Reconstruction of past conditions provides important information on how ecosystems have been impacted by climate change, but generally for microhabitats worldwide there are no long‐term empirical measurements. In these cases, there has been protracted debate about how various large‐scale environmental proxies can best be used to reconstruct local temperatures. Here we help resolve this debate by examining how well environmental proxies hindcast sand temperatures at nest depths for five sea turtle nesting sites across the world. We link instrumental air temperature and sea surface temperature records with empirical sand temperature observations in the Atlantic (Ascension Island and Cape Verde), the Indian Ocean (Chagos Archipelago), the Caribbean (St Eustatius) and the Pacific (French Polynesia). We found strong correlations between sea surface temperatures, air temperatures and sand temperatures at all our study sites. Furthermore, Granger causality testing shows variations in sea surface temperature and air temperature precede variations in sand temperatures. We found that different proxies (air or sea temperature or a combination of both) predicted mean monthly sand temperatures within <0.5°C of empirical observations. Reconstructions of sand temperatures over the last 170 years reveal a slight warming of temperatures (maximum 0.5°C per century). An analysis of 36 published datasets revealed that the gradient of the relationship between sand temperature and air temperature is relatively constant, suggesting long‐term changes in sand temperature could be extended around the world to include nesting sites where there are no empirical measurements of sand temperature. Our approaches are likely to have utility for a range of microhabitats where there is an interest in long‐term changes in temperature.