Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Tempus2 Modeling topic
No spam. Unsubscribe anytime.
Researchers present Tempus 2: satellite‑based, regional stream temperature estimates for ungauged sites
Summary
Tempus 2 uses satellite land‑surface temperature, humidity, topography and land‑cover to estimate daily stream temperatures at ~1 km resolution for ungauged sites; the model typically shows an RMSE around 2°C for ungauged validation gauges, with limitations in reservoirs and groundwater‑dominated sites.
Get email alerts on the Tempus2 Modeling topic
No spam. Unsubscribe anytime.
Daniel (presenting for the Tempus 2 team) described a remote‑sensing approach to fill gaps in stream temperature coverage. The model uses satellite land‑surface temperature, humidity, topography and land‑cover to predict daily mean and maximum stream temperatures at approximately 1‑kilometer resolution without requiring local calibration.
"The model outputs 1 kilometer resolution," Daniel said, adding that Tempus 2 is designed to be usable at scale and to run where local monitoring is absent. In U.S. validation against USGS gauges, Daniel reported typical root‑mean‑squared errors for ungauged sites of about 2 degrees Celsius and noted that the model captures roughly 90% of variance in daily mean temperature across sites and days.
Daniel outlined practical uses and limits: Tempus 2 can support large‑scale assessments of thermal suitability, generate long time series for thousands of points, and prioritize locations for new monitoring. It performs less well in areas strongly affected by large reservoirs or atypical groundwater dynamics; local observations can be used to fine‑tune model output when long periods of record are available.
The team provides automated data retrieval (currently via Google Earth Engine), a pretrained model and the option to produce regional datasets. Daniel said they are open to requests for specific derived metrics (for example, cumulative degree days, 7‑ or 30‑day maxima, and uncertainty estimates) and for tailored datasets covering many points in Southern California or nationwide.

