LINKING WITHIN-FIELD CROP RESPONSE WITH SOIL CHARACTERISTICS TO DEFINE CROPRESPONSE ZONES FOR MANAGEMENT ZONE DELINEATION

Authors: Walthall, Charles; Kaul, Monisha; Timlin, Dennis; Daughtry, Craig

Submitted to: Geospatial Information in Agriculture and Forestry International Conference
Publication Type: Proceedings
Publication Acceptance Date: 8/27/2001
Publication Date: N/A
Citation: N/A

Interpretive Summary: Zones defining similar response of a crop to environmental factors are needed to define management zones for precision farming. This can be accomplished by relating an observable plant characteristic such as foliage density expressed as leaf area index (LAI) to the ability of soil to hold water expressed as soil water holding capacity (SWHC). Given a model that predicts LAI as a function of SWHC (and other lesser factors), maps of SWH may be possible by using LAI inputs to a model inversion. A traditional multiple linear regression, a neural network and a plant growth simulation model were tested for their ability to predict LAI from inputs that included SWHC, using two years of data from a corn field in Beltsville, Maryland. Results in order of best performance were the neural network, the multiple linear regression and the crop simulation model. All three methods appeared to capture the general appearance of the relationship between LAI and SWHC observed with actual data. These results showed that such a model is possible given further research using additional years of data. These findings will be of special interest to anyone trying to link vegetative LAI derived from remotely sensed data to soil characteristics.

Technical Abstract: Zones defining consistent crop response to environmental factors are needed to define management zones for precision farming. This is possible by relating an observable plant characteristic such as foliage density (as leaf area index or LAI) to the ability of soil to hold water (as soil water holding capacity or SWHC). LAI can be linked to SWHC via inversion of a model relating SWHC to LAI. Prediction of LAI using multiple linear regression, a neural network and a simple mechanistic plant growth simulation model were tested using two years of data that included SWHC from a corn field in Beltsville, Maryland. Results in order of best performance were the neural network, the multiple linear regression and the crop simulation model. All methods appeared to capture the general non-linear relationship between LAI and SWHC observed with actual data. Further research using additional years of data is needed to further explore these models for LAI prediction and the level of accuracy needed for operational use. These findings will be especially useful to anyone trying to link LAI derived from remote sensing data to soil characteristics.