Page Banner

United States Department of Agriculture

Agricultural Research Service

Research Project: Predicting Crop Water Use in the San Joaquin Valley of California Using Remote Sensing
2012 Annual Report


1a.Objectives (from AD-416):
To develop algorithms relating remotely sensed canopy cover to basal crop coefficients and software for conversion of remote sensing images to crop coefficient maps.


1b.Approach (from AD-416):
Canopy ground cover of various crops from the west side of the San Joaquin Valley will be derived from Landsat and/or other multispectral images collected under acceptably clear weather conditions. Image analyses will be required to process these data. Correlation between the satellite based canopy cover data and basal crop coefficient will be made and algorithms established on crop specific bases. A prototype user interface program will be developed to facilitate the translation of the satellite imagery to canopy cover or crop coefficient maps based on the relationships and algorithms developed in the previous phases of the project. An economic analysis will also be carried out to compare cost and benefit of employing remote sensing for improving crop water use efficiency. Documents Grant with CSU-Seaside.


3.Progress Report:

This project supports objective 1 of the parent project. Satellite imagery can be used to estimate crop growth conditions, and when combined with ground-based measurements, can also provide assessment of crop water status. This project investigated using remote sensing information to improve crop coefficient estimates over current methods and make irrigation scheduling more practical, convenient, and accurate. The objectives were to develop technology to estimate near real-time crop coefficients for fields/crops in the San Joaquin Valley from remotely-sensed data and to develop and demonstrate a prototype decision support system that can efficiently deliver crop coefficient and estimated crop water use information to agricultural producers and water suppliers. Field experiments were carried out from April to October 2008 in a 27 by 35 km area southwest of Fresno, CA, with measurements taken for fractional canopy cover of 20 row, tree, and vine crops from five large farming operations on the same days of satellite overpass. A linear correlation was made between the canopy cover data and normalized difference vegetation index values calculated from the satellite imagery. Conversions of canopy to crop coefficient were summarized for nine different vegetable crops. A generic non-linear relationship was established with reasonable accuracy for predicting crop coefficients from just fractional canopy cover values for these vegetable crops. An analysis was undertaken to identify an optimal relationship between canopy cover and basal crop coefficient that might be used to support broad-area satellite mapping, and to quantify resulting errors. An FAO-56 interpolation method was used to relate the two parameters for several major annual crop classes (vegetables, tubers, legumes, fibers, oils, and cereals) using a “density coefficient” based on canopy cover and crop height. Good agreement was noted between this approach and a second analytical method based on published lysimeter results for lettuce, pepper, garlic, broccoli, melons, tomato, and bean. In addition, temporal profiles of crop coefficients and evapotranspiration were developed for several study fields. Crops included bellpepper, watermelon, cantaloupe, onion, corn, alfalfa, garlic, pistachio, almond, safflower, tomato, cotton, carrot, and grape. Satellite images were atmospherically corrected to surface reflectance via the Landsat Ecosystem Disturbance Adaptive Processing System, and converted to vegetation index on a per-pixel basis. The index values were converted to crop fractional cover based on an equation developed earlier in this project. Then, conversion equations previously developed on weighting lysimeters were used to transform fractional cover to crop coefficients. Spatially contiguous maps of crop coefficients and evapotranspiration were generated for the San Joaquin Valley for selected satellite dates. The satellite-based approach, as implemented in regions with an available evapotranspiration network, potentially enables timely estimation of biological crop water demand for resource monitoring, evaluation of irrigation efficiencies, and scheduling of irrigation events. The satellite-based method was also compared to Surface Energy Balance Algorithm for Land method, which derived evapotranspiration through a surface energy balance approach. Ground-based measurements of fractional ground cover and surface soil moisture were compared with satellite-based vegetation indices and combined with ground-based reference evapotranspiration measurements to calculate actual evapotranspiration estimates. This method was compared to the Surface Energy estimates on a field-by-field basis. Surface Energy method uses satellite radiances and meteorological data to solve earth-surface energy balance, and computes evapotranspiration per-pixel by applying radiative, aerodynamic and energy balance physics. Although the results from the two methods were correlated fairly well, the Surface Energy estimates of evapotranspiration tended to be about 25% lower than those estimated from the satellite method, probably due in part to water stress in perennial crops. Satellite data were merged over time and space to form an 8-day composite to facilitate use of overlapping portions of each scene to increase the frequency of observations and reduce data gaps due to cloud cover. Landsat satellite provides the spatial resolution necessary to produce information at individual-field scale, while the higher temporal resolution of MODIS satellite outputs provide a gap-filling capability to ensure data availability. Gap-filling algorithms were employed to ensure spatially continuous data over agricultural regions. Current algorithms included the use of moving window averages and linear interpolation to fill small data gaps over agricultural fields. Prior to further processing, non-agricultural areas were masked out using the USDA NASS Cropland Data Layer, which was also used to differentiate annual crops from perennials. Each Landsat scene was atmospherically corrected. Vegetation index was calculated from the composited scenes using the red and near infrared wavelengths which provides a measure of photosynthetic capacity. Vegetation index data were then transformed to fraction of ground cover by the empirical relationship developed under this study. Ground cover was then converted to basal crop coefficient based on a physical description of the crop canopy. For perennial crops in the initial prototype release, results from two multi-year studies, one for trees one for vines, recently performed on two large weighing lysimeters at the University of California Kearney Agricultural Center were used. Both studies reported strong positive relationships between mid-day canopy light interception and canopy cover for trees and vines. As an additional step, basal crop evapotranspiration can be estimated as the product of crop coefficient and evapotranspiration, as retrieved from the daily statewide grids generated by California Department of Water Resources. Growers and water managers can then use the daily crop coefficients to make more informed decisions. Project findings were presented by members of the project personnel at the American Geophysical Union fall meeting in San Francisco, the American Society of Agricultural and Biological Engineers meeting in Phoenix, AZ, and the 28th International Horticultural Congress in Portugal.


Last Modified: 10/31/2014
Footer Content Back to Top of Page