Abstract
The aim of this study was to validate evaporation models that can be used for palaeo‐reconstructions of large lake water levels. Lake Titicaca, located in a high‐altitude semi‐arid tropical area in the northern Andean Altiplano, was the object of this case study. As annual evaporation is about 90% of lake output, the lake water balance depends heavily on the yearly and monthly evaporation flux. At the interannual scale, evaporation estimation presents great variability, ranging from 1350 to 1900 mm year^−1^. It has been found that evaporation is closely related to lake rainfall by a decreasing relationship integrating the implicit effect of nebulosity and humidity. At the seasonal scale, two monthly evaporation data sets were used: pan observations and estimations derived from the lake energy budget. Comparison between these data sets shows that (i) there is one maximum per year for pan evaporation and two maxima per year for lake evaporation, and (ii) pan evaporation is greater than lake evaporation by about 100 mm year^−1^. These differences, mainly due to a water depth scale factor, have been simulated with a simple thermal model θ~w~(h, t) of a free‐surface water column. This shows that pan evaporation (h = 0·20 m) is strongly correlated with direct solar radiation, whereas the additional maximum of lake evaporation (h = 40 m) is related to the heat restitution towards the atmosphere from the water body at the end of summer. Finally, five monthly evaporation models were tested in order to obtain the optimal efficiency/complexity ratio. When the forcing variables are limited to those that are most readily available in the past, i.e. air temperature and solar radiation, the best results are obtained with the radiative Abtew model (r = 0·70) and with the Makkink radiative/air temperature model (r = 0·67). Copyright © 2007 John Wiley & Sons, Ltd.