Authors: Daniel Philippus, Claudia R. Corona, and Terri S. Hogue – Colorado School of Mines f Mines
Title: Large-Domain Coastal and Estuarine Water Temperature Prediction in NextGen
Presentation Type: Poster Presentation
Abstract: Coastal and estuarine water temperature (CEWT) strongly influence ecological system resilience. For example, robust wetland health is intricately linked to CEWT, where CEWT influences biological cycles such as germination rates, growth cycles and activity levels of cold-blooded animals, including keystone species. CEWT further plays an important role in the reproductive cycles of many species that transition from the inland to the sea, such as anadromous fish migration, growth, and survival. Furthermore, because CEWT strongly influences coastal hydrodynamics (e.g., flooding) and ice formation, CEWT data are essential for ensuring the safety of coastal infrastructure (e.g., roads and ports) and supporting effective fisheries management. However, CEWT data are often unavailable in operational contexts because many river outlets lack temperature monitoring and there are no validated techniques for CEWT prediction at unmonitored sites over large domains. Existing inland stream water temperature (SWT) models can generate CEWT predictions in unmonitored locations, but their performance in modeling temperature changes in estuarine/coastal environments is unknown. Moreover, no extant CEWT model provides a Basic Model Interface implementation, so NextGen-based national water modeling efforts supporting inundation, ice, or ecological forecasts cannot currently incorporate CEWT predictions. In this study, we evaluate the performance of available NextGen-compatible, CONUS-scale SWT models for predicting CEWT and identify modifications to effectively represent CEWT-specific processes. We evaluated the TempEst 2 SWT model (hindcasting) and the TempEst-NEXT SWT model (forecasting), which are inland SWT models developed in an earlier CIROH project. In preliminary testing for sites on the East and West coasts of the CONUS, the two TempEst models exhibited moderately reduced performance for CEWT prediction (NSE ~0.7 for daily mean CEWT, compared to NSE ~0.9 for inland SWT). However, both models performed poorly in the Great Lakes (NSE ~0.3-0.4). Based on these preliminary findings, we anticipate that estuarine/coastal modifications to the TempEst models will involve adjusted representation of CEWT seasonality. We further hypothesize that short-term variation will be more sensitive to the ocean/lake boundary condition relative to local weather variations. The goal of the finalized models is to support operational prediction through NextGen for improved fisheries management and real-time planning for coastal water levels and ice formation.