Reconstructing past precipitation from lake levels and inverse modelling for Andean Lake La Cocha

Knowledge of paleoclimates and past climate change is important to put recent and future climate change in perspective. In the absence of well-developed methodology to reconstruct paleoprecipitation the majority of climate reconstructions focus on temperature, whereas precipitation is an equally important climate parameter. This paper explores the possibility of inferring paleoprecipitation from lake-level records by inverse hydrological modelling. Pollen spectra of a lacustrine sediment core were used to infer changes in past temperatures and lake levels during the past 14,000 years. A hydrological model that calculates lake levels using meteorological parameters and a digital terrain model were developed for the catchment area of Lake La Cocha. After calibration the model accurately simulated modern lake levels. A sensitivity analysis shows that the model results are most sensitive to temperature and precipitation. This hydrological model was subsequently used to estimate mean annual precipitation needed to reproduce the pollen-based reconstructed lake levels (inverse modelling). The lake currently discharges through the permanent Guamués River, with a modelled mean annual discharge of 3.6 m³ s^-1. However, past lake levels and hydrological modelling results suggest that Lake La Cocha has been free of discharge during most of the Holocene, and after an intermittent phase only recently started discharging permanently. The uncertainty in the inferred precipitation during the discharge-free period is estimated at ~22 mm. Quasi stable lake levels seem to justify using equilibrium conditions when reconstructing precipitation. Early Holocene lake levels were ~10 m lower than modern values, implying that precipitation must have been 30-40 % less than today.