JAWRA 2025 - LSTM attributes analysis

These are LSTM predictions coming from different sets of static attributes. The "PC" indicates the principal component of the HydroAtlas (Linke et al., 2019). The number after PC indicates the number of PCs used in the LSTM model. For instance, PC1 is the first principal component, PC2 is the first two principal components, etc. Also included here are LSTM runs with randomly selected HydroAtlas attributes (HA_10randomAttributes_[a through i]). We also have LSTM runs that use NOAH-MP parameters as static attributes. Results from these data will be included in an upcoming peer reviewed publication in the Journal of the American Water Resources Association, a posted with preiliminary results is available online (Frame et al., 2024). LSTM training and testing was done with NeuralHydrology (Kratzert et al., 2022). Included are zip files with ensemble mean predictions, which are used in the published figures and analasis, but we also include the individual model runs (individual_runs.zip). Code is available to analyze these runs, which is available at this GitHub repository: https://github.com/NWC-CUAHSI-Summer-Institute/attribute_pca AKNOWLEDGEMENTS: NOAA NWC Innovators Program, Summer Institute. Administered by the CUAHSI andthe University of Alabama under the CIROH; NOAA Cooperative Agreement NA22NWS4320003. This research waspartially supported by NOAA cooperative agreement (grant number NA19NES4320002). The research resultspresented herein represent the personal opinion of the authors, and not official policy of the U.S. Department ofCommerce or the National Oceanic and Atmospheric Administration References: Frame, J. M., Araki, R., Bhuiyan, S. A., Bindas, T., Rapp, J., Bolotin, L., Deardorff, E., Liu, Q., Haces-Garcia, F., Liao, M., Frazier, N., & Ogden, F. L. (2024). Machine learning for a heterogeneous water modeling framework. ESS Open Archive. https://doi.org/10.22541/essoar.171690771.18409532/v1 Kratzert, F., Gauch, M., Nearing, G., & Klotz, D. (2022). NeuralHydrology---A Python library for Deep Learning research in hydrology. Journal of Open Source Software, 7(71), 4050. Linke, S., Lehner, B., Ouellet Dallaire, C., Ariwi, J., Grill, G., Anand, M., Beames, P., Burchard-Levine, V., Maxwell, S., Moidu, H., Tan, F., Thieme, M. (2019). Global hydro-environmental sub-basin and river reach characteristics at high spatial resolution. Scientific Data 6: 283. doi: https://doi.org/10.1038/s41597-019-0300-6