Submitted to: Meeting Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 5/1/2009
Publication Date: 9/5/2009
Citation: French, A.N., Hunsaker, D.J., Thorp, K.R., Clarke, T.R. 2009. Spatially distributed evapotranspiration estimation using remote sensing and ground-based radiometers over cotton at Maricopa, Arizona. In: Proceedings of New Approaches to Hydrological Prediction in Data-sparse Regions, September 6-12, 2009, Hyderabad, India. p. 267-272. Interpretive Summary: The increasing sophistication of land surface hydrological models based on remote sensing will make spatial mapping of evapotranspiration (ET) operationally feasible in the near future. This capability could have a major impact upon water management and conservation in agricultural regions where water supplies are scarce. However, these capabilities are not matched by current, nor near-future, remote sensing systems. In most instances farm-scale ET could be mapped with satellite data no more frequently than 16 days, an interval at least twice as long as needed for operational use. To cope with this shortcoming an ET modeling approach has been developed that fills the gap between remote sensing images. The method combines existing visible, near infrared, and thermal infrared data together with ground-based plant temperature observations to predict daily ET. Using experimental data collected over cotton in 2003 at Maricopa, Arizona, the method was tested and found potentially accurate to within 20% over a two-week interval. This research will benefit growers in irrigated lands and regional water managers.
Technical Abstract: Spatially distributed estimates of evapotranspiration (ET) over agricultural lands could be valuable for water management in arid environments and for monitoring irrigated croplands. In recent year various ET estimation approaches have been developed that utilize remote sense data to provide the needed spatially distributed information. Commonly these include visible and near infrared observations for vegetation indices and thermal infrared observations for surface temperatures. However, these approaches have a major constraint: high-resolution remote sensing data (<100 m) are typically unavailable at sufficiently frequent intervals t allow the ET estimates to e used operationally. For example, Landsat TM data have 16-day repeat interval, but where cloudy skies prevail the interval could readily exceed a month. Clearly ET estimated based on the infrequent observations alone would have very limited practical value for crop water management. To help reduce the impact of this constraint, and ET estimation approach is developed that combines surface temperature data from ground-based radiometers with the most recently available remote sensing images. The approach projects vegetation cover from the latest NDVI scenes using a simplified crop model while spatially distributed surface temperatures are estimated from these cover projections and continuously recording ground-based radiometers. The resulting image pairs are then input to a two-source energy balance model, which estimates ET at daily time steps. The approach is demonstrated using remotely sensed data over cotton experiment conducted in 2003 at Maricopa, Arizona. Using soil moisture depletion observations for ET validation, the ET modeling approach is shown to be accurate within 1 mm/day for projections up to two weeks beyond the latest remote sensing acquisition and thus, could be a useful way to cope with infrequent remote sensing image data.