Moisture availability and ecological restoration limit fine fuels and modelled wildfire intensity following non‐native ungulate removal in Hawaii

Improving the prediction of wildfire susceptibility on Hawai'i Island, Hawai'i, using explainable hybrid machine learning models

This study presents a comparative analysis of four Machine Learning (ML) models used to map wildfire susceptibility on Hawai'i Island, Hawai'i. Extreme Gradient Boosting (XGBoost) combined with three meta-heuristic algorithms - Whale Optimization (WOA), Black Widow Optimization (BWO), and Butterfly Optimization (BOA) - were employed to map areas susceptible to wildfire. To generate a wildfire inventory, 1408 wildfire points were identified within the study area from 2004 to 2022. The four ML models (XGBoost, WOA-XGBoost, BWO-XGBoost, and BOA-XGBoost) were run using 14 wildfire-conditioning factors categorized into four main groups: topographical, meteorological, vegetation, and anthropogenic. Six performance metrics - sensitivity, specificity, positive predictive values, negative predictive values, the Area Under the receiver operating characteristic Curve (AUC), and the average precision (AP) of Precision-Recall Curves (PRCs) - were used to compare the predictive performance of the ML models. The SHapley Additive exPlanations (SHAP) framework was also used to interpret the importance values of the 14 influential variables for the modeling of wildfire on Hawai'i Island using the four models. The results of the wildfire modeling indicated that all four models performed well, with the BWO-XGBoost model exhibiting a slightly higher prediction performance (AUC = 0.9269), followed by WOA-XGBoost (AUC = 0.9253), BOA-XGBoost (AUC = 0.9232), and XGBoost (AUC = 0.9164). SHAP analysis revealed that the distance from a road, annual temperature, and elevation were the most influential factors. The wildfire susceptibility maps generated in this study can be used by local authorities for wildfire management and fire suppression activity.

Shifting Limitations to Restoration across Dryland Ecosystems in Hawaiʻi

Hawaiian dryland ecosystems are important for global biodiversity conservation and contain numerous species threatened with extinction. Over the past century, wildfire frequency and size have increased dramatically because of invasion by fire-promoting non-native invasive species, greatly threatening these ecosystems. Native species restoration is a tool that can disrupt the cycle of increased fire and invasion in lowland dry forest communities, but restoration prescriptions have not been studied systematically in other dryland plant communities. We examined the restoration of three Hawaiian dryland plant communities (a high-productivity Diospyros sandwicensis and Metrosideros polymorpha lowland dry forest (HP), a moderate-productivity Myoporum sandwicense and Sophora chrysophylla dry forest/woodland (MP), and a low-productivity Dodonaea viscosa shrubland (LP)), using a community-assembly framework to understand the abiotic and biotic constraints to species establishment and growth in each community. Because active restoration methods are often needed, at both high and low levels of productivity, we also examined restoration treatments and outcomes across the three sites, which spanned a gradient of rainfall and substrate age. At each site, we used the same factorial field experiment with three factors: habitat quality (high or low), weed control (yes or no), and species addition (none, seeding, or outplanting). Outplants (cohort 1) and seeds were added in the winter of 2009–2010, and outplants were added again in March 2011 (cohort 2). Dispersal limitation was apparent at the LP and HP sites, but was not observed in the MP site, which had, overall, greater native diversity and abundance. Outplant survival was greater in high-quality habitats at the HP site, likely due to reduced abiotic stress. Invasive species were found in greater abundance in certain types of microsites at the LP and MP sites, suggesting that shade or topography can be used to plan restoration and weed-control activities. Overall, active restoration methods improved restoration outcomes at the high- and low-productivity sites, and less so at the moderately productive site. Weed removal and outplanting were effective restoration prescriptions at the LP and HP sites, and habitat quality could also be used to increase survival at the HP site. Active restoration could be a lower priority for moderately invaded, moderate-productivity communities, which have the capability to maintain a native ecosystem state.