Quantifying model uncertainty to improve watershed-level ecosystem service quantification: a global sensitivity analysis of the RUSLE uri icon

abstract

  • Ecosystem service-support tools are commonly used to guide natural resource management. Often, empirically based models are preferred due to low data requirements, simplicity and clarity. Yet, uncertainty produced by local context or parameter estimation remains poorly quantified and documented. We assessed model uncertainty of the Revised Universal Soil Loss Equation ? RUSLE developed mainly from US data. RUSLE is the most commonly applied model to assess watershed-level soil loss. We performed a global sensitivity analysis (GSA) on RUSLE with four dissimilar datasets to understand uncertainty and to provide recommendations for data collection and model parameterization. The datasets cover varying spatial levels (plot, watershed and continental) and environmental conditions (temperate and tropical). We found cover management and topography create the most uncertainty regardless of environmental conditions or data parameterization techniques. The importance of other RUSLE factors varies across contexts. We argue that model uncertainty could be reduced through better parameterization of cover management and topography factors while avoiding severe soil losses by targeting soil conservation practices in areas where both factors interact and enhance soil loss. We recommend incorporating GSA to assess empirical models? uncertainty, to guide model parameterization and to target soil conservation efforts

publication date

  • 2017
  • 2017