AEZ entails the linking of a number of logical procedures to arrive at a quantitative estimate of yield or production for a particular agro-ecological zone or agro-ecological cell. Such a methodology is particularly suited to computer- ization, and mainframe computers were used in the early FAO continental scale studies (FAO,1978) because of the large amounts of data involved. Subsequent- ly the methodology has been implemented on minicomputers, and most recently on microcomputers. Most advanced AEZ investigations incorporate a series of databases, linked to GIS and dedicated computer models, which have multiple potential applications to natural resource management and land-use planning.
Software tools can be grouped into databases, geographical information systems, models, and integrated packages.
In the compilation of inventories of land and land use, AEZ studies normally use large quantities of data. For direct viewing of information and for access by models for land suitability and productivity assessment, these data are most conveniently stored in databases. Databases can either be constructed using commercially-available software, or dedicated pre-programmed packages can be used. Relevant databases available from FAO are:
� multilingual soil database (FAO/ISRIC/CSIC, 1995)
� crop environmental requirements database (FAO, 1994b)
� land use database (de Bie, van Leeuwen and Zuidema, 1995).
Most recent FAO AEZ studies have used databases incorporated into shell programmes such as the Agricultural Planning Toolkit (APT), which is described under the heading of Integrated Packages (p. 61).
Once the essential data are stored in the databases, AEZ uses models to derive quantitative outputs describing productivity and land suitability. Models represent a simplification of a more complex reality and the level of detail of the model should be consistent with the objectives of the study, the availability of data, and the knowledge base from which inferences can be drawn. Summary mechanistic models, based on relationships between external variables and the intermediate or ultimate products are particularly suited to land evaluation (Dumanski and Onofrei,1989). As plants obey similar physiological rules, sets of parameters can be input for individual crops, and for inputs and operations which describe the production system, and the results can be directly compared for different production systems and different land units or AECs.
A number of suitable models are available for use in AEZ studies. CYPPAC (De Baveye,1988), and CYSLAMB (crop yield simulation and land assessment model for Botswana) (Tersteeg, 1994) have been developed in the course of FAO projects. An updated version of the former program is incorporated in the APT shell. Most national AEZ studies use simpler models for crop productivity estimation in which the crop water balance is not crop specific. Further models are used to assess erosion induced production loss, and to estimate livestock and fuelwood productivity.
Geographic information systems have emerged as powerful tools in the management and analysis of the large amount of basic data and information, statistical, spatial and temporal, needed to generate in a flexible, versatile and integrated manner, information products in the form of maps as well as tabular and textual reports for land use decisions. In recent years FAO has been developing GIS in linkage with its agro-ecological zoning and similar models, applying these to tackle issues of land, food and people at global, national and sub-national levels. So far the applications have mainly addressed issues linking land-use outputs with other development goals in such areas as food production, food self sufficiency, cash crop requirements, population supporting capacity, taking into account soil fertility constraints, soil salinity, soil erosion risks and land degradation hazards. Good progress has been made in developing GIS- based tools for land resources planning, management and monitoring at different scales.
The development of these and other related applications involve the analysis and interpretation of large quantities of biophysical and socio-economic data, statistical, spatial and temporal, in order to produce the diverse kinds of information products required in the form of images, maps and both tabular and textual reports for decision making at the various application scales of interest. Up-to-date computing tools of spatial analysis allowing easy access to data and information and their manipulation are necessary to produce these.
Rapid development in information technology in the last decade has created a unique opportunity for the development of such a tool in the form of a multi- purpose land resource information system (LRIS) which can be used to generate quickly and efficiently various kinds of information according to the require- ments of different users. The LRIS contains computerized databases, models, decision-support tools and a user interface to facilitate its operation.
A GIS is the central element in the configuration of a LRIS. GIS's utility derives from a capacity for dynamic functionality based on the following three main qualities:
1. the physical computing capacity to manipulate data, including overlay, join, disaggregate;
2. the related capacity to query the data by formulating hypotheses for testing assumptions, defining potential relationships and developing theoretical constructs;
3. the capacity to relate two-dimensional and three-dimensional location of earth features, including atmosphere, lithosphere/hydrosphere/ecosphere, along with dynamic (space/time) four-dimensional processes, such as represented by functional operations of systems of land resources appraisal, planning, management and monitoring.
GIS/LRIS is a multidisciplinary undertaking which integrates databases of various kinds and sources, models for data analysis, decision-support tools, computer hard and software and the human resources and institutional framework to operate the system. Remote sensing provides data and maps on land cover and land use and enables rapid and efficient monitoring of land use change, which is an essential element of land degradation assessments and a determinant of land use sustainability.
The integration of AEZ and GIS, in combination with procedures and expert guidance, enables AEZ analysis to be performed more efficiently, and allows a flexible presentation of results according to user needs. The FAO AEZ study in Kenya (FAO,1993a) developed an integrated software package which could be adapted for use elsewhere, provided the expertise is available to reset the parameters. Alternatively, APT is a package which integrates databases and models, but the results require separate importation into GIS.
The integrated systems used in the Kenya AEZ study have two principal components :
� a computerized land resource database;
� a set of (mainly empirical and heuristic) models in the form of computer programs.
The land resource database is obtained by combining various data layers (map and tabular data) on the physical aspects of agricultural environments such as soil, landform and climate. The models are used to create the land resource database, calculate land suitabilities and land productivity, and to determine optimum land resources allocations (Figure 8). Various outputs are generated in both tabular and map form. The power of the AEZ methodology is based on the multipurpose integrated resources database it creates.
The linkages between GIS and AEZ models can be called ad hoc and partial. GIS and models are developed separately. Map input/overlay and map output capabilities of the GIS are used for preparation of the land resources database required by the models. Model processing is outside the GIS. Data flow from the GIS-created databases into the AEZ model and vice versa. Modelling results are transferred to GIS for further processing and presentation.
The software package used in the detailed country AEZ methodology consists of five computer programs to implement the AEZ models and a number of utility programs of various kinds related to database management, statistical analysis and display of results. The AEZ programs analyse land suitability and land productivity including cropping patterns, linkage to livestock and forestry production systems and soil erosion considerations. A linear programming program for land-use optimization at cell and district levels is incorporated in the package.
One major area of development has been in applying optimization models to sets of AEZ/GIS outputs in order to examine alternative regional or district level land-use patterns. Such models suggest feasible land-use allocation patterns that best satisfy specified development objectives, e.g., target food consumption patterns, population supporting capacities or rural employment levels. A mathematical programming approach is taken as there are many feasible land- use allocations e.g., maximize population supporting capacity (production of calories and proteins and the cell level), subject to a district level crop mix constraint, and a district level limit on the use of fertilizer.
The continued development of AEZ/GIS has also served to expand the spatial ranges, or scales, of its application. While the underlying concepts of AEZ are valid at any scale, the specific methods and tools of implementation must often differ in order to reflect the changing nature and complexity of decision making at national, district, farm and even plot level.
AEZ/GIS approaches are suited to any application in which the relationship between land resources and land uses needs to be explored - either in the context of assessing the suitability of land resources for specific uses, or of assessing the likely impact of those uses on the land resources themselves. Furthermore, the ways in which these relationships can be explored are constantly being enriched. Other applications in the policy analysis and planning areas pose "what if ....?" questions. The two main types of questions are: (1) what if I could modify one or more land resource characteristics? (e.g., by terracing, drainage, fertilizer application, liming) or (2) what if I could modify current or proposed land-use characteristics? (e.g., by the use of genetic materials that are more drought resistant, or that have a shorter growth cycle, or by the use of more machinery and less labour, or by the use of crop residues for feed and not for mulching). AEZ/GIS can estimate the changes either in land-use suitability or in environmental degradation hazard that arise from the "what if ....... ?" scenario being tested. The broader socio-economic costs and benefits of proposed modifications can then be evaluated. For this type of application a GIS and model are developed in close interaction. The model is implemented using exclusively input, processing and output functions of the GIS.
This methodology continues to develop, and the further recent enhancements include the following:
� Improved model of climatic data analysis to take into account the effects of cold temperatures in LGP calculation.
� Refined models of crop suitability to:
� take into account CO2 enrichment and its effects on rate of photosynthesis and crop water use efficiency in the biomass calculation model depending on crop cycle length;
� better evaluate agro-climatic constraints and quantify soil moisture deficit at various stages of crop growth;
� enable artificial increments in temperature and precipitation under existing and evaluated CO2 concentrations to test the sensitivity of the AEZ models to climatic variations;
� enable inclusion of sustainability considerations in the formulation of the planning scenarios;
� fully integrate potential evapotranspiration (PET) calculation, length of growing period (LGP) determination (water balance model), biomass and yield calculation into suitability/productivity assessment.
The latest version of the integrated GIS/AEZ system is shown in Figure 12.
FAO is preparing an improved tool, which incorporates these upgraded AEZ models and multi-criteria decision support techniques for a more generalized use in different agro-ecological and socio-economic settings to provide more effective assistance to various stakeholders in their land-use decision making and land use-negotiations. The software will be able to run on PC computers which are readily available in developing countries.
