A Modeling Framework for Assessing the Value of Learning in Dynamic Adaptive Planning: Application to Green Infrastructure Investment Evaluation

Abstract

AbstractTo cope with the uncertainty of green infrastructure planning, many cities take an adaptive approach and use learning‐by‐doing to improve estimates of the cost and efficacy of stormwater management practices (SMPs) and use that information to improve stormwater plans. However, deciding whether that learning is worth its expense has been a challenge for practitioners. We propose a modeling framework to assess the economic value of learning. Methodologically, we present a generalized adaptive planning method that includes learning from direct and indirect investments and multiple degrees of learning. The formulation enables users to specify possible knowledge gains from near‐term actions and quantify its value by assessing its impacts on subsequent decisions and their performance. Further, we quantify the values of both learning and adaptability by calculating differences in expected system performance between three types of decision making: non‐adaptive (no learning between decisions), passive adaptive (adaptive planning that passively accepts incidental learning), and active adaptive (adaptive planning that considers potential learning opportunities when choosing investments). For illustration, we apply the framework to an example inspired by a real stormwater management setting in Philadelphia, PA. In the example, the ability of SMPs to reduce runoff is evaluated by hydrological simulation. The literature and expert opinions inform estimates of costs, SMP performance deterioration over time, and predictions of possible knowledge gains. The results show that active adaptive planning supported by stochastic optimization can achieve substantial cost savings.