Seasonal dynamics of water source use and multi-objective precision irrigation optimization for greenhouse rose cut-flower cultivation

Greenhouse rose production requires irrigation strategies that balance water conservation with plant nutritional status, yet such strategies depend on understanding seasonal shifts in plant water uptake and the pathways linking irrigation, soil water, and plant responses. Here, we conducted a one-year intensive case study in a commercial greenhouse rose system in Jinning, southwest China, using a purposively selected and fully instrumented site within a major local production cluster. The site was chosen to resolve seasonal process relationships under a defined production context rather than to serve as a statistical proxy for all greenhouses in the district. We developed an integrated workflow that combined stable isotope measurements and nonlinear programming for water-source partitioning, partial least squares structural equation modeling (PLS-SEM) for pathway identification, and Pareto optimization for irrigation scenario screening. The isotope results indicated clear seasonal changes in uptake depth: shallow soil layers (0–30 cm) dominated during the wet season, whereas reliance on deeper layers (50–80 cm) increased during the dry season. The PLS-SEM results further suggested that irrigation effects on composite plant status were transmitted mainly through soil-water conditions, and the scenario analysis identified a seasonally differentiated balanced strategy as a candidate under the modeled nutrient constraints. Relative to the current management benchmark, this candidate occupied an intermediate trade-off position with only moderate adjustment in irrigation frequency and annual input while maintaining nutrient thresholds and favorable plant P, plant K, and leaf δ¹ ³C values in the scenario space; here, leaf δ¹ ³C is interpreted conservatively as a complementary carbon-isotope indicator related to plant water-use behavior rather than as a direct measurement of water-use efficiency (WUE). Because the evidence comes from one intensively monitored greenhouse and simulation-based scenario comparison, the proposed strategy should be interpreted as a context-dependent, mechanism-informed candidate rather than as a universally validated prescription. Overall, the integrated stable isotope, PLS-SEM, and Pareto workflow offers a transparent, case-based screening framework for linking observed water-source dynamics with candidate precision irrigation strategies in similarly managed greenhouse rose systems.