Engineer details watershed-model calibration and validation using Navarro River data
Get AI-powered insights, summaries, and transcripts
Subscribe
Summary
At a technical briefing, engineer Akash Rasher explained how the team's LSPC watershed models were calibrated and validated, citing calibration years 2018–2023 and validation years 2005–2017 for the Navarro River watershed and noting vetted metrics fell mostly in a 'very good' category.
Akash Rasher, an engineer in the supply and demand section, described how the LSPC interface lets users customize model-run parameters and recommended using the team’s prebuilt watershed models when there have been no major land-use changes to the watershed.
Rasher said a key advantage of the provided models is that “they've been calibrated and validated for current conditions.” He explained calibration tunes model parameters to match observed data while validation tests the tuned model against independent datasets to ensure robustness in other scenarios.
Using the Navarro River watershed as an example, Rasher said the team used USGS streamflow gauge data from water years 2018–2023 for calibration and tested performance against data from 2005–2017 for validation. He described one common check as visually comparing modeled and observed hydrographs to see whether peaks and valleys align in timing and magnitude, and said the team also applies statistical metrics, including R-squared and other thresholds drawn from published literature, to measure accuracy of flow volumes and timings.
Rasher presented a table from the Navarro River model development report showing monthly-scale metrics and said most measures fell into a “very good” category. He emphasized that the team evaluates both daily and monthly timescales but prioritizes monthly results because diversion records are recorded by month.
Rasher also said the Navarro River analysis incorporated limited low-flow data provided by the North Coast Regional Water Quality Control Board; that dataset was insufficient for calibration but sufficient for additional validation checks and is summarized in the model development report. He pointed listeners to the team’s model development reports on the website for citations, detailed methods and thresholds.
Rasher concluded by advising users who choose not to use the provided configuration to learn how to modify LSPC model parameters before running alternate scenarios. The presentation did not include motions or formal decisions.