New water accounting reveals why the Colorado River no longer reaches the sea

The study addresses why the Colorado River no longer reaches the Gulf of California by providing a comprehensive accounting of consumptive uses across sectors, sub-basins, exports, and environmental losses. Despite the river’s critical role in supplying water to more than 40 million people and extensive irrigated agriculture, long-term overconsumption has led to frequent annual deficits—16 of 21 years during 2000–2020—necessitating withdrawals from Lake Mead and Lake Powell, which fell to roughly three-quarters empty by the end of 2022. Historical complexities in the river’s governance, including ambiguities in the 1922 Colorado River Compact, exclusion of the Gila River from Compact accounting, and delayed specification of Mexico’s allocation until the 1944 Treaty, have hindered full sectoral and crop-specific accounting. The purpose is to produce a complete water budget for 2000–2019 that includes within-basin uses, exports, Mexico’s use, reservoir evaporation, and riparian/wetland evapotranspiration to inform policy and balancing use with supplies.

Prior accounting efforts by the US Bureau of Reclamation (e.g., the 2012 Colorado River Basin Water Supply and Demand Study) did not include the Gila River basin due to its exclusion from Compact allocations and did not detail sectoral uses in Mexico. Earlier work by Richter et al. provided a basin water budget with crop-specific consumption but omitted Mexico’s uses, reservoir evaporation, riparian evapotranspiration, and exports outside the physical basin. Given that roughly 19% of river water is exported or used in Mexico and that the Gila is a major tributary, previous omissions produced inaccuracies and uncertainty in discussions about where the river’s water goes. The paper builds on and corrects these shortcomings by incorporating all major use categories and indirect losses.

Study period and aggregation: Data aggregated annually for 2000–2019 and averaged. The study compiles direct human uses (MCI and agricultural), exports, and indirect losses (reservoir evaporation; riparian/wetland evapotranspiration). The authors note inconsistencies in measurement/estimation methods across sources and over time and plan updates as methods evolve. - MCI and agricultural water consumption (Upper Basin): Primary data from US Bureau of Reclamation’s Upper Basin Consumptive Uses and Losses (CUL) reports. Because direct measurement is incomplete, Reclamation uses estimation methods to compute consumptive use; New Mexico provides its own estimates reviewed with Reclamation. Thermoelectric consumptive use is provided by power companies. Municipal and industrial consumption is derived from USGS five-year water use reports. Groundwater included only where hydraulically connected to surface flows in CRB per Reclamation; deep/isolated groundwater excluded. - MCI and agricultural water consumption (Lower Basin excluding Gila): From Reclamation’s annual Colorado River Accounting and Water Use Reports for Arizona, California, Nevada based on measured deliveries and return flows. Where sectoral attribution was not explicit, the authors classified end-uses from water user information. For entities averaging <1,230 million m³ (1000 acre-feet) per year, sectoral allocation followed basin-level MCI/ag percentages. - Central Arizona Project (CAP) disaggregation: All subcontract deliveries (2001–2019) were classified by final end-use (including leases and groundwater storage for later use). Native American agricultural allocations leased to cities were counted as MCI. - Gila River basin: Surface-water uses from Arizona Department of Water Resources for the five AMAs (Prescott, Phoenix, Pinal, Tucson, Santa Cruz), separated by sector and source; only surface water from the Gila hydrologic system included. Upper Gila data from University of Arizona. Assumed essentially full consumption within the Gila (on average <1% of flow exits to the Colorado River). - Mexico: Sectoral uses based on Hernandez-Cruz et al. (deliveries 2008–2015), with agricultural use estimated from Mexican agricultural surveys and agency consultations; cross-checked against Reclamation’s cross-border delivery volumes. - Exported water (inter-basin transfers): Annual average transfers for 46 canals/pipelines obtained from Reclamation’s CUL data (Upper Basin) and from Siddik et al. (Colorado River Aqueduct). All exported water considered consumptive relative to CRB. End-uses (MCI vs agriculture) determined from project documentation/operator communications; for 17 small transfers (<247,000 m³ or 200 acre-feet each), applied average sectoral shares from larger transfers. - Reservoir evaporation: Estimated using Reclamation HydroData and standardized Penman–Monteith method (Lower Colorado annual summaries). Salt River Project supplied Gila reservoir evaporation. Potential bank storage/seepage in/out of reservoirs recognized but not quantified, so excluded from budgets. - Riparian and wetland evapotranspiration (ET): Included ET for riparian/wetland vegetation along mainstem and tributaries, using recent riparian mapping and remote-sensing/ET datasets. Accounted for drainwater-supported systems (e.g., Salton Sea; Cienega de Santa Clara in Mexico) in riparian/wetland ET tallies. - Crop-specific consumption: Total agricultural consumption by sub-basin from above sources was allocated to crops using modeled crop water requirements. The team used AquaCrop-OS (v5.0a) to simulate monthly water requirements (1981–2019) for 15 major crops (covering most irrigated area), supplemented by a simple model for a validation crop lacking parameters. Inputs included USDA crop area and calendars, FAO parameters, gridMET climate (precipitation, Tmin/Tmax, reference ET), soil data (NRCS/SoilGrids), and crop parameters from AquaCrop-OS. Results partitioned into blue/green water following established frameworks. Data and code availability: Data compiled on HydroShare; AquaCrop-OS model used; riparian mapping via Dryad and processed with ArcGIS Pro/Google Earth Engine.

- Comprehensive 2000–2019 water budget including within-basin uses, exports, Mexico, reservoir evaporation, and riparian/wetland ET. - Irrigated agriculture is the dominant consumer: 52% of overall basin consumption (text) and 54% per Fig. 4. Agriculture accounts for 74% of direct human use. - MCI direct uses constitute about 9% (Fig. 4) of total consumptive use. - Indirect environmental and system losses are substantial: riparian/wetland evapotranspiration ~19% of total; reservoir evaporation reported as 11% in Results text but shown as 18% in Fig. 4. - Cattle-feed crops: Grass hays/alfalfa collectively account for 46% of all direct human water consumption; in the Upper Basin, cattle-feed crops consume ~90% of agricultural water and generate over three times the economic value compared to MCI uses combined. - Exports and Mexico: Nearly one-fifth of river water is exported or used in Mexico. Sectoral accounting now includes Mexico and all exported volumes, resolving prior omissions. - Temporal variability: Inter-annual variability across sectors is high (ranges equivalent to 24–47% of 20-year averages). Notable decreases in Lower Basin consumption from 2000 to 2019 for MCI (−38%) and agriculture (−15%), partly due to shortage and conservation agreements (2001–2023). - Policy-relevant volumes: Average annual consumptive use during 2000–2019 estimated at 23.7 billion m³ (193 MAF). Stabilizing reservoirs likely requires 3–4 billion m³/yr (2.4–3.2 MAF/yr) reduction in Upper and Lower Basins, with an additional 1–3 billion m³/yr (0.81–2.4 MAF/yr) by 2050 under continued warming.

By incorporating exports, Mexico’s uses, reservoir evaporation, and riparian/wetland ET, the revised accounting provides a more complete picture than prior efforts. Compared to the 2020 analysis, the share of direct consumption attributed to cattle-feed crops decreases from 53% to 46% due to inclusion of indirect losses and improved crop modeling. Nevertheless, the core conclusion remains: irrigated agriculture dominates basin water consumption. The inclusion of riparian/wetland ET (about one-fifth of total) highlights the substantial environmental water dependence along the river and its tributaries, reinforcing the need to leave more water in-channel to sustain ecosystems, including the degraded delta. Documented reductions in Lower Basin uses since 2000 indicate that agreements can lower consumption, but scientists argue that current cuts are insufficient to balance with 21st-century supplies. The quantified reductions needed (3–4 billion m³/yr, potentially rising by 2050) provide concrete targets for ongoing binational negotiations toward achieving a sustainable supply-demand balance.

This study delivers the most comprehensive Colorado River water budget to date for 2000–2019, resolving long-standing accounting gaps (Gila River basin, Mexico, exports, reservoir evaporation, riparian/wetland ET) and refining crop-specific use estimates. It confirms that irrigated agriculture remains the dominant consumer, that riparian/wetland ET and reservoir evaporation are major components, and that cattle-feed crops drive a large share of direct human use. The results provide an empirical foundation for current and post-2026 management negotiations. Future research and updates should: incorporate improved measurement methods (e.g., updated ET and crop models), refine sectoral allocations for small transfers, better quantify reservoir bank storage/seepage, and update Mexico and Gila accounting as new data emerge, especially under continued aridification and changing demands.

- Data heterogeneity: Different agencies and methods (e.g., Reclamation CUL, USGS, AMA reports, Mexico sectoral data) use varying estimation/measurement approaches; updates may change figures. - Temporal coverage mismatches: Some Mexico sectoral estimates (2008–2015) do not perfectly align with the 2000–2019 averaging window. - Groundwater treatment: Only hydraulically connected groundwater to surface flows is included; deep/isolated groundwater uses are excluded, potentially underrepresenting total local water use though appropriate for a river-consumptive accounting. - Reservoir bank storage/seepage not quantified: Gains/losses to reservoir banks are recognized but excluded due to insufficient measurement, potentially biasing net evaporation losses. - Small inter-basin transfers: For 17 low-volume exports, sectoral uses were inferred from average shares, introducing uncertainty. - Model uncertainties: Crop water estimates (AquaCrop-OS and supplemental simple models), ET estimates (Penman–Monteith, riparian mapping/remote sensing), and sectoral allocations carry parameter and input uncertainties; inter-annual variability is high, and climate change may alter future baselines.