Abstract A number of analytical models have been presented in the contemporary literature to describe and predict the thermal behavior of salt-gradient solar ponds under steady- and unsteady-state conditions. This paper presents a detailed theoretical comparison between three different analytical models proposed to simulate the thermal behavior of solar ponds. These models are slightly modified to represent the gel pond configuration. For experimental comparison, a gel pond constructed at the University of New Mexico, which has been in operation for several years, has been used as the reference pond. The gel pond differs from conventional salt-gradient ponds in that the nonconvective insulating layer is replaced by a transparent polymer gel. Numerical computations have been made to optimize the geometric and operational parameters of the pond using the three different models. The optimum thickness of the nonconvective layer (gel) and the thin upper convective layer (fresh water) and their effects on pond performance at a given ambient temperature, isolation and storage temperature have been calculated using all the models. The three models have also been used to calculate the absorptivity— transmissivity product (ατ), a parameter which represents the transparency of the nonconvective and the upper convective layers combined together. These calculated values have been compared with experimental values as measured through an actual gel layer to test the accuracy of the different models as applied to the gel pond. The results show that, under the same temperature difference between the storage zone and the ambient (20°C) and a yearly average insolation of 250 W/m 2, the model proposed by Wang and Akbarzadeh predicts an efficiency of approx. 32% as compared to the high values of 37.2 and 39% predicted by the Kooi model and Kaushik and Bansal model, respectively. Under the same conditions, the optimum gel thicknesses predicted by the three models are 62, 55, and 75 cm, respectively.