Attribution of heritage deterioration risks at UNESCO Cultural World Heritage Sites

Dataset description This dataset provides the results and Python code used to assess how anthropogenic climate change has shifted cultural heritage risk indicators at UNESCO World Heritage Sites. The analysis compares factual all-forcing climate experiment (ALL) with counterfactual natural-only experiment (NAT) and calculates both intensity and probability attribution for five heritage risk indicators: Scheffer Index, High Humidity Days, Salt Transitions, Frost Intensity, and Growing Season Length. Files included: CulturalWHS_RiskIndicators.csv: A master results table for 1,049 UNESCO World Heritage Site entries that inscribed with cultural values (cultural and mixed WHSs). For each entry, the table includes site location, value criteria, UNESCO World Heritage region, and calculated heritage risk indicator results. The results include preindustrial values, recent-past (2000-2014) values under natural-only and factual climate simulations, anthropogenic intensity change, preindustrial 95th percentile thresholds, recent-past exceedance probabilities, Fraction Attributable Risk, and Attributable Risk Reduction. CulturalWHS_RiskIndicators_CODEBOOK.csv: A corresponding codebook defining the column headers in the master results table (CulturalWHS_RiskIndicators.csv). Attribution of Heritage Risk Indicators.ipynb — A Python notebook illustrating the full workflow used in the study. It loads MPI-GE CMIP6 climate variables, reconstructs relative humidity (RH) from air temperature and dew-point temperature, validates reconstructed RH, calculates the five heritage risk indicators, maps values at World Heritage Site locations, calculates intensity and probability attribution metrics, creates regional summaries, exports the final tables, and reproduces the main figures. Data acquisition The calculations use external datasets: the MPI-GE CMIP6 climate ensemble and the UNESCO World Heritage Site list. The climate data were obtained from the MPI-GE CMIP6 archive through the Earth System Grid Federation platform: https://aims2.llnl.gov/search/cmip6/. The analysis uses daily mean air temperature (tas_day), daily precipitation (pr_day), and daily dew-point temperature (tdps_day) for the historical, hist-nat, and SSP2-4.5 experiments. Native MPI-GE historical relative humidity (hurs_day) is used only to validate the reconstructed relative humidity. Data acquired in December 2025. The World Heritage Site list and attributes were obtained from the UNESCO World Heritage List: https://whc.unesco.org/en/list/xlsx/. Data acquired in December 2025. Computation and visualisation All code is written in Python and uses xarray, numpy, pandas, geopandas, scipy, and matplotlib. The notebook first loads MPI-GE CMIP6 daily climate variables for historical, hist-nat, and SSP2-4.5 simulations while preserving the ensemble dimension. We used the first 30 ensemble members from MPI-GE CMIP6 (r1 to r30). Relative humidity is reconstructed from daily mean air temperature and dew-point temperature using the Magnus formula, then checked against native MPI-GE historical relative humidity. Five heritage risk indicators are calculated annually. Indicator values are extracted at World Heritage Site locations using the nearest native MPI-GE grid cell. Intensity attribution is calculated by comparing recent-past factual all-forcing conditions with recent-past natural-only conditions. Probability attribution is calculated by defining extreme heritage risk conditions using preindustrial 95th percentile thresholds and estimating recent-past exceedance probabilities under factual and natural-only simulations. Fraction Attributable Risk is reported where anthropogenic forcing increases exceedance probability, while Attributable Risk Reduction is reported for Frost Intensity where anthropogenic forcing reduces exceedance probability. Result dataset The main result dataset is stored in CulturalWHS_RiskIndicators.csv. This table contains the full list of World Heritage Site entries used in the analysis, together with World Heritage Sites metadata and calculated climate attribution results for the five heritage risk indicators. The CulturalWHS_RiskIndicators_CODEBOOK.csv defines each column or column-name pattern in the master table.