Addressing Issues of Model Scale with New E3SM Land Model Parameterizations and a Novel Ecosystem Mapping and Modeling Approach
Authors
Eve Gasarch1* (egasarch@lanl.gov), Claire Bachand1* (cbachand@lanl.gov), Rich Fiorella1, Ryan Crumley1, Katrina Bennett1, Colleen Iversen2
Institutions
1Los Alamos National Laboratory, Los Alamos, NM; 2Oak Ridge National Laboratory, Oak Ridge, TN
URLs
Abstract
Scaling is a ubiquitous challenge in Earth system models (ESMs). In the Arctic, the spatial distribution of snow depends on fine-scale processes and has impacts on underlying permafrost, regional hydrology, and ecology. Incorporating these fine-scale relationships into coarse-scale ESMs is difficult due to the challenges of accessing and observing the vast Arctic system. In this work, the project aims to improve representation of snow in the Arctic in the E3SM Land Model (ELM) through improved process representation and novel snow-scaling techniques. To improve ELM, the team leveraged and modified its existing subgrid structure and snow parameterizations. Results show that downscaling radiation and precipitation forcings based on topographic characteristics improves E3SM’s snow estimates. Researchers also find that modeled vegetation heights are too tall in the Arctic. Taken together, combined with ELM’s lack of wind redistribution of snow, the team finds that ELM exhibits biases in its representation of fine-scale snow processes. To complement these ELM developments, researchers propose a novel clustering technique to upscale snow characteristics from watershed to regional scales using remotely sensed landscape characteristics. This construct, which is referred to as ecosystem mapping and modeling units (EMMUs), is a promising way to infer snow distributions across large domains and to evaluate ESMs at the grid cell level (e.g., >1 km2).
EMMUs are derived using a k-means clustering method on features derived from airborne light detection and ranging (LiDAR) data collected on Alaska’s Seward Peninsula in summer 2021 and winter 2022. The team examined the relationships between snow depth and LiDAR-derived landscape characteristics to inform feature selection. The results presented include 11 clusters based on topographic position index, shrub height, and shrub density. Average snow depth differs significantly between each cluster, demonstrating that the EMMU approach could be valuable for informing and evaluating ELM developments at larger scales.