Field-year realized evapotranspiration and a selected reference-field benchmark for corn, South Platte River Basin, Colorado, 2020–2025: derived analysis database and code

Derived data and analysis code supporting the article "Sprinkler-Irrigated Corn Fields Show Higher Realization of a Reference-Field Evapotranspiration Benchmark: A Six-Season Satellite Analysis of Colorado's South Platte Corridor". The archive contains the field-year analysis database of 12,430 commercial corn field-years on 5,091 irrigated fields across six processing domains of the South Platte River Basin, Colorado, for the 2020-2025 growing seasons. For each field-year it gives seasonal realized evapotranspiration from Landsat surface-energy-balance retrievals, seasonal benchmark evapotranspiration from a documented double-logistic crop-coefficient curve fitted to a selected reference-field population, the realization ratio, the benchmark gap in millimetres and acre-feet, irrigated area on both the parcel and measured-corn bases, and generalized water-source and irrigation-method classes. The benchmark curve is fitted to Reference Corn Fields selected by peak seasonal NDVI, groundwater or conjunctive water source, and the upper quartile of mean productive-season observed crop coefficient. Because that final step ranks on the observed coefficient itself, the benchmark is endogenous to that extent and describes the upper quartile of observed behaviour among eligible fields rather than an independently established potential; the article states this explicitly and the archive ships the frozen membership list so the selection can be inspected directly. Also included: the frozen reference-field membership list (1,105 field-years on 831 fields, flagging the 949 field-years and 11,408 observations the adopted curve was fitted to); the daily reconstruction series (1,998,538 field-days of interpolated crop coefficients, planting-adjusted growing degree days and daily rescaled reference evapotranspiration) from which both seasonal integrals are re-derived exactly; the field-scene observation table (189,676 records); the benchmark crop-coefficient parameter table; the daily reference-ET rescaling series and its station-based verification; gate-by-gate inclusion counts; all mixed-model output (coefficients, adjusted means, contrasts, omnibus tests, variance components) and field-clustered bootstrap replicates; benchmark-parameter and reference-ET sensitivity analyses; drought-context tables; the manuscript figures; and the analysis code, including self-contained scripts that reproduce the headline results and refit the baseline additive and geometry-adjusted models from the archived tables alone. Fields are identified only by a stable surrogate key. Field boundaries, coordinates, parcel identifiers, surface-water diversion identifiers and well identifiers are not released, because together they link each field to specific decreed water rights and identifiable agricultural operations. Primary inputs (Landsat Collection 2, the Cropland Data Layer, Colorado Decision Support Systems irrigated-lands coverages, CoAgMET, gridMET) are not redistributed and are cited to their originators in DATA_AVAILABILITY_STATEMENT.md within the archive. Code under code/ is additionally available under the MIT License. The two GeoJSON layers under gis/ are public-domain U.S. Government works. Scripts under code/pipeline_archive/ are preserved unedited as a record of what was executed, so their comments retain earlier working terminology. Start with README.md. VARIABLES.md documents every variable in every table with units, definition and provenance. MANIFEST.md lists every file with a one-line description, and CHECKSUMS.sha256 covers every file in the archive. Changes in this version: the volumetric contrasts are re-derived on a corrected, CDL-measured corn acreage basis rather than the full-parcel boundary area; the verification script's expectation set and unit conversion are corrected, the earlier version having printed every realization figure at 0.938 times its true value against a superseded four-region expectation set; the frozen reference-field membership list and the field-scene observation table are added; the reference-ET series is correctly identified as the CoAgMET Penman-Kimberly alfalfa-reference series; and terminology follows the revised article, which reports realization against a selected reference-field benchmark rather than a near-potential one.