Irrigation priority, late hedging and an undocumented minimum-flow rule: twelve years of gauged allocation at Poechos, Chira basin, Peru

How multipurpose reservoirs divide water among competing uses is rarely observed and mostly simulated. We read it from 12.6 years of daily gauging of inflow and every outlet of the Poechos reservoir (northern Peru), whose balance closes by two independent routes (r² = 0.979 and 0.977). Under low flow, irrigation canals receive 80% of the release and the river 12.5%; from high to low flow, they lose 39%, and the river 81%. Hedging starts late: models with seasonal controls place the trigger at 26-38% of maximum storage. The river is held at exactly 5.0 m³/s on 1463 days (32%) and falls below it on 232, yet the operator's normative documents available to us contain no minimum-flow provision. In 2024, the river was dry on 70 days while canals received three times the volume the setpoint required. Regulation removes the dry half of the regime: median flow falls 87% and zero-flow days rise from 0 to 92. Warm ENSO summers precede failing dry seasons (r = +0.71, n = 12): a preliminary hypothesis with four months' lead. The record exposes a de facto allocation priority that no regulation states and sets a gauged baseline for auditable environmental flows.