Distributionally-Shifted Soil-Moisture Imputation for Computer Agriculture Irrigation Inference Under Sensor Network Attrition: An Artificial Intelligence Missing-Not-At-Random Regret Bound

Soil-moisture sensor networks in irrigated agriculture lose nodes to battery depletion, rodent damage and radio dropout, and the resulting missingness is rarely missing-at-random: probes fail preferentially in the wetter and more saline zones that matter most for irrigation decisions. The framework couples a covariate-shift-corrected conditional imputer with a distributionally robust irrigation policy, and the bound decomposes regret into an imputation error term, a shift term and a policy-optimism term. A replication-based evaluation of the before-after-control-impact estimator was conducted under a declared variance structure with between-cluster SD 6.0, residual SD 4.5 and intraclass correlation 0.64, across 2000 replications per design cell for cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points. The evaluation indicates that 56 clusters are required for 80% power at a 5 percentage-point effect. The contribution is a bound that makes the attrition mechanism, rather than the imputation algorithm alone, the object of design.