Water‑supply staff outline automated demand methodology for DRAT modeling

Peymon Olami, senior specialist in the Water Supply and Demand Assessment and Instreamflow Section, presented the team's demand‑data methodology used to build inputs for the Drought Water Right Allocation Tool (DRAT), saying the approach has been applied to the Russian River and to datasets for nine other watersheds the unit is working on.

The procedure combines automated scripts that pull ERIMS flat files with manual GIS and record review. "I'd say about 80% of the process is automated and 20% is manual," Olami said, and he described a workflow that flags suspect records, applies standardized correction actions, and produces a CSV demand table DRAT consumes.

Why it matters: DRAT allocates available supply while respecting the water‑rights priority system. More consistent, QA/QC'd demand inputs affect modeled allocations and curtailment recommendations, making the accuracy of ERIMS reporting and the team's corrections consequential for downstream modeling and management.

How the process works: Staff begin by pulling multiple ERIMS (RMS) flat files using R scripts, then filter to include only active reporting points of diversion (PODs) and selected water‑right types. Included types were described as registrations, appropriatives, misstatements and stock bonds; excluded categories included adjudicated rights (managed by a Watermaster), appropriative state filings, groundwater recordations, very old section‑12 filings and wastewater change petitions. The presenters emphasized that the inclusion/exclusion rules follow the division's original demand methodology.

After filtering, the team runs a GIS preprocessing step to check POD locations. Olami said ERIMS plotting is "about 98% accurate," but a few misplotted high‑diversion PODs can materially change results. The team flags four common issues: PODs plotted inside the watershed that actually lie outside, PODs plotted outside that actually lie inside, PODs plotted inside that deliver to a different watershed, and PODs outside but hydrologically connected. They use PLSS overlap, coordinate overlap and a name/text search of ERIMS fields to flag records; flagged items are reviewed manually and corrected in ERIMS when needed.

Detecting reporting errors: The scripts look for duplication (same totals repeated across multiple water‑right reports), unit‑conversion mistakes (for example reporting gallons when the field expects acre‑feet) and empty or monthly‑missing reports. The team described a standardized spreadsheet workflow: scripts export flagged records to review spreadsheets, reviewers record standardized QAQC actions and the scripts then apply those actions to the dataset. For unit conversions the current automated flagging thresholds are reports that differ by more than 100 times from a reference value or differ from average/median totals by more than 100 acre‑feet; flagged records are examined and converted when the evidence (for example gallons reported in a direct column) shows a units error.

Handling multiple PODs and subbasin assignment: DRAT models supply and demand across interconnected subbasins. For a single POD, the POD coordinates determine the subbasin assignment. When a water right has multiple PODs in the same flow path, the presenters said they assign the most downstream subbasin to maximize modeled availability. If PODs lie in disconnected flow paths, staff split the right into subrights and apportion reported diversion volumes to subbasins based on relative drainage area (including upstream contributing area). The team uses a watershed connectivity matrix (built from stream data) to determine flow relationships.

Scope, history and collaboration: Presenters said the team intentionally began analyses with 2017 reporting because earlier ERIMS records are structurally inconsistent and generally less reliable. "We decided to stick with 2017 as the starting point," Olami said. The scripts and review materials are hosted in a GitHub repository; staff offered to add collaborators and said they will continue updating demand tables when new water‑year data are available (presenters noted an update to include water‑year 2024 later in the year). A contractor (Paradigm) is producing an updated, watershed‑agnostic version of DRAT that the group intends to use in future work.

Questions and caveats raised in the session included whether older records should be incorporated, the tradeoffs between automation and manual review, and how splits and downstream assignment interact with priority‑date curtailment. Presenters and other staff cautioned that the scripted rules produce false flags that still require human judgment (for example frost‑protection rights that legitimately vary widely year to year), and noted that adjudicated rights and Watermaster‑administered records have different reporting pathways and were excluded from the automated demand table.

Next steps: The team will share slides, the recording and repository access to interested staff; they plan periodic updates as new RMS/ERIMS submissions arrive and expect to adopt the contractor's updated DRAT implementation when it becomes available.