Benchmarking Evaluation of Graph-Neural-Network Soil-Moisture Imputation for Computer Agriculture Irrigation Advisories Under Sparse Sensor Networks: An Artificial Intelligence Error-Propagation Bound for Water-Use Efficiency

Sparse soil-moisture sensor networks force irrigation advisories to rely on imputed values whose errors propagate into water-use decisions. The framework is evaluated through a replication-based design study of the before-after-control-impact estimator under a declared variance structure (between-cluster SD = 6.0, residual SD = 4.5, intraclass correlation = 0.64), with power estimated across 2000 design cells.54 clusters are required to detect a 5-percentage-point effect at 80% power, and that the intraclass correlation of 0.64 substantially inflates the cluster requirement relative to independent sampling. The derived bound indicates that advisory water-use efficiency is concave in imputation accuracy, so that marginal improvements in GNN imputation yield diminishing returns once the error falls below a threshold set by the between-cluster variance. The contribution is a formal bridge between imputation benchmarking and irrigation decision safety, together with a design table that practitioners can use to size sparse sensor deployments.