Accelerating Fine-Scale Forest Modeling
Across the North American Boreal Zone

NASA scientists are using high-performance computing (HPC) resources at the NASA Center for Climate Simulation (NCCS) at Goddard Space Flight Center to test and accelerate forest modeling and achieve a better understanding of forest response to climate change and wildfires.

Project Overview

Using the Advanced Data Analytics Platform (ADAPT) and ADAPT’s Arctic Boreal Vulnerability Experiment (ABoVE) Science Cloud, this research focuses on improving simulations of forest dynamics and characteristics within the North American boreal zone, a region where climate change is occurring faster than anywhere else on Earth.

Scientific Modeling

High-resolution, ecological modeling helps scientists simulate ecosystem dynamics by considering important vegetation drivers at a fine scale.

An individual-tree based forest model—the University of Virginia Forest Model Enhanced (UVAFME)—was updated for improved simulation of forest dynamics and biotic-abiotic interactions within the boreal region. Model updates were tested against field observations of climate and soil characteristics (e.g., soil moisture, evapotranspiration) as well as forest characteristics (e.g., biomass, species composition) at hundreds of sites across Alaska. Results showed good agreement between model output and field observations.

The updated model was then applied across the Tanana River Basin within interior Alaska (~115,000 km2 area) on a grid of 2 km2, equaling about 28,500 sites. The model was used to simulate forest dynamics under two climate change scenarios: Representative Concentration Pathway (RCP) 4.5, a more conservative prediction for climate change; and RCP 8.5, a more extreme prediction. Under both scenarios, UVAFME predicted overall biomass decline within the Tanana River Basin, with some areas of increasing biomass in historically very cold, moist areas. UVAFME also predicted increasing deciduous forest cover across the region.

HPC Resources Used

The ABoVE Science Cloud was used to conduct code refinements and code testing, as well as to run model simulations. The model was tested on four nodes on the cloud, running jobs interactively. A large simulation was run across the Tanana River Basin using a Slurm script to submit batches of jobs to 10 nodes. Each job consisted of about 2,850 sites to simulate, using only one core and minimal RAM. If run in sequence on a personal computer, running all of the Tanana would take 317 hours. When distributed across 10 nodes, the runs took only 32 hours, or 1/10 of the time. Thus, the ABoVE Science Cloud was an invaluable resource for accelerating the testing and production of this forest model.

Scientific Approach

UVAFME simulates each individual tree’s establishment, growth, and death across a forested landscape. Trees grow and respond to environmental conditions (e.g. temperature, soil moisture, etc.) and compete with other trees for resources (e.g. light). Trees can also die through stress or disturbances such as wildfires. Trees also interact with their abiotic environment by influencing litter decay, soil organic layer depth, and litter flammability. By explicitly considering the important vegetation-soils-fire-climate interactions at their inherent scale (i.e., individual trees), UVAFME can better represent current and future forest environmental conditions.

Project Scientist Adrianna Foster flying over Denali National Forest, on the way to the Tanana River Valley study area. Photo courtesy A. Foster.

Results and Impact

This finding demonstrates the importance of a fine-scale model like UVAFME in simulating future forest state. “In the past, applying a high-resolution model like UVAFME across such a large region would have been prohibitively costly in terms of computing time,” observed project scientist Adrianna Foster. “However, with advances in computing power and with the availability of HPC systems like those of the NCCS, detailed and highly interactive models can be applied across whole region and across continents.”

More Information:

University of Virginia Forest Model Enhanced

Foster, Adrianna C., Amanda H. Armstrong, Jacquelyn K. Shuman, Herman H. Shugart, Brendan M. Rogers, Michelle C. Mack, Scott J. Goetz, K. Jon Ranson, 2019: Importance of Tree- and Species-Level Interactions With Wildfire, Climate, and Soils in Interior Alaska: Implications for Forest Change Under a Warming Climate. Ecological Modelling, 409, 108765, doi:10.1016/j.ecolmodel.2019.108765.

Sean Keefe, NASA Goddard Space Flight Center