Cooperators:
Blackland Research Center
Texas Agricultural Experiment Station
Temple, Texas
Period:
Budget:
Objectives . Methodology . Results
Comments:
Please send your comments/questions by email to Dr. Arnold or Dr. Srinivasan or
Dr. Follett
The analysis involved one of the most widely applied and accessible databases, the Major Land Resource Regions (MLRR's) and their subunits, the Major Land Resource Areas (MLRA's) were done. MLRA's are spatially organized information on land as a resource for farming, ranching, forestry, engineering, recreation, and other uses. The databases contained environmental variables (weather, soil, and topographic attributes) which were treated as "independent" or "driving" variables. The "response" or "cropping data" variables considered included crop yield parameters and types of crops grown. In addition, latitude and longitude information to locate Agricultural Experiment Stations (AES) and Agricultural Research Service stations (ARS) on maps was also collected.
Three approaches were used to identify the agroecozones: (i) Delineation using environmental variables and MLRA boundaries, (ii) Delineation using environmental and cropping variables and MLRA boundaries, and (iii) Delineation using environmental variables only.
(i) Delineation using environmental variables and MLRA boundaries:
The environmental variables considered to develop mapping information included precipitation, potential ET, heat units, available soil water holding capacity, soil depth, water per unit of soil depth, curve number, and slope. Climate, soils, and topography data bases were used to obtain these variables. A clustering analysis was performed on the standardized data using the SAS cluster procedures. The cluster procedures started with individual cases, and at each subsequent step, the two most similar cases or clusters were merged. Initial number of clusters to examine were based on "jumps" and/or a change in the trend of the maximum distance measure indicating a shift in the dissimilarity incorporated into each cluster had occurred. To avoid bias, a minimal level of scrutiny was applied; only minimal changes were made to account for "bad" data, outliers, and improper clustering.
(ii) Delineation using environmental and cropping variables and MLRA boundaries:
This section applied the theory developed within ecosystem science to the problem of deriving a framework for agroecozone delineation. Statistically, "environmental" variables are the independent variables while the "crops" are the dependent variables. Conceptully, the crops are not the only components of the ecosystem being studied. Pests and soil nutrients are examples of other components, as are farmers who are influenced by other parts of the system, including social, political, and economic factors. Weather does not directly determine what is grown but rather farmers decide what crops will exist where. For this section, the focus was on the response of the agroecosystem to the environmental variables; that is the distribution of crops. The ultimate goal was to determine the relationship between the environmental variables and the crop responses.
(iii) Delineation using environmental variables only:
The two general guidelines followed for this approach were: (a) The zones should be based on environmental variables that specifically influence the agricultural research and related activities, and (b) The separation of ecoregions should be based on remotely sensed satellite imagery products developed for landcover characterization. The clustering procedure used was run in an image processing system to obtain the best spatial resolution. Several subsets using less than eight parameters at several levels of classes (zones) were run to obtain the optimum number of classes. The signature file is a set of spatial data extracted from each Area of Interest (AOI). The four major subsets consisting of areas in the west included Western Forest, Alpine Tundra, and Barren lands. Once the number of classes was decided for the AOI's, the classes were defined by conducting a supervised classification on the rest of the area outside the AOI's.
This study indicated that the currently available tools (e.g. GIS, simulation models, statistical procedures) and databases (including exisitng climate, soils, crops data, and satellite imagery) could be used to develop systems that can meet a variety of needs for ecosystem-based management ans policy decisions for agricultural land of the US.
Results of approach (i) indicate that, with regard to levels of clustering, 11 and 21 clusters are too few to adequately provide boundaries for agroecozones in the continental US while 53 and 61 clusters are too many. Selection of a cluster level of 30 to 40 appears to be about the range of clusters that can adequately describe the agroecozone boundaries. The map with 44 clusters is recommended as providing the most suitable information about the agroecozones to policy makers for this type of analysis.
Approach (ii) indicated that the use of both environmental and crop classes of variables produced more contiguous regions compared with the zones determnined with cropping data alone and was better able to segregate several important production areas. Work described in this approach should be regarded as highly preliminary with the main outcome being the development and testing of tools and databases to search for ecologically meaningful patterns in the distribution of crops and important environmental driving variables.
In approach (iii), Map C3 provided the greatest separability between classes that describe agroecological zones and represented six variables with 25 classes. Comparisons of these analyses with those that established discrete boundaries indicated the high amount of variability that existed within boundaries that might be drawn around any ecozone. Again, the work described in this approach should also be regarded as highly preliminary.
The technology and knowledge base exists to develop a system for characterizing and analyzing the true "dynamic" nature of agroecosystems at the regional and national levels. This transdisciplinary systems approach would provide more power and flexibility and could be tailored to meeta wide variety of needs. Such an approach is recommended for future efforts of this kind.
It is likely unrealistic to assume that a single system exists for agroecological classification that can meet needs as diverse as, for example, prioritizing the geographic distribution of agricultural research at the national level versus determining the transferability of agricultural practices and technologies to different locations.
The approaches considered by this study are a type of "emerging technology" that should be considered seriously for further development. However, if significant progress is to be made, future efforts need to be conducted as a serious research activity that is supported with adequate funding, expert personnel, and time to allow significant progress to be made.
