Search for: All records

Abstract The Chesapeake Bay is the largest estuary in the continental United States. Extreme temperature events, termed marine heatwaves, are impacting this ecologically important zone with increasing frequency. Although marine heatwaves evolve across space and time, a complete spatial picture of marine heatwaves in the Bay is missing. Here, we use satellite sea surface temperature to characterize marine heatwaves in the Chesapeake Bay. We consider three products: NASA MUR, NOAA Geo-Polar, and Copernicus Marine OSTIA, and validate their effectiveness using in situ data from the Chesapeake Bay Program. We find that Geo-Polar SST is the most suitable dataset for marine heatwave analysis in this location, with a root mean squared error of 1.6$$^\circ $$ $_{}^{\circ }$C. Marine heatwaves occur on average of 2.3 times per year and last 10.8 days per event. A north-south (along estuary) gradient is identified as a common pattern of spatial variability. Seasonally, summer marine heatwaves are shorter, more frequent, and have a more consistent duration, with an inter-quartile range of 6–11 days (median=8 days). December marine heatwaves have a much larger inter-quartile range of 6–28 days (median=13 days). Marine heatwaves are increasing at a rate of 4 events/year in the upper Bay and 2 events/year in the main stem of the lower Bay. Our analysis suggests that the major observed spatial patterns are a result of long-term warming, not shifts in the spread of the temperature distribution. Overall, the qualitative character of marine heatwaves in the Chesapeake Bay is not changing but they are becoming more frequent. 

Important natural resources in the Arctic rely heavily on sea ice, making it important to forecast Arctic sea ice changes. Arctic sea ice forecasting often involves two connected tasks: sea ice concentration at each pixel and overall sea ice extent. Instead of having two separate models for two forecasting tasks, in this report, we study how to use multi-task learning techniques and leverage the connections between ice concentration and ice extent to improve accuracy for both prediction tasks. Because of the spatiotemporal nature of the data, we designed two novel multi-task learning models based on CNNs and ConvLSTMs, respectively. We also developed a custom loss function which trains the models to ignore land pixels when making predictions. Our experiments show our models can have better accuracies than separate models that predict sea ice extent and concentration separately, and that our accuracies are better than or comparable with results in the state-of-the-art studies.