Two decades of Landsat images reveal spatial and temporal dynamics of invasion and improve species distribution models

Abstract

Developing accurate and cost‐effective methods to detect and predict the spread of invasive species remains an ongoing challenge. Species distribution models (SDMs) are widely used to predict invasion under current and future climates and inform management strategies. However, occurrence datasets are often spatially and temporally biased due to unstructured and opportunistic data collection. Remote sensing of individual species over broad spatial and temporal scales has become increasingly feasible with the accumulation of satellite images and the development of convolutional neural networks. In this study, we used 21 years of archival Landsat satellite imagery (2000–2020) to train a land use/land cover (LULC) classification algorithm using a temporal convolutional neural network to predict occurrence of the invasive species, leafy spurge ( Euphorbia virgata ), across Minnesota. We validated our predictions with an independent dataset. First, we show that leafy spurge has expanded from 1067 km ² in 2000–2002 to 7156 km ² in 2018–2020, a 570% increase. Drought severity modulated the predicted area invaded and was associated with fluctuations over time. Second, we tracked changes in probability over time for individual pixels and showed that invasion has been concentrated in two largely disjunct regions that differ in climate. Third, our remotely sensed predicted occurrence dataset and community science dataset were closer to roadsides, suggesting that roadside disturbance facilitates invasion. Last, we showed that SDMs built using remotely sensed occurrences had higher discrimination, were less overfit, and performed better outside of urban areas. Overall, 21 years of archival satellite imagery showed rapid expansion in two geographically distinct regions, invasion associated with roads, and improved predictions of invasion.