A comprehensive strategy for modeling watershed restoration priority areas under epistemic uncertainty: A case study in the Atlantic Forest, Brazil

Abstract

Hydrological models are useful to support ecosystem services restoration. However, when planning the expansion of Nature-based Solutions and Payments for Ecosystem Services, there are requirements including the principle of additionality, representation of multiple runoff mechanisms and calculation of farm-scale spatial outputs and epistemic uncertainty. While addressing these aspects is important to make future planning more effective (investment costs and water benefits) it is also challenging. This paper integrates, in a single modeling framework, the main requirements highlighted. This allows an improved selection of top priority areas with farm-scale spatial resolution and a deeper understanding of how epistemic modeling uncertainty affects the results, which is relevant to evaluate the risks of overestimating ecosystem services benefits. The modeling approach is demonstrated in a case study watershed in the Atlantic Forest biome (Brazil). The PLANS model is introduced, which uses the design of TOPMODEL to simulate both simulates both saturation-excess and infiltration-excess runoff at the farm-scale resolution. Also, it uses a novel saturation index based in a composition of Height Above the Nearest Drainage and Topographical Wetness Index terrain descriptors. We applied the Generalized Likelihood Uncertainty Estimation method, aided by an evolutionary algorithm, to estimate output epistemic uncertainty. Results show that uncertainty substantially affects priority definition (97% ranking change), and that simulated topographic effects can overcome local effects of land cover and soil type. With better data, uncertainty evaluation indicates that the restoration program in the study case could have the cost of its allocation policy reduced by up to 27%, showing that it can be made more cost- effective with the methods proposed.