A Markov Chain-Monte Carlo framework with hybrid-order transitions for generating Typical Wind Year datasets across global tropical climates

Standard Typical Meteorological Year (TMY) datasets often fail to capture the complex stochastic dynamics of wind vectors, leading to significant uncertainties in building energy simulations for tropical climates. This study addresses the structural conflict between capturing multi-hour temporal persistence and maintaining computational stability by introducing a Hybrid-Order Markov Chain–Monte Carlo framework. Unlike static-order models, the developed framework employs an order-adaptive transition mechanism that dynamically prioritizes second-order memory while reverting to first-order robustness in sparse data regions, thereby resolving the memory-sparsity paradox in generating continuous 8760-h wind sequences. Validated across 27 globally distributed tropical cities, the framework demonstrates superior fidelity in reproducing joint speed-direction distributions (Jensen–Shannon divergence <0.052) and preserving directional inertia compared to conventional baselines. Crucially, building energy simulations reveal a systematic infiltration gap in current TMY3 datasets, which are shown to underestimate wind-driven infiltration loads in equatorial regions due to the neglect of high-frequency wind occurrences. By correcting these biases and preserving the coupled wind energy signature, the developed Typical Wind Year (TWY) provides a physically rigorous meteorological boundary condition for climate-responsive design and HVAC sizing in wind-sensitive tropical climates.