In 1981 the U.S. Geological Survey established an experimental project to assess the possible and practical use of remote-sensing data to estimate evapotranspiration as an approximation of consumptive use of water in the lower Colorado River flood plain. The project area was in Parker Valley, Arizona. The approach selected was to measure the areas covered by each type of vegetation, using remote-sensing data in various types of analyses, and to multiply each area by a predetermined water-use rate.
Two calibration and six remote-sensing methods of classifying crop types were compared for cost, accuracy, consistency, and labor requirements. Included were one method each for field reconnaissance using 1982 data, low-altitude (less than 5,000 feet) aerial photography using 1982 data, and visual photointerpretation of Landsat satellite images using 1981 and 1982 data; two methods for medium-altitude (15,000-18,000 feet) aerial photography using 1982 data; and three methods for digital Landsat satellite images using 1981 data. A test of the most promising digital-processing method, which used three image dates, was made in part of Palo Verde Valley, California, where 1981 crop data were more complete than in Parker Valley.
Of the eight methods studied, the two-date digital-processing method was the most consistent and least labor intensive for identifying two or three major crops; visual photointerpretation of Landsat images was the least expensive. Evapotranspiration estimates from crop classifications by all methods differed by a maximum of 6 percent. Total evapotranspiration calculated from crop data and phreatophyte maps in 1981 ranged from 11 percent lower in Palo Verde Valley to 17 percent lower in Parker Valley than consumptive use calculated by water budgets. The difference was greater in Parker Valley because the winter crop data were not included.