Integrating Artificial and Collective Intelligence in Hydro-Economic Modeling for Sustainable Irrigation and Drought Adaptation in Colombia

Water-scarce smallholder regions require irrigation strategies that are both biophysically credible and socially legitimate. We propose an integrated hydro-economic decision-support framework for the Cesar Department (Colombia) that couples (i) AI emulation of FAO AquaCrop crop–water responses, (ii) Bayesian aggregation of collective intelligence elicited under Shared Socioeconomic Pathways (SSP) using an AHP-anchored, two-round Delphi protocol, and (iii) a household linear program enforcing diversification and a Top-M social-alignment rule with penalized slack. XGBoost surrogates generate household-specific yield and net irrigation requirement coefficients for four crops under a dry-year baseline (test R 2 : 0.89–0.93 for yield; 0.87–0.91 for NIR). Expert-derived SSP-specific crop priority vectors are modeled on the simplex and combined via a Dirichlet posterior to produce crop-preference weights with 95% credible intervals, operationalized as social-weight layers (Base/Low/High/Uniform/Favored). The model solves 2,520 household–scenario combinations (168 households × 3 SSPs × 5 layers) with full feasibility and zero slack use. Increasing normative leverage through the minimum-alignment quota (α) and social-reward scaling (μ) raises the social welfare component with limited income disruption in SSP1 and SSP4: in SSP4, the mean objective increases from USD 8,334 (Base) to USD 11,427 (Favored), driven by the social term (USD 3,283→6,382) while net income remains stable (∼USD 5,045–5,052). SSP3 provides a stress test, revealing tighter feasibility (alignment 0.582 and net income USD 3,661 under Low; binding rate 22%). Across Socioeconomic Vulnerability Index (SEVI) terciles, mean income declines by ∼42% while alignment remains high, motivating complementary investments that relax constraints for high-vulnerability households. Overall, embedding uncertainty-aware social priorities into hydro-economic optimization enables policy-relevant, scenario-conditional irrigation guidance while transparently revealing when preference-consistent targeting is compatible with feasibility and when it requires enabling investments. • Develops drought-focused sustainability indicators linking farm income, irrigation-water use, and stakeholder crop priorities, aligned with UN Sustainable Development Goals (food security—SDG 2; water—SDG 6; climate action—SDG 13). • Quantifies crop “priority indicators” with Bayesian uncertainty (Dirichlet posterior; 95% credible intervals). • Validates crop performance indicators using XGBoost surrogates (test R 2 : 0.89–0.93 yield; 0.87–0.91 net irrigation). • Runs 2,520 farm–future–policy cases (168×3×5) to compare indicator trade-offs across settings. • Stronger social guidance raises the overall objective (USD 8,334→11,427); water scarcity binds 22% of farms and high-vulnerability income is ∼42% lower.