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.

Métodos de aprendizaje automático para evaluar la vulnerabilidad y el potencial de restauración de las transiciones de estados de los ecosistemas Resumen Las transiciones de estado de los ecosistemas pueden tener consecuencias ecológicas graves o ser un éxito de restauración. Estas transiciones son comunes en los sistemas acuáticos, especialmente si consideramos los cambios mediados por humanos en el uso del suelo y del agua. Generamos un marco conceptual transferible para permitir las evaluaciones multiescalares de la resiliencia del estado y las alertas tempranas de la transición de estado que sirvan para guiar las restauraciones estratégicas y evitar el colapso del ecosistema. Este marco integró predicciones de aprendizaje automático a los conceptos de estado del ecosistema (p. ej.: clasificación del estado, gradientes de vulnerabilidad y potencial de restauración que lleve a transiciones de estado) y fue diseñado para investigar los posibles factores ambientales. Para aplicar el marco, generamos probabilidades de predicción de la presencia de vegetación acuática sumergida (VAS) en casi 10,000 sitios del Alto Río Mississippi en los Estados Unidos. Después usamos un método de interpretabilidad para explicar las predicciones del modelo y obtener información de los posibles factores ambientales y umbrales o respuestas lineales de la presencia y ausencia de la VAS. La precisión del modelo fue de 89% sin sesgo espacial. Los mejores pronosticadores y probables factores ambientales de la idoneidad de hábitat de la VAS fueron la profundidad promedio del agua, los sólidos suspendidos, el sustrato y la distancia a la VAS más cercana. Estos factores ambientales exhibieron respuestas no lineales de tipo umbral para la VAS. Todos los resultados también están presentados en un tablero virtual para explorar los resultados a varias escalas espaciales. Los resultados del modelo de idoneidad de hábitat y las explicaciones de las predicciones de varias de las escalas espaciales (4 m-400 km de alcance del río) pueden guiar la investigación y la planeación de la restauración.

Exotic-Dominated Grasslands Show Signs of Recovery with Cattle Grazing and Fire.

In grasslands, overgrazing by domestic livestock, fertilization, and introduction of exotic forage species leads to plant communities consisting of a mixture of native and exotic species. These degraded grasslands present a problem for land managers, farmers, and restoration ecologists concerned with improving biodiversity while continuing to use the land for livestock production. Here we assessed the response of butterfly and plant community composition to the use of fire and moderate grazing by domestic cattle on degraded grasslands dominated by exotic plants. We evaluated change by comparing experimental pastures to two reference sites that were grasslands dominated by native plants. We used two burning and grazing treatments: 1) patch-burn graze, a heterogeneously managed treatment, where one third of the pasture is burned each year and cattle have free access to the entire pasture, and 2) graze-and-burn, a homogenously managed treatment, where the entire pasture is grazed each year and burned in its entirety every three years. We tested for change in the butterfly and plant community composition over seven years using Bray-Curtis dissimilarity measures. Over the course of seven years, degraded pastures in both treatments became more similar to reference sites with respect to the butterfly and plant communities. Only two butterfly species and two plant functional guilds exhibited significant linear trends over time, with varying responses. Compositional changes in both the butterfly and plant communities indicate that the use of moderate grazing and fire may shift butterfly and plant communities of exotic-dominated grasslands to be more similar to reference tallgrass prairies over time.

Bee Abundance and Nutritional Status in Relation to Grassland Management Practices in an Agricultural Landscape.

Grasslands provide important resources for pollinators in agricultural landscapes. Managing grasslands with fire and grazing has the potential to benefit plant and pollinator communities, though there is uncertainty about the ideal approach. We examined the relationships among burning and grazing regimes, plant communities, and Bombus species and Apis mellifera L. abundance and nutritional indicators at the Grand River Grasslands in southern Iowa and northern Missouri. Treatment regimes included burn-only, grazed-and-burned, and patch-burn graze (pastures subdivided into three temporally distinct fire patches with free access by cattle). The premise of the experimental design was that patch-burn grazing would increase habitat heterogeneity, thereby providing more diverse and abundant floral resources for pollinators. We predicted that both bee abundance and individual bee nutritional indicators (bee size and lipid content) would be positively correlated with floral resource abundance. There were no significant differences among treatments with respect to bee abundance. However, some of the specific characteristics of the plant community showed significant relationships with bee response variables. Pastures with greater abundance of floral resources had greater bee abundance but lower bee nutritional indicators. Bee nutritional variables were positively correlated with vegetation height, but, in some cases, negatively correlated with stocking rate. These results suggest grassland site characteristics such as floral resource abundance and stocking rate are of potential importance to bee pollinators and suggest avenues for further research to untangle the complex interactions between grassland management, plant responses, and bee health.