Document Type

Article

Publication Date

5-2023

Abstract

Context Addressing ecosystem degradation in the Anthropocene will require ecological restoration across large spatial extents. Identifying areas where natural regeneration will occur without direct resource investment will improve scalability of restoration actions.

Objectives An ecoregion in need of large scale restoration is the Great Basin of the Western US, where increasingly large and frequent wildfires threaten ecosystem integrity and its foundational shrub species. We develop a framework to forecast where postwildfire regeneration of sagebrush cover (Artemisia spp.) is likely to occur within the burnt areas across the region (>900,000 km2).

Methods First, we parameterized population models using Landsat satellite-derived time series of sagebrush cover. Second, we evaluated the out-of-sample performance by predicting natural regeneration in wildfres not used for model training. This model assessment reproduces a management-oriented scenario: making restoration decisions shortly after wildfires with minimal local information. Third, we asked how accounting for increasingly fine-scale spatial heterogeneity could improve model forecasting accuracy.

Results Regional-level models revealed that sagebrush post-fire recovery is slow, estimating > 80-year time horizon to reach an average cover at equilibrium of 16.6% (CI95% 9–25). Accounting for wildfre and within-wildfre spatial heterogeneity improved out-ofsample forecasts, resulting in a mean absolute error of 3.5 ± 4.3% cover, compared to the regional model with an error of 7.2 ± 5.1% cover.

Conclusions We demonstrate that combining population models and non-parametric spatial matching provides a fexible framework for forecasting plant population recovery. Models for population recovery applied to Landsat-derived time series will assist restoration decision-making, including identifying priority targets for restoration.

Copyright Statement

This version of the article has been accepted for publication, after peer review (when applicable) and is subject to Springer Nature’s AM terms of use, but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at: https://doi.org/10.1007/s10980-023-01621-1

Download

Share

COinS