Calibration of a distributed land surface model for a semi-arid basin using remotely sensed data.

Abstract

A meso-scale medium-resolution land surface model using the NCEP's NOAH code has been setup over the San Pedro basin in Arizona. The model is driven using the 50-year hydrologically balanced land surface data set developed at the University of Washington (UW), precipitation data from the PERSIANN (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks) system, and solar radiation influx from the University of Maryland's SRB (Surface Radiation Budget). The remotely sensed forcings are introduced in stages to explore the sensitivity on the simulation results to the remote sensing information. The model was calibrated, using the available remote sensing forcings, to ground observations of the turbulent heat fluxes and ground temperatures using multi-criteria calibration techniques. The simulation results for the distributed model using the default (uncalibrated) and the calibrated parameters are intercompared against each other and against the UW outputs. Results show the PERSIANN data substitution to be a major improving factor in the results in capturing the surface runoffs and the soil moisture diurnal variations because of its fmer sub-hourly temporal resolution as against the UW precipitation forcing which is basically at a daily temporal resolution. Separately, the default seasonal vegetation variation in the NOAH, i.e., constant greenness leaf area index (LAI) and variable greenness fraction (GF) was changed to the opposite case of constant GF and varying LAI using LAI data from the MODIS sensor. Visual inspection of the simulation results show the incompatibility between the NOAH and the MODIS LAI as per the MODIS LAI substitution method used in this study.