Prediction and uncertainty in restoration science

Abstract

Restoration outcomes are notoriously unpredictable and this challenges the capacity to reliably meet goals. To harness ecological restoration's full potential, significant advances to predictive capacity must be made in restoration ecology. We outline a process for predicting restoration outcomes, based on the model of iterative forecasting. We then describe six challenges that impede predictive capabilities in restoration and, for each, an agenda for overcoming the challenge. Key challenges include the lack of clear goals, insufficient knowledge of why restoration outcomes vary, difficulty quantifying known drivers of variation prior to initiation of restoration projects, model uncertainty, the need to scale up local understanding to guide large‐scale restoration efforts, and temporally variable conditions that hinder long‐term forecast accuracy. Meeting these challenges will require research to resolve key drivers of variation in restoration outcomes; however, there is also a critical need to begin forecasting efforts in restoration ecology immediately. Although early efforts may be of limited practical utility, iterating between model development and evaluation will resolve data needs, minimize uncertainty, and lead to predictions that practitioners can confidently embrace. In turn, a robust predictive capacity will help to reliably meet goals, enhance cost‐effectiveness, and guide policy decisions to help see out the promise of the Decade on Ecosystem Restoration.