Rain-Gated Retractable-Roof Control in Semi-Arid Greenhouses: Why Threshold Rules Fail and What Predictive Aperture Control Buys

Abstract

Protected cultivation in semi-arid regions is irrigated almost entirely from groundwater that is frequently brackish, while rainfall is shed off the greenhouse cover and lost. Retractable-roof greenhouses make it physically possible to admit that rainfall directly to the crop, and a natural control policy is to gate roof opening on a rain signal and a soil-moisture threshold. We formalise this policy, embed it in a coupled soil-water, salt and energy balance driven by eight years of hourly synthetic weather calibrated to Algerian high-plateau climate normals (423 mm/yr rainfall, 1225 mm/yr reference evapotranspiration), and evaluate it against four alternatives. The threshold policy fails, and it fails structurally rather than through mistuning: operated alongside a conventional moisture-triggered drip system it admits 2.2 mm/yr of rainfall, because the drip system holds the root zone above the opening threshold so the gate is almost never satisfied; operated as the primary water source it reduces groundwater use by 79 % but collapses actual transpiration from 908 to 474 mm/yr, with a water-stress index of 42 %. An exhaustive sweep of 40 threshold pairs leaves the best standalone configuration at 33 % stress, so no choice of thresholds recovers the policy: a gate on current moisture cannot reserve storage for forecast rain. We therefore formulate roof operation as a receding-horizon decision over a continuous aperture with an explicit, decomposable cost. The resulting controller reduces groundwater use by 30.0 % relative to threshold drip while lowering crop water stress (1.61 % versus 2.74 %) and mean root-zone salinity (ECe 3.30 versus 3.69 dS/m), at a cost of 303 h/yr of canopy wetness and 88 roof cycles per year. It is nevertheless dominated on water by passive rainwater harvesting into a tank, which reaches 35.2 % savings with no canopy wetting at all. The open roof is preferable only where irrigation water is saline: its advantage in ECe grows monotonically with ECiw and becomes material above 1.8 dS/m. We conclude that rain-gated roof opening is not a general water-saving technology but a targeted salinity-management technology, and we report the conditions delimiting its usefulness.