Simple-anoxia-model

Ten thousand lakes are chosen for the upper Midwest region of the US. For these lakes, water quality data are downloaded from the USGS Water Quality Portal. These data are sparse in space and time for most lakes, but some lakes have high quality long-term time series. For these same 10,000 lakes, the USGS has already modeled hydrodynamics (water temperature) over 1979-2015. We will develop a simple model for lake dissolved oxygen concentration and test it against high quality time series for lakes from the NTL LTER. This model will be calibrated and validated for lakes within the 10,000 lakes data set that have sufficient dissolved oxygen data. For all lakes, a combined process-guided machine learning approach will be used within a “transfer learning” framework to predict dissolved oxygen dynamics and anoxia probability for each lake. For more details on the science and the progress, please see the “prospectus” located here: https://drive.google.com/drive/u/1/folders/1lndLnjLOQYW4z5JIv51xqtjoIE3rLnEt