Causal-Forest Inference for Computer Agriculture Cover-Crop Termination Timing Under Unobserved Residue Decomposition: An Artificial Intelligence Heterogeneous-Treatment Effect Bound

Cover-crop termination timing is a management decision whose agronomic and environmental consequences depend on residue decomposition, a process that is rarely observed at the resolution required for causal inference. The analysis is theoretical and design-based: no observational data are used or claimed. Under a declared variance structure with between-cluster standard deviation 6.0, residual standard deviation 4.5 and intraclass correlation 0.64, power was estimated across 2000 design cells for cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points. The design required 54 clusters to detect a 5-percentage-point effect at 80% power. The paper derives the bound, states the assumptions under which it is identified, and reports the design's operating characteristics. The contribution is a bound that remains valid when the decomposition moderator is unobserved, together with a transparent account of the sample sizes required to estimate it.