|Talebizadeh, Mansour - Orise Fellow|
|Brauer, David - Dave|
|Starks, Patrick - Pat|
Submitted to: ASA-CSSA-SSSA Annual Meeting Abstracts
Publication Type: Abstract Only
Publication Acceptance Date: 7/11/2017
Publication Date: 10/25/2017
Citation: Talebizadeh, M., Moriasi, D.N., Gowda, P.H., Marek, G.W., Steiner, J.L., Brauer, D.K., Tadesse, H.K., Nelson, A.M., Starks, P.J. 2017. Implication of varying parameter sensitivities and temporal aggregation on estimating evapotranspiration in the semi-arid conditions [abstract]. ASA-CSSA-SSSA Annual Meeting Abstracts. Available at: https://scisoc.confex.com/crops/2017am/webprogram/Paper106714.html.
Interpretive Summary: Abstract only.
Technical Abstract: Sensitivity analysis is generally used to examine the effects of changes in one or more model parameters on model output. Knowledge gained from this analysis greatly assists in simplifying model parameterization through calibration. One of the problems with this approach is that the sensitivity of input parameters may vary when the model is used to simulate environmental impacts under different agricultural systems (e.g. different climate, land use and management, etc.). In addition, both parameter sensitivities and their calibrated values may vary under different temporal scales at which model outputs are calculated. In this study, a dynamic sensitivity analysis approach with different Monte Carlo simulation settings was applied to identify sensitive parameters as well as their contribution to the model performance in estimating evapotranspiration (ET) using the Agricultural Policy/Environmental eXtender (APEX) model. Results indicated that the two most sensitive parameters (“Soil evaporation – plant cover factor” and “Root growth-soil strength”, responsible for reducing soil evaporation due to leaf area index (LAI) development and root growth inhibition due to unfavorable soil characteristics) were found responsible for the largest error reduction (71% reduction in estimating daily ET). Use of more than four calibration parameters did not improve model performance significantly. Meanwhile, the posterior parameter distribution showed that the time aggregation level could affect the posterior parameters distribution. However, for less sensitive parameters or the parameters that become sensitive in certain conditions during the simulation period, the posterior distributions tend be closer to the prior distribution. The proposed methodology has the advantage of capturing those input parameters that are sensitive under certain conditions which may not be detected with static sensitivity analysis. In addition, it provides an objective way of translating the concept of parameter sensitivity into model performance improvement.