Geochemical evidence for the use of magnetic susceptibility as a paleorainfall proxy in the tropics

Abstract

Several workers have estimated paleomonsoonal rainfall using proxies that are related to wind strength, upwelling, productivity, chemical weathering and pedogenesis. In an earlier work on sediments of Thimmannanayakanakere (TK) — a lake in Southern India [Shankar, R., Prabhu, C.N., Warrier, A.K., Vijaya Kumar, G.T., Sekar, B., 2006. A multidecadal rock magnetic record of monsoonal variations during the past 3700 years from a tropical Indian tank. Journal of the Geological Society of India 68, 447–459], we suggested that magnetic susceptibility ( χ lf) may be used as a proxy to estimate the paleomonsoonal rainfall in tropical regions. This was based on the positive correlation between magnetic susceptibility of TK sediments and instrumental rainfall data and backed up by historical records and paleoclimate proxy data from widely separated regions of India. In this paper, we provide geochemical data (results of CBD extraction and elemental analysis) and their statistical analysis (PCA) to support our earlier proposition that χ lf may be used as a proxy for paleomonsoonal rainfall in the tropical regime. The CBD treatment of sediment samples reduced the average χ lf from 132 to 64 × 10 − 8 m 3 kg − 1 , average χ fd from 12.91 to 2.01 × 10 − 8 m 3 kg − 1 and average χ fd (%) from 8.9 to 3%, showing that the samples do contain pedogenic magnetite. Further, samples of high- (low-) rainfall periods showed a higher (lower) reduction in the magnetic parameters, suggesting a higher (lower) content of pedogenic magnetite. Metal/Al ratios (Fe/Al, Ti/Al, Na/Al, Mn/Al, Pb/Al, Zn/Al, Cu/Al, K/Al, Ca/Al, Mg/Al and Sr/Al) are positively correlated (most of them at the 1% level) with χ lf, suggesting that the TK sediments are derived mainly from the catchment. In particular, Ti/Al, K/Al and K/Na ratios that are good indicators of chemical weathering intensity (CWI), start increasing from ~ 2 cal. ka B.P. onwards, corresponding to increasing χ lf (= increasing rainfall) from this time. As chemical weathering is principally driven by rainfall (temperature being constant) in tropical regions, data on CWI indicators provide support to the hypothesis that χ lf estimates paleomonsoonal rainfall. Principal component analysis of geochemical and magnetic data shows that three important factors explain the variance in the data set: terrigenous, pedogenic (= rainfall) and carbonate (= low rainfall or drought). Thus, our data strengthen Shankar et al.'s [Shankar, R., Prabhu, C.N., Warrier, A.K., Vijaya Kumar, G.T., Sekar, B., 2006. A multidecadal rock magnetic record of monsoonal variations during the past 3700 years from a tropical Indian tank. Journal of the Geological Society of India 68, 447–459] hypothesis that magnetic susceptibility may be used as a proxy for monsoonal rainfall in the tropical milieu. We have proposed a model to explain pedogenesis, chemical weathering intensity and carbonate precipitation, which are all related to rainfall in the TK catchment.