Understanding the patterns and relationships between vegetation productivity and climatic conditions is essential for predicting the future impacts of climate change. Climate change is altering precipitation patterns and increasing temperatures in the Southwest United States. The large-scale and long-term effects of these changes on vegetation productivity are not well understood. This study investigates the patterns and relationships between seasonal vegetation productivity, represented by Moderate Resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI), and the Standardized Precipitation Evapotranspiration Index (SPEI) across the Mojave, Sonoran, and Chihuahuan Deserts and the Apache Highlands of the Southwest United States over 16 years from 2000 to 2015. To examine the spatiotemporal gradient and response of vegetation productivity to dry and wet conditions, we evaluated the linear trend of different SPEI timescales and correlations between NDVI and SPEI. We found that all four ecoregions are experiencing more frequent and severe drought conditions in recent years as measured by negative SPEI trends and severe negative SPEI values. We found that changes in NDVI were more strongly correlated with winter rather than summer water availability. Investigating correlations by vegetation type across all four ecoregions, we found that grassland and shrubland productivity were more dependent on summer water availability whereas sparse vegetation and forest productivity were more dependent on winter water availability. Our results can inform resource management and enhance our understanding of vegetation vulnerability to climate change.

We present an analysis of the problem of identifying biolog- ical context and associating it with biochemical events in biomedical texts. This constitutes a non-trivial, inter-sentential relation extraction task. We focus on biological context as descr iptions of the species, tissue type and cell type that are associated with biochemical events. We present a new cor pus of open access biomedical texts that have been annotated by biology subject matter exper ts to highlight context- event relations. Using this cor pus, we evaluate several classiers for context-event association along with a detailed analysis of the impact of a var iety of linguistic features on classier perfor mance. We finnd that gradient tree boosting performs by far the best, achieving an F1 of 0.865 in a cross-validation study.

We extend first-order model agnostic meta-learning algorithms (including FOMAML and Reptile) to image segmentation, present a novel neural network architecture built for fast learning which we call EfficientLab, and leverage a formal definition of the test error of meta-learning algorithms to decrease error on out of distribution tasks. We show state of the art results on the FSS-1000 dataset by meta-training EfficientLab with FOMAML and using Bayesian optimization to infer the optimal test-time adaptation routine hyperparameters. We also construct a small benchmark dataset, FP-k, for the empirical study of how meta-learning systems perform in both few- and many-shot settings. On the FP-k dataset, we show that meta-learned initializations provide value for canonical few-shot image segmentation but their performance is quickly matched by conventional transfer learning with performance being equal beyond 10 labeled examples. Our code, meta-learned model, and the FP-k dataset are available at https://github.com/ml4ai/mliis .