Evaluation of a Hybrid Reflectance-Based Crop Coefficient and Energy Balance Evapotranspiration Model for Irrigation Management

Authors

J. Burdette Barker, University of Nebraska-LincolnFollow
Christopher M. U. Neale, University of Nebraska-LincolnFollow
Derek M. Heeren, University of Nebraska-LincolnFollow
Andrew E. Suyker, University of Nebraska - LincolnFollow

Document Type

Article

Date of this Version

4-2018

Citation

Barker, J. B., C. M. U. Neale, D. M. Heeren, and A. E. Suyker. 2018. Evaluation of a hybrid reflectance-based crop coefficient and energy balance evapotranspiration model for irrigation management. Transactions of the ASABE 61(2): 533-548, doi: 10.13031/trans.12311.

doi 10.13031/trans.12311

Comments

© 2018 American Society of Agricultural and Biological Engineers. Used by permission.

Abstract

Accurate generation of spatial soil water maps is useful for many types of irrigation management. A hybrid remote sensing evapotranspiration (ET) model combining reflectance-based basal crop coefficients (K_cbrf) and a two-source energy balance (TSEB) model was modified and validated for use in real-time irrigation management. We modeled spatial ET for maize and soybean fields in eastern Nebraska for the 2011-2013 growing seasons. We used Landsat 5, 7, and 8 imagery as remote sensing inputs. In the TSEB, we used the Priestly-Taylor (PT) approximation for canopy latent heat flux, as in the original model formulations. We also used the Penman-Monteith (PM) approximation for comparison. We compared energy balance fluxes and computed ET with measurements from three eddy covariance systems within the study area. Net radiation was underestimated by the model when data from a local weather station were used as input, with mean bias error (MBE) of -33.8 to -40.9 W m^-2. The measured incident solar radiation appeared to be biased low. The net radiation model performed more satisfactorily when data from the eddy covariance flux towers were input into the model, with MBE of 5.3 to 11.2 W m^-2. We removed bias in the daily energy balance ET using a dimensionless multiplier that ranged from 0.89 to 0.99. The bias-corrected TSEB ET, using weather data from a local weather station and with local ground data in thermal infrared imagery corrections, had MBE = 0.09 mm d^-1 (RMSE = 1.49 mm d^-1) for PM and MBE = 0.04 mm d^-1 (RMSE = 1.18 mm d^-1) for PT. The hybrid model used statistical interpolation to combine the two ET estimates. We computed weighting factors for statistical interpolation to be 0.37 to 0.50 for the PM method and 0.56 to 0.64 for the PT method. Provisions were added to the model, including a real-time crop coefficient methodology, which allowed seasonal crop coefficients to be computed with relatively few remote sensing images. This methodology performed well when compared to basal crop coefficients computed using a full season of input imagery. Water balance ET compared favorably with the eddy covariance data after incorporating the TSEB ET. For a validation dataset, the magnitude of MBE decreased from -0.86 mm d^-1 (RMSE = 1.37 mm d^-1) for the K_cbrfalone to -0.45 mm d^-1 (RMSE = 0.98 mm d^-1) and -0.39 mm d^-1 (RMSE = 0.95 mm d^-1) with incorporation of the TSEB ET using the PM and PT methods, respectively. However, the magnitudes of MBE and RMSE were increased for a running average of daily computations in the full May-October periods. The hybrid model did not necessarily result in improved model performance. However, the water balance model is adaptable for real-time irrigation scheduling and may be combined with forecasted reference ET, although the low temporal frequency of satellite imagery is expected to be a challenge in real-time irrigation management.