Emulating the Global Change Analysis Model with Deep Learning

TL;DR

This work presents a high-fidelity, differentiable neural emulator for GCAM, enabling rapid predictions of $22{,}528$ outputs (across $44$ quantities, $32$ regions, and $16$ years) from $12$ inputs, of which $9$ are continuously varied in $[0,1]$ and $3$ remain binary. The emulator, a four-layer neural network with $256$ hidden units per layer, achieves a median $R^2$ of $0.998$ on predictions and $0.812$ on input–output sensitivities, validating both predictive accuracy and alignment with GCAM's sensitivities. By enabling faster runs and differentiable input-space exploration, the approach facilitates scenario discovery and large-ensemble design while reducing computational costs. The work demonstrates how emulator-in-the-loop strategies can guide multi-sector climate–energy analyses, providing a tool for efficient, targeted exploration of GCAM outputs and their drivers.

Abstract

The Global Change Analysis Model (GCAM) simulates complex interactions between the coupled Earth and human systems, providing valuable insights into the co-evolution of land, water, and energy sectors under different future scenarios. Understanding the sensitivities and drivers of this multisectoral system can lead to more robust understanding of the different pathways to particular outcomes. The interactions and complexity of the coupled human-Earth systems make GCAM simulations costly to run at scale - a requirement for large ensemble experiments which explore uncertainty in model parameters and outputs. A differentiable emulator with similar predictive power, but greater efficiency, could provide novel scenario discovery and analysis of GCAM and its outputs, requiring fewer runs of GCAM. As a first use case, we train a neural network on an existing large ensemble that explores a range of GCAM inputs related to different relative contributions of energy production sources, with a focus on wind and solar. We complement this existing ensemble with interpolated input values and a wider selection of outputs, predicting 22,528 GCAM outputs across time, sectors, and regions. We report a median $R^2$ score of 0.998 for the emulator's predictions and an $R^2$ score of 0.812 for its input-output sensitivity.

Figures (2)

• Figure 1: Diagram of the input-output relationship using GCAM. The emulator approximates the dashed box, mapping directly from inputs to outputs
• Figure 2: GCAM (left) vs. Emulator (right) local sensitivities of inputs vs Years (Top), Quantities (Middle), and Regions (Bottom).