The purpose of this study was to investigate which climate/heat indices perform best in predicting heat-induced loss of physical work capacity (PWC-loss). Integrating data from earlier studies, data from 982 exposures (75 conditions) exercising at a fixed cardiovascular load of 130b.min^-1, in varying temperatures (15-50°C), humidities (20-80%), solar radiation (0-800W.m^-2), wind (0.2-3.5m.s^-1) and two clothing levels, were used to model the predictive power of ambient temperature, Universal Thermal Climate Index (UTCI), Wet Bulb Globe Temperature (WBGT), Modified Equivalent Temperature (mPET), Heat Index, Apparent Temperature (AT), and Wet Bulb Temperature (T_wb) for the calculation of PWC-loss, skin temperature (T_skin) and core-to-skin temperature gradient, and Thermal perception( TSV) in the heat. R², RMSD and Akaike stats were used indicating model performance. Indices not including wind/radiation in their calculation (T_a, Heat Index, AT, T_wb) struggled to provide consistent predictions across variables. For PWC-loss and TSV, UTCI and WBGT had the highest predictive power. For T_skin, and core-to-skin temperature gradient, the physiological models UTCI and mPET worked best in semi-nude conditions, but clothed, AT, WBGT and UTCI worked best. For all index predictions, T_a, vapor pressure and T_wb were shown to be the worst heat strain predictors. While UTCI and WBGT had similar model performance using the full dataset, WBGT did not work appropriately in windy, hot-dry, conditions where WBGT predicted lower strain due to wind, whereas the empirical data, UTCI and mPET indicated that wind in fact increased the overall level of thermal strain. The findings of the current study highlight the advantages of using a physiological model-based index like UTCI when evaluating heat stress in dynamic thermal environments.