An assessment of mean annual precipitation in Rajasthan, India needed to maintain 
Mid-Holocene lakes

Emily C. Gill, Balaji Rajagopalan1, and Peter H. Molnar

Paleo-climate literature reports evidence of freshwater lakes over Rajasthan, in 
northwestern India, during the mid-Holocene (~6ka), where desert conditions prevail in 
present time.  It’s suggested that mid-Holocene temperatures were warmer, precipitation was 
nearly double current levels, and there was an enhanced La Niña-like state.  While previous 
analyses infer the lakes were sustained by generally high precipitation and low evaporation, 
we provide a systematic analysis on the relationship between the relevant energy budget 
quantities. 
We have built a hydrological lake model to reconstruct lake levels throughout the Holocene.  
Model output is evaporation from the lake.  Inputs are precipitation over the lake and 
catchment runoff, determined using precipitation, Preistley-Taylor evapotranspiration, 
interception and infiltration. 
Initial tests of the model have been completed with current climate conditions to ensure 
accurate behavior.  Contemporary runs used station precipitation and temperature data for 
the region surrounding Lake Didwana, LD (27N 74E).  Digital elevation maps were used to 
compile LD lake bathymetry.  Under current climate conditions, a full LD (~ 9 m) empties 
over the first several years.  While lake depth varies yearly, increasing with each monsoon 
season, variations following the initial decline are minimal (~ ± 1.0 m). 
We ran the model with a 2000-year sequence of precipitation and temperature generated by 
resampling the observed weather sequences, with a suite of base line fractions of vegetation 
cover and increased precipitation, with mid-Holocene solar insolation. Initial runs revealed 
that precipitation amount and percent of vegetated catchment area influence lake levels, but 
insolation alone does not.  Incrementally changing precipitation (between current levels and 
a 75% increase) and percent of vegetated area (between 10-90%) reveals that a 50% increase 
in precipitation alone is not enough to reach the maximum lake levels reported by Enzel et 
al. [1999] of 7m during the mid-Hoocene.  For LD to reach maximum levels, both at least 50% 
more precipitation than today and a vegetated fraction of the catchment of at least 50% is 
required, but if precipitation were twice that today, and vegetation covered 50% of the 
area, the lake would have been deeper than 9 m. 
Future work involves generating P and temperature series for 2000-year long sequences 
representing the early-, mid-, and late-Holocene using two approaches: k-nearest neighbor 
and generalized linear model.  Using these, we’ll run the lake model to determine what 
combinations of P, E, and other variables are necessary to sustain the lakes. While model 
runs suggest that monsoon rainfall should increase in a warming world, observations show we 
are currently in the longest epoch of below-normal south-Asian monsoonal rainfall.  By using 
the mid-Holocene as an analog for a future warming world, this study could expand the 
understanding of the south-Asian monsoon’s potential response to warming.