Satellites and Fuzzy Logic Reveal Who Really Gains When Scarce Irrigation Water Is Shared

In one of the driest corners of the Mediterranean, a team of economists and water scientists has built a framework that asks a deceptively simple question with profound consequences: when a government allocates a scarce cubic metre of irrigation water, who actually benefits, and by how much? The answer, according to new research published in Social Indicators Research, is that it depends—critically and measurably—on the state of the land, the infrastructure, and the season in which that water arrives. The work, led by Besma Belhadj of the University of Tunis El Manar and Bechir Fridhi of Majmaah University, turns satellite measurements of crop water use into causal estimates of the value of water, and then uses those estimates to compare different philosophies of water allocation across Tunisia.

The starting point is a technical innovation in how water itself is measured. Rather than relying on irrigation quotas or rainfall records, the researchers quantify effective crop water using actual evapotranspiration, or ETa, derived from the FAO’s WaPOR satellite database. Evapotranspiration captures the water a crop genuinely consumes—the sum of evaporation from soil and transpiration through leaves—making it a far more honest accounting unit than the water that is nominally delivered. Dekadal ETa values are summed to seasonal totals, converted into volumes using intersected administrative and crop-mask areas, and screened with quality flags so that low-confidence satellite pixels are masked rather than imputed. Every governorate-crop-year cell in the resulting panel carries provenance and quality-control documentation, a level of transparency that the authors treat as essential for any policy-facing conclusion.

But measuring water is only half the problem. The other half is causal: how much extra yield does an additional unit of effective crop water actually produce? Simply correlating water use with yields would be misleading, because farmers who irrigate more may differ systematically from those who do not—in soil quality, capital, or institutional access. To solve this, the researchers assembled a governorate-by-crop-by-year panel spanning 2010 to 2024 and identified yield responses using instruments that plausibly shift water availability without directly affecting yields: lagged rainfall anomalies aggregated over upstream basins, with pixels inside the target governorate excluded, and reservoir filling ratios matched to downstream irrigated perimeters through an assignment matrix built from scheme geography and operator records. The logic is elegant. Rain falling upstream of a governorate, and water stored in dams feeding its canals, change how much water crops can consume, but they do not directly change the agronomic or economic conditions inside the governorate itself.

On top of this causal backbone sits the study’s most distinctive component: a fuzzy inference layer that captures the state dependence of water’s marginal value. The idea is grounded in agronomic reality. The same proportional increase in effective crop water does not produce the same yield gain everywhere. Where irrigation delivery is already adequate and technology is efficient, an extra cubic metre buys relatively little; where water stress is high and a binding constraint exists, the same increment can be transformative. To formalise this, the authors built a fuzzy system with three antecedents—water stress, irrigation adequacy, and technology efficiency—each expressed through trapezoidal membership functions that allow a range of values to be unambiguously high or low rather than a single sharp threshold. Three linguistic terms per antecedent generate twenty-seven rules, each encoding an economically and agronomically plausible statement about when marginal returns to water should be larger or smaller. The full rule base, calibrated membership-function parameters, and alternative rule structures are published in the paper’s appendices, an unusually complete disclosure for a component that might otherwise be dismissed as researcher discretion.

The results upend the conventional ranking of where water should go. Average water productivity—the yield per unit of water that dominates most irrigation planning—would direct scarce water toward the regions and crops that already use it most efficiently. But when the state-dependent, causal marginal value of water is estimated instead, the ranking changes. A proportional increase in effective crop water yields different gains across Tunisia’s governorates, and accounting for this heterogeneity reverses some of the priorities implied by average productivity. In practical terms, the water that looks most productive on paper may not be the water that generates the most value at the margin, because productivity averages conceal the local hydrologic and technological states that determine what one more cubic metre is actually worth.

The framework’s second pillar connects agricultural water to human wellbeing. The authors construct governorate-level multidimensional deprivation indicators using household survey and regional accounts data from Tunisia’s National Institute of Statistics, drawing on a fuzzy measurement tradition in poverty analysis that Belhadj has developed in earlier work. They then compare water allocation scenarios against three criteria simultaneously: efficiency, in the total value generated; distribution, in how deprivation changes across territories; and institutional feasibility, scored by a second fuzzy system over five criteria including the metered delivery headroom that existing infrastructure can absorb. This triple test is what elevates the paper from an econometric exercise to a statement about territorial quality of life. Water allocation, the authors argue, should be treated as an environmental-social determinant of wellbeing—one link in a chain that runs from the canal to the household.

The headline finding concerns the trade-off between efficiency and equity. In the scenario accounting, a Rawlsian maximin allocation—one that prioritises the worst-off territories, in the spirit of philosopher John Rawls’s difference principle—retains 90.9 percent of the total value gain that a purely utilitarian allocation would achieve. When the comparison is restricted to crop groups whose elasticity estimates pass over-identification tests, the maximin allocation retains 97.0 percent of the utilitarian gain, while producing a larger reduction in mean deprivation. In other words, the equity-efficiency trade-off that policy makers usually assume to be painful may be remarkably mild: choosing the fairer allocation sacrifices almost nothing in total value while delivering substantially greater reductions in territorial deprivation. For water-stressed Mediterranean governments wrestling with allocation politics, that is a strikingly hopeful result.

The authors are notably careful about what their evidence can and cannot support. The Hansen over-identification test rejects for vegetables and high-value annuals at p equals 0.007, meaning the two instrument families do not identify a common effect for that group; the corresponding coefficient is therefore reported as a diagnostic rather than a validated causal estimate, and the Nabeul citrus value, which uses the same coefficient as a proxy, inherits that limitation. The transmission step from agricultural value added to deprivation is explicitly framed as accounting rather than causal inference: the income-deprivation gradient used to convert value gains into deprivation changes is an estimated cross-sectional association from a single household survey, applied uniformly across governorates because cell sizes do not support separate estimation. The paper claims only the productive and distributional stages of the water-to-wellbeing pathway, not wellbeing outcomes themselves—a restraint that strengthens rather than weakens its credibility.

The robustness apparatus is equally thorough. Sensitivity analyses vary the fuzzy specification, testing triangular and Gaussian membership functions against the trapezoidal baseline and alternative rule structures, and re-estimate the central results without the fuzzy layer entirely, showing that the substantive conclusions survive even when the discretionary component is removed. Leave-one-reservoir-out reconstructions test the storage instrument, alternative SPEI drought horizons test the climate inputs, and cells failing satellite coverage thresholds are masked rather than imputed so that data uncertainty remains traceable and separate from sampling uncertainty. Scenario capacities are checked against one-season metered delivery headroom records held by Tunisia’s regional agricultural development commissariats and agricultural development groups, with each pilot module capped at the minimum of a conservative 200,000 cubic metre ceiling and the actual infrastructure headroom—an upper bound on what existing systems could absorb without new construction.

What emerges is a template that extends well beyond Tunisia. The Mediterranean basin is among the world’s most water-stressed regions, and allocation decisions there are increasingly made under conditions of genuine scarcity, where every cubic metre redirected is a cubic metre taken from someone. By combining satellite-derived water accounting, credible causal identification, transparent state-dependent modelling, and multidimensional deprivation measurement, the framework offers a way to make those decisions with evidence about their distributional consequences rather than averages that hide them. The finding that a fairness-first allocation can retain nearly all of the efficiency gains—while cutting deprivation more deeply—suggests that the hardest choices in Mediterranean water policy may be less agonising than feared, provided the value of water is measured where it actually lands, in the states and places where crops and communities actually live.

Subject of Research: Causal, state-dependent estimation of the marginal value of irrigation water and its distributional effects on territorial quality of life in Tunisia

Article Title: From Crop Water Productivity to Territorial Quality of Life: A Causal, State-Dependent ETa Framework for Water Allocation in Mediterranean Agriculture

Article References: From Crop Water Productivity to Territorial Quality of Life: A Causal, State-Dependent ETa Framework for Water Allocation in Mediterranean Agriculture. (n.d.). https://doi.org/10.1007/s11205-026-03949-y

Image Credits: AI Generated

DOI: 10.1007/s11205-026-03949-y

Keywords: water allocation, evapotranspiration, Tunisia, Mediterranean agriculture, causal inference, instrumental variables, fuzzy logic, multidimensional deprivation, water productivity, quality of life, irrigation policy, satellite data