Abstract
Standard linear regression methods frequently fail to detect complex spatial patterns in agricultural economic data, leading to incomplete assessments of geographic variation in resource prices. This study addresses this methodological gap by analyzing spatial and temporal dynamics of irrigation water pricing across 16 Queensland water supply schemes over eleven years (2015–2025), employing three complementary analytical approaches applied to panel data comprising 176 observations. First, a fixed effects panel model identified a structural break in 2019, with real water prices rising 115%–164% above 2015 baseline levels through to 2024. Second, Generalized Additive Models (GAMs) revealed highly significant non-linear spatial and temporal price patterns, explaining 40% of price variation (adjusted R² = 0.40; deviance explained = 42.7%), nearly three times the explanatory power of linear specifications (R² = 0.15). The spatial smooth required an effective degree of freedom of 7.591, characterising a geographically complex price gradient across Queensland's north-south latitude range that a linear specification would fail to detect. Third, cross-sectional log-linear regression identified substantial sectoral and regional price variation (adjusted R² = 0.527), with cotton-dominant schemes recording prices 258% above non-specialized baseline schemes, horticulture schemes recording an 81% premium, and the Burdekin region recording a 63% discount. GAM robustness checks confirmed that this spatial complexity reflects genuine geographic structure rather than model overfitting (ΔAIC = +60.52; leave-one-scheme-out instability ratio = 0.138). These findings have direct implications for water pricing policy, market regulation, and cost recovery frameworks in Queensland and other jurisdictions with geographically segmented irrigation markets.