Estimating Invasion Success by Non-Native Trees in a National Park Combining WorldView-2 Very High Resolution Satellite Data and Species Distribution Models

Abstract

Invasion by non-native tree species is an environmental and societal challenge requiring predictive tools to assess invasion dynamics. The frequent scale mismatch between such tools and on-ground conservation is currently limiting invasion management. This study aimed to reduce these scale mismatches, assess the success of non-native tree invasion and determine the environmental factors associated to it. A hierarchical scaling approach combining species distribution models (SDMs) and satellite mapping at very high resolution (VHR) was developed to assess invasion by Acacia dealbata in Peneda-Gerês National Park, the only national park in Portugal. SDMs were first used to predict the climatically suitable areas for A. dealdata and satellite mapping with the random-forests classifier was then applied to WorldView-2 very-high resolution imagery to determine whether A. dealdata had actually colonized the predicted areas (invasion success). Environmental attributes (topographic, disturbance and canopy-related) differing between invaded and non-invaded vegetated areas were then analyzed. The SDM results indicated that most (67%) of the study area was climatically suitable for A. dealbata invasion. The onset of invasion was documented to 1905 and satellite mapping highlighted that 12.6% of study area was colonized. However, this species had only colonized 62.5% of the maximum potential range, although was registered within 55.6% of grid cells that were considerable unsuitable. Across these areas, the specific success rate of invasion was mostly below 40%, indicating that A. dealbata invasion was not dominant and effective management may still be possible. Environmental attributes related to topography (slope), canopy (normalized difference vegetation index (ndvi), land surface albedo) and disturbance (historical burnt area) differed between invaded and non-invaded vegetated area, suggesting that landscape attributes may alter at specific locations with Acacia invasion. Fine-scale spatial-explicit estimation of invasion success combining SDM predictions with VHR invasion mapping allowed the scale mismatch between predictions of invasion dynamics and on-ground conservation decision making for invasion management to be reduced. Locations with greater potential to suppress invasions could also be defined. Uncertainty in the invasion mapping needs to be accounted for in the interpretation of the results.