Vegetation Pattern Fields as Constraint Traces A Seed-Based Framework for Archaeological Prospection

Positioning Note This paper introduces a conceptual and methodological framework rather than a new sensing technique or detection algorithm. Its purpose is to reframe how vegetation patterns observed through remote sensing are interpreted in archaeological prospection. Conventional approaches treat such patterns as anomalies to be identified and classified. In contrast, this work treats them as partial surface expressions of underlying constraint fields and formulates inference as a constrained reconstruction problem: identifying the minimal set of environmental and anthropogenic constraints sufficient to generate the observed pattern structure. The framework is intentionally conservative. It: • does not assume unique solutions • does not claim direct recovery of past activity • does not replace existing sensing modalities Instead, it provides a disciplined inferential pipeline that explicitly separates observation, reconstruction, and validation, with formal controls on admissibility, minimality, and equifinality. The contribution of this paper is therefore not empirical discovery, but a structured method of inference: a constraint-first, seed-based approach for generating, evaluating, and pruning candidate explanations of vegetation pattern fields under bounded ecological dynamics. This framework intersects with existing work in archaeological prospection, spatial pattern analysis, and model-based inference. It relates to prior approaches in crop mark interpretation and landscape archaeology, while extending concepts of equifinality and inverse reasoning by introducing explicit minimality constraints and reconstruction-based evaluation. It also draws on methods from spatial statistics and ecological modeling, particularly in the use of structural invariants and generative simulation for pattern interpretation. *This framework is an application of the constraint-first architecture described in prior work, specifically corresponding to System B (the pruning and evaluation system) within a dual-system structure. In that architecture: • System A generates candidate structures • System B evaluates and prunes them under host-domain constraints The present work operates entirely within System B. It does not generate hypotheses about archaeological features, but instead: • evaluates candidate constraint fields, • enforces minimality, • and prunes non-admissible reconstructions through simulation. The seed-based reconstruction process described here corresponds directly to the REM-Evo stage: • candidate constraint sets are generated, • subjected to falsifiers, • and pruned based on invariant reproduction and domain admissibility. The reconstruction–pruning loop can also be interpreted as a domain-specific instance of the Epistemic Collider: • candidate constraint models are “collided” with ecological dynamics, • invalid models are destroyed, • and surviving structure is retained as minimal seeds. This paper therefore does not introduce a standalone method, but a domain instantiation of a broader constraint-governed inference architecture. ——— Abstract Archaeological prospection using remote sensing is typically framed as the detection of spectral, thermal, or spatial anomalies. This paper reframes that task as a constraint-trace problem, in which vegetation patterns are interpreted not as isolated indicators, but as surface expressions of underlying environmental and anthropogenic constraint fields. A seed-based, constraint-first framework is introduced to infer