• Mississippi State University Extension Service
• USDA National Agricultural Statistics Service

• Scott A. Samson
• Thomas L. Gregory

The intent of this project is to evaluate the capability of three commercial satellite remote sensing platforms to provide quality spectral and temporal data needed to identify, delineate and assess crop acreage estimations on fields managed by private farm operators. The USDA National Agricultural Statistics Service employs these data to estimate large area crop acreage estimates.

• evaluate the capability of commercial satellite remote sensing systems to deliver image data on a time-needed basis for crop classification without detrimental impact associated with cloud cover;
• capability of commercial sensors to spectrally discriminate, by end of the growing season, between major row crops raised in Mississippi;
• apply objectives #1 and #2 to fields managed by private farm operators rather than controlled test plots;
• assess impact, at varying pixel resolutions, of mixed pixels on the definition of field boundaries;
• assess impact and compensate for within field anomalies in order to accurately assess acreage enumeration at field level;
• compare results of commercial satellite remote sensing systems as inputs into the NASS system for crop acreage estimation; and
• develop prototype of crop calendar for the Mississippi Delta area.

In collaboration with the National Agricultural Statistics Service office in Jackson, Mississippi, 10 segments will be selected that contain the 3 predominant row crops found in Mississippi: corn, cotton and soybeans. These crops account for approximately 77% of the total harvested cropland and over 82% of the value of harvested crops. The selected sites will represent dominant terrain and climate zones within the state. The segments will be divided into 2 subsets of equal size and crop-type variety. One subset will be used to develop the classification model. The second subset will be used to validate the effectiveness of the model since these segments were not used in the development of the classification model.

IKONOS data will be spatially integrated at resolutions between 4-meters and 30-meters in order to evaluate the impact of changing spatial resolution on crop type identification, classification, field boundary delineation and acreage enumeration. This process will assist in future sensor platform design for crop inventory applications. One segment, or a cluster of segments if they are found within a 50-km x 50-km area, will be used to evaluate the effectiveness of radar image data to discriminate between corn, cotton and soybeans, delineate field boundaries and estimate crop acreage. RADARSAT fine beam C-band image data will be acquired for this task.

Several approaches to classification model development will be evaluated for their effectiveness in spectral discrimination between corn, cotton and soybeans. Because 'a priori' information is available about the crop rotation trends and local cropping patterns from the development of the NASS segments, a synthesis of 'data-driven' and 'knowledge-driven' models will be examined. In the context of the intended market for this model (i.e., the NASS), the model will incorporate only ancillary data that are nationally available to ensure transportability to other areas throughout the United States. In addition, the user processes will be carefully documented and structured in such a way that future users will not require technical knowledge of the model in order to successfully implement the model.

A second year of data compilation and analysis is requested in order to evaluate methodology developed during the first year and to accommodate any potential impact associated with crop damage attributed to natural events (e.g., drought, flood) or sensor malfunctions on the Landsat, IKONOS or RADARASAT platforms. During this time, the crop calendar developed for the Mississippi Delta area will be reviewed and refined, if needed. The second year will also offer an opportunity to limit 'a priori' information provided to the classification model to determine the amount of field-level information needed for acceptable crop identification and delineation.