Intercomparisons of 10 heat stress metrics within a global land surface model

Heat stress occurs when the human body loses the ability to internally regulate heat balance, and an increase of internal temperatures of ~3ºC can be lethal. Many different heat diagnostic indices have been developed to diagnose heat stress, and policy makers have made decisions to incorporate these indices in weather warning systems. Systematic evaluation of multiple heat stress metrics is, however, a difficult task. Furthermore, heat stress metrics are designed with differing philosophical approaches: comfort, empirical, and physiological responses. We present results from model–model comparisons of heat stress metrics that cover a wide spectrum of thermal approaches and assumptions, using the National Center for Atmospheric Research (NCAR) Community Land Model 4.5 (CLM4.5). CLM4.5 is forced with a variety of reanalysis data products (CRUNCEP, MERRA, etc.) and the NCAR Community Atmospheric Model (CAM4 and CAM5), to characterize uncertainty in present-day heat stress. CLM4.5 has a full range of Earth climate, from urban city representations, to high altitude mountain environments. 10 different heat stress metrics are implemented directly into CLM4.5 and are calculated at every time step. We show that the responses to heat stress have low variability between model runs, even out to the 99th Percentile (hottest 3-4 days of a year). The covariance of relative humidity and temperature has a limited parameter space that is due to model balances from conservation of mass, the equation of state, and entropy. Analyzing the joint distributions between temperature and heat stress metrics, temperature alone is a poor indicator of heat stress. We show that all heat stress metrics reach a maximum within regions of high convective activity (monsoonal regions). Additionally, this corresponds to maximum wet bulb temperatures, however, in desert regions, wet bulb temperatures fail as an adequate index. Comfort based algorithms rate desert regions as ‘warmer' than their empirically derived counterparts. Overall, CLM4.5 when forced with a large range of reanalysis products and model data sets has a small range of variability in heat stress metrics. Regional climate plays an important factor in applications of heat stress metrics.