EVALUATING CLIMATE ADAPTATION IN COMPLEX SYSTEMS

Climate adaptation programmes are expanding rapidly, yet the methodological foundations for evaluating them remain unsettled. The problem is not one of missing metrics. It is one of misaligned reasoning: evaluative frameworks designed for stable, bounded interventions are routinely applied to programmes that operate through long, contingent causal chains, across heterogeneous contexts, and within institutional systems whose quality, incentives, and politics shape outcomes as powerfully as any technical design choice. This paper advances three related arguments. First, that climate adaptation cannot be evaluated adequately through performance-measurement logic alone; it requires a theory-based, contribution-oriented, and mixed-methods architecture capable of reasoning about complex causal pathways under conditions of genuine uncertainty. Second, that many adaptation programmes are evaluability-constrained before they are evaluated — poorly theorised, poorly baselined, and poorly evidenced by design — and that evaluators must treat this constraint as a methodological diagnosis rather than a residual weakness. Third, that Bayesian reasoning and Bayesian networks can serve as powerful advanced tools for integrating heterogeneous evidence and formalising confidence judgments, but only as disciplined components within a broader theory-based design rather than as substitutes for it. The framework developed here is structured around eight linked tasks: theory of change reconstruction; contextual and system mapping; layered results differentiation; evidence architecture design; method delineation by inferential role; portfolio synthesis; causal integration through contribution analysis; and explicit uncertainty management including Bayesian updating. It also treats representation, epistemic power, and distributional equity as constitutive of evaluative rigour rather than as sensitivity overlays. An illustrative evaluation matrix operationalises the framework across six evaluation dimensions. The paper concludes with implications for evaluation offices, funders, and programme managers — and argues that stronger adaptation evaluation depends as much on institutional architecture as on methodological sophistication.