Using explainable machine learning methods to evaluate vulnerability and restoration potential of ecosystem state transitions

Ecosystem state transitions can be ecologically devastating or be a restoration success. State transitions are common within aquatic systems worldwide, especially considering human-mediated changes to land use and water use. We created a transferable conceptual framework to enable multiscale assessments of state resilience and early warnings of state transitions that can inform strategic restorations and avoid ecosystem collapse. The conceptual framework integrated machine learning predictions with ecosystem state concepts (e.g., state classification, gradients of vulnerability, and recovery potential leading to state transitions) and was devised to investigate possible environmental drivers. As an application of the framework, we generated prediction probabilities of submersed aquatic vegetation (SAV) presence at nearly 10,000 sites in the Upper Mississippi River (United States). Then, we used an interpretability method to explain model predictions to gain insights into possible environmental drivers and thresholds or linear responses of SAV presence and absence. Model accuracy was 89% without spatial bias. Average water depth, suspended solids, substrate, and distance to nearest SAV were the best predictors and likely environmental drivers of SAV habitat suitability. These environmental drivers exhibited nonlinear, threshold-type responses for SAV. All the results are also presented in an online dashboard to explore results at many spatial scales. The habitat suitability model outputs and prediction explanations from many spatial scales (4 m to 400 km of river reach) can inform research and restoration planning.

Unraveling a century of global change impacts on winter bird distributions in the eastern United States

One of the most pressing questions in ecology and conservation centers on disentangling the relative impacts of concurrent global change drivers, climate and land-use/land-cover (LULC), on biodiversity. Yet studies that evaluate the effects of both drivers on species' winter distributions remain scarce, hampering our ability to develop full-annual-cycle conservation strategies. Additionally, understanding how groups of species differentially respond to climate versus LULC change is vital for efforts to enhance bird community resilience to future environmental change. We analyzed long-term changes in winter occurrence of 89 species across nine bird groups over a 90-year period within the eastern United States using Audubon Christmas Bird Count (CBC) data. We estimated variation in occurrence probability of each group as a function of spatial and temporal variation in winter climate (minimum temperature, cumulative precipitation) and LULC (proportion of group-specific and anthropogenic habitats within CBC circle). We reveal that spatial variation in bird occurrence probability was consistently explained by climate across all nine species groups. Conversely, LULC change explained more than twice the temporal variation (i.e., decadal changes) in bird occurrence probability than climate change on average across groups. This pattern was largely driven by habitat-constrained species (e.g., grassland birds, waterbirds), whereas decadal changes in occurrence probabilities of habitat-unconstrained species (e.g., forest passerines, mixed habitat birds) were equally explained by both climate and LULC changes over the last century. We conclude that climate has generally governed the winter occurrence of avifauna in space and time, while LULC change has played a pivotal role in driving distributional dynamics of species with limited and declining habitat availability. Effective land management will be critical for improving species' resilience to climate change, especially during a season of relative resource scarcity and critical energetic trade-offs.

Spatial Heterogeneity in the Occurrence Probability of Rainstorms over China

Detecting the spatial heterogeneity in the potential occurrence probability of water disasters is a foremost and critical issue for the prevention and mitigation of water disasters. However, it is also a challenging task due to the lack of effective approaches. In the article, the entropy index was employed and those daily rainfall data at 520 stations were used to investigate the occurrences of rainstorms in China. Results indicated that the entropy results were mainly determined by statistical characters (mean value and standard deviation) of rainfall data, and can categorically describe the spatial heterogeneity in the occurrence of rainstorms by considering both their occurrence frequencies and magnitudes. Smaller entropy values mean that rainstorm events with bigger magnitudes were more likely to occur. Moreover, the spatial distribution of entropy values kept a good relationship with the hydroclimate conditions, described by the aridity index. In China, rainstorms are more to likely occur in the Pearl River basin, Southeast River basin, lower-reach of the Yangtze River basin, Huai River basin, and southwest corner of China. In summary, the entropy index can be an effective alternative for quantifying the potential occurrence probability of rainstorms. Four thresholds of entropy value were given to distinguish the occurrence frequency of rainstorms as five levels: very high, high, mid, low and very low, which can be a helpful reference for the study of daily rainstorms in other basins and regions.