Water and Energy Exchanges in Semi-Arid Environments using Land-Surface Models: Results from the PILPS San Pedro Experiment
Location
Eccles Conference Center
Event Website
http://water.usu.edu/
Start Date
3-28-2006 10:00 AM
End Date
3-28-2006 10:20 AM
Description
Predicting the availability of water resources in hydrologically stressed regions, like the U.S. Southwest, depends fundamentally on the ability to understand and adequately reproduce the interaction of vegetation processes with climate and its effects on the water cycle. Within the framework of the PILPS Semi-Arid Experiment (PILPS San Pedro), two 4+ year long data sets from the USDA Experimental Watershed in the Walnut Gulch, Arizona, are used for an intercomparison analysis of eight state-of-the-art Land-Surface Models (LSM)1. Off-line runs using the provided forcing and default parameters (a priori established model parameters for semi-arid regions) that included several energy and water balance model outputs and state variables allow to perform split sample validation tests using goodness-of-fit measures (R2, IA, NSE, RMSE, Bias), similarity measures (Hausdorff Norm), statistical correlation (Taylor diagrams), etc. The considered time scales for the analysis were: monthly, daily average, annual, interannual. Because the data sites (Lucky Hills and Kendall) correspond to two different semi-arid environments (shrub and grass, respectively) we attempt to diagnose the extent to which the standard representation of the participant parameterization schemes is adequate. We focus on evaluating both, the consistency of net energy partition into latent, sensible and ground heat flux components as well as closure and decomposition of the water balance. Interannual closure check showed that all of the schemes have energy balance residuals less than ± 2 wm-2. The mean annual ground heat flux is zero for three models (C,H,I). Three models (A,E,D) lose ground heat flux and two (B,G) have positive mean annual ground heat fluxes. Models fail to reproduce the annual net radiation. Large differences in the surface energy partitioning between latent and sensible heat fluxes have been observed among models and between sites. The Bowen ratio for Lucky Hills site (shrub) is 4.86 opposed to 2.23 for the Kendall site (grass). The among-model range of Bowen ratios for the grassland site varies from 1.38 to 5.30. In the shrub site it ranges between 2.65 and 6.27. Four models present the same partitioning independently of the site. Those differences suggest the possibility that different schemes predict different energy balance climatologies and that depending on the model probably different physical mechanism may be controlling the exchanges at the surface. In Lucky Hills most of the models (except H,B) represent well the annual latent heat flux but underestimate on the sensible by more than 15%. In Kendall, only one model (D) represents the mean annual latent heat adequately and the rest of them underestimate it. Three (B, E, G) overestimate on the sensible heat and the underestimation of the others is not larger than 20%. The monthly diurnal cycle of net radiation and sensible heat is better modeled than the cycle of latent heat or ground heat. Interannual water balance check showed that residuals of only three of the schemes (G,H,I) are less than ± 3 mm. All others fail to fulfill the criteria for closure. The partition of available water into runoff and evaporation reveal that models fail to plot along the total annual precipitation line. Models present a great variability in the individual terms of the 1 A: BATS2, B: CBM, C:ISBA, D:NOAH, E:SEWAB, G:SPONSOR, H:SSIB, I:SWAP water balance, being estimates of runoff and soil moisture the most different. Comparison between model predicted runoff events and observations from stream gauges nearby the flux towers have been performed. In general, models present runoff when it has been recorded at the gauges and the majority of them also when no event has been observed. However, the differences in runoff quantity for a given event are very large among the models. The soil moisture estimates show very different behaviors and differences here are of orders of magnitude. These results point out serious shortcomings in the ability of LSM to reproduce water and energy exchanges in different semi-arid environments. Not only have we concluded that the parameterization need revision but that semi-arid regions constitute a complex environment within which an easy transferability of characteristics is not appropriate.
Water and Energy Exchanges in Semi-Arid Environments using Land-Surface Models: Results from the PILPS San Pedro Experiment
Eccles Conference Center
Predicting the availability of water resources in hydrologically stressed regions, like the U.S. Southwest, depends fundamentally on the ability to understand and adequately reproduce the interaction of vegetation processes with climate and its effects on the water cycle. Within the framework of the PILPS Semi-Arid Experiment (PILPS San Pedro), two 4+ year long data sets from the USDA Experimental Watershed in the Walnut Gulch, Arizona, are used for an intercomparison analysis of eight state-of-the-art Land-Surface Models (LSM)1. Off-line runs using the provided forcing and default parameters (a priori established model parameters for semi-arid regions) that included several energy and water balance model outputs and state variables allow to perform split sample validation tests using goodness-of-fit measures (R2, IA, NSE, RMSE, Bias), similarity measures (Hausdorff Norm), statistical correlation (Taylor diagrams), etc. The considered time scales for the analysis were: monthly, daily average, annual, interannual. Because the data sites (Lucky Hills and Kendall) correspond to two different semi-arid environments (shrub and grass, respectively) we attempt to diagnose the extent to which the standard representation of the participant parameterization schemes is adequate. We focus on evaluating both, the consistency of net energy partition into latent, sensible and ground heat flux components as well as closure and decomposition of the water balance. Interannual closure check showed that all of the schemes have energy balance residuals less than ± 2 wm-2. The mean annual ground heat flux is zero for three models (C,H,I). Three models (A,E,D) lose ground heat flux and two (B,G) have positive mean annual ground heat fluxes. Models fail to reproduce the annual net radiation. Large differences in the surface energy partitioning between latent and sensible heat fluxes have been observed among models and between sites. The Bowen ratio for Lucky Hills site (shrub) is 4.86 opposed to 2.23 for the Kendall site (grass). The among-model range of Bowen ratios for the grassland site varies from 1.38 to 5.30. In the shrub site it ranges between 2.65 and 6.27. Four models present the same partitioning independently of the site. Those differences suggest the possibility that different schemes predict different energy balance climatologies and that depending on the model probably different physical mechanism may be controlling the exchanges at the surface. In Lucky Hills most of the models (except H,B) represent well the annual latent heat flux but underestimate on the sensible by more than 15%. In Kendall, only one model (D) represents the mean annual latent heat adequately and the rest of them underestimate it. Three (B, E, G) overestimate on the sensible heat and the underestimation of the others is not larger than 20%. The monthly diurnal cycle of net radiation and sensible heat is better modeled than the cycle of latent heat or ground heat. Interannual water balance check showed that residuals of only three of the schemes (G,H,I) are less than ± 3 mm. All others fail to fulfill the criteria for closure. The partition of available water into runoff and evaporation reveal that models fail to plot along the total annual precipitation line. Models present a great variability in the individual terms of the 1 A: BATS2, B: CBM, C:ISBA, D:NOAH, E:SEWAB, G:SPONSOR, H:SSIB, I:SWAP water balance, being estimates of runoff and soil moisture the most different. Comparison between model predicted runoff events and observations from stream gauges nearby the flux towers have been performed. In general, models present runoff when it has been recorded at the gauges and the majority of them also when no event has been observed. However, the differences in runoff quantity for a given event are very large among the models. The soil moisture estimates show very different behaviors and differences here are of orders of magnitude. These results point out serious shortcomings in the ability of LSM to reproduce water and energy exchanges in different semi-arid environments. Not only have we concluded that the parameterization need revision but that semi-arid regions constitute a complex environment within which an easy transferability of characteristics is not appropriate.
https://digitalcommons.usu.edu/runoff/2006/AllAbstracts/31