultural Experiment Station $200,000

WSSD 48,000

Rockefeller Foundation 67,000

USDA NRCS 200,000

Total (2 years) 515,000

- Weather tools

A near real-time weather data website was developed in cooperation with the Livestock Early Warning System Project of the Global Livestock CRSP. This website enables users to download near real-time weather data for the continent of Africa (http://cnrit.tamu.edu/rsg/rainfall/rainfall.cgi). Since January 18, 2002, the webpage has had 1449 downloads, totaling 18.75 megabytes of data, with an average of 13 downloads a day. Domains most frequenting the website include 1047 downloads from .edu, 75 from .com, 16 from .gov, 67 from .net, and 2 from .org. These queries were also from 12 countries.

In order to develop long-term climate profiles on a spatially explicit basis, a tool was developed to geo-correct weather generator coefficients to user-defined points for both East and West Africa. The geo-correction tool extracts weather parameters for user defined points and replaces these same coefficients in the weather generator file for the nearest WMO weather station having similar climatology. The weather generator can then be run for a pre-defined set of years and the output can be used as driving variables for simulation modeling.

All models have been delivered to our partners in each country along with all run files. These run files have been cataloged in a "read me" files and supplied on CDROM. These same data are available on the http://cnrit.tamu.edu web site. These same data are also reflected in .shp files for use in the ACT tool or ARCVIEW. Much of the data can access via ACCESS databases or within the models in their native data structures. The various databases and model runs are highlighted below.

West Africa

1. Comprehensive spatial analysis of the country of Mali with creation of simulation zones reflecting unique combinations of soils groups and climate surface clusters.
2. Complete parameterization of dominant soils across the entire country of Mali and Senegal ready for use in the EPIC/SWAN crop model. Subsets of these data were transformed into data structures for the PHYGROW model. The PIRL soils classification system was captured and filled in digital form.
3. Complete parameterization of all the major weather stations of Mali and Senegal will all daily data filled via the WxGEN weather generator.
4. Parameterization of over 100 native grass, forbs and woody species of rangelands in the Sikasso region of Mali including growth attributes and preferences by cattle, sheep and goats.
5. Complete parameterization of monthly nutrient requirements and physiological profiles of all major classes of livestock for major breeds of cattle, sheep and goats in West Africa.
6. Updated secondary economic data for all major district and province regions of Mali and Senegal.
7. Farm level data on production, budgets and cost/revenues on pre-stratified farms across Sikasso, as well as farms in other major provinces of Mali.

East Africa

1. Comprehensive spatial analysis of agro-ecological simulation zones reflecting the way that Kenya, Uganda and Tanzania think of their natural resources and agricultural production zones. This however, is build upon biophysical data from climatic surface analysis and soils mapping.
2. Complete parameterization of soils throughout Kenya, Uganda and Tanzania in digital form, transforming paper data to model ready data.
3. Complete parameterization of all the major weather stations in Kenya with most of the major weather stations in Uganda and Tanzania stratified by agro-ecological zone.
4. Parameterization of over 200 native grass, forbs and woody species of rangelands in throughout the highlands and rangelands of Kenya, Uganda and Tanzania. including growth attributes and preferences by cattle, sheep and goats.
5. Complete parameterization of monthly nutrient requirements and physiological profiles of all major classes of livestock for major breeds of cattle, sheep and goats in East Africa.
6. Updated secondary economic data for all major district and province regions of Kenya to 1999 and 1996 in Uganda and Tanzania.
7. Farm level data on production, budgets and cost/revenues on pre-stratified farms across agro-ecological zones of Kenya and Uganda.

The following number of model runs are available and delivered to our partners in East and West Africa.

EPIC INSORMIL CRSP Mali/Senegal 3200

EPIC PEANUT CRSP Mali/Senegal 300

EPIC SMALLHOLDER DAIRY Kenya/Uganda 1200

PHYGROW SMALLHOLDER DAIRY Kenya/Uganda 285

SWAN EXT/INTENSIFICATION Sikasso, Mali 3000

PHYGROW EXT/INTENSIFICATION Sikasso, Mali 985

SWAN INTENSIFICATION Rift Valley, Kenya 2700

PHYGROW INTENSIFICATION Rift Valley, Kenya 635

NUTBAL EXT/INTENSIFICATION Mali 3460

NUTBAL SMALLHOLDER DIARY Kenya 3460

NUTBAL INTENSIFICATION Rift Valley, Kenya 3460

These data were also repackaged in spreadsheets interfaces where the large number of model runs were condensed into mathematical meta-models which were linked to pre-parameterized agricultural sector model (ASM) and farm level analysis model (FLAM) for Mali and Kenya reflecting the analysis for the intensification studies in Sikasso and Rift Valley.

Assessing impact of technologies and policies requires the use of a suite of decision support models that address the complexity of economic systems at the sector level, farm level economics and human welfare, crop production, grazing land production, livestock performance and resulting environmental processes. Typically "integration" of these processes involved manual transfer of data files between applications or limited digital integration in a subset of modules. Further, there was limited ability to modify models in a manner that allowed tighter "digital" integration. There is growing need within SANREM and with other partners, including FAO, to package a number of different research simulations together to develop a more detailed and holistic view of non-homogeneous activities and environments. This need to package integrated suites of models so that they can be run on a single computer or internet/intranet, led to the development of the Common Modeling Environment (CME) concept. CME is an evolving set of information technology that is modular and designed to grow in sophistication as needs are identified within organizations. This system brings cross-platform delivery, a scalable

distributed computing model, and shared common input data to many research models with minimal model modification (Figure 11) Using a custom scripting language or "middleware", a model server process can be run on any platform that has a JAVA virtual machine installed, allowing incorporation of stand alone models without undue stress on research model developers. This product is available on the internet for download at http://cnrit.tamu.edu/CME. Since its release in 1999 there have been 301 downloads of the software to over 32 countries with 18 downloads to .edu, 59 .com and 98 unresolved locations. The application of the technology was demonstrated by linking the grazingland production model, PHYGROW, and the crop model, EPIC, to the CME environment.

Partnering with CIMMYT and USAID Office of Foreign Disaster Assistance Almanac Characterization Tool (ACT) produced Country Almanacs for 13 countries in Africa. They include: Mali, Kenya, Botswana Mozambique, Uganda, Zimbabwe, Ethiopia, Malawi, Tanzania, Zambia, Angola, Liberia, and Sierra Leone. In addition country ACT’s and continental wide data for Africa is maintained and available on the Characterization Assessment and Application Group (CAAG) web site at www.brc.tamus.edu/char/. This web site is accessed through secured passwords before data can be downloaded. To date we have issued 230 passwords to scientist and GIS specialists from 45 countries.