Parameter conditioning with a noisy Monte Carlo genetic algorithm (NMCGA) to estimate effective soil hydraulic properties from space

Effective soil hydraulic parameters and their uncertainties are of paramount importance in large scale hydro- climatic applications. We applied the concept of parameter conditioning of soil hydraulic parameter distributions to estimate these effective parameters and their uncertainties at the remote sensing footprint such that the time series of the mean simulated soil moisture (from Monte Carlo simulations) approximates the satellite-based (AMSR-E) remotely sensed soil moisture data. Here, we developed a noisy Monte Carlo genetic algorithm (NMCGA) and linked it with a soil-atmosphere-plant-model SWAP to perform the parameter conditioning. We present the results of our NMCGA applications using numerical experiments (with synthetic datasets) and regional remote sensing experiments. Numerical experiments include parameter conditioning with: (i) unique soil signatures (sandy loam, silt loam, clay loam) and (ii) mixed soil signatures in a remote sensing pixel under irrigated and rainfed conditions. For the regional remote sensing experiments, we used the SMEX05 region, Iowa, USA using AMSR-E soil moisture data, MODIS LAI, and TRMM precipitation as conditioning data. The parameter conditioning include using: (i) restricted parameter spaces based on soil hydraulic parameter values derived from laboratory measurements, (ii) relaxed parameter spaces based on the min-max values of the soil hydraulic parameters, (iii) relaxed parameter spaces with restrictions on thetares/thetasat ratio, (iv) relaxed parameter space with restrictions on thetares/thetasat ratio, and minimum values of parameters alpha and n, and (v) by varying penalty coefficients in NMCGA. We envisaged that the soil parameter estimation and conditioning method presented would be very useful in exploring the full potentials of AMSR-E soil moisture data for hydro-climatic modeling applications in the future.