Precipitation signal in pollen rain from tropical forests, South India

We have analyzed the pollen content of 51 surface soil samples collected in tropical evergreen and deciduous forests from the Western Ghats of South India sampled along a west to east gradient of decreasing rainfall (between 11 degrees 30-13 degrees 20'N and 75 degrees 30-76 degrees 30'E). Values of mean annual precipitation (Pann, mm/yr) have been calculated at each of the 51 sampling sites from a great number of meteorological stations in South India, using a method of data interpolation based on artificial neural network. Interpolated values at the pollen sites of Pann range from 1200 to 5555mm/yr, while mean temperature of the coldest month (MTCO) remains >15 degrees C and humidity factor (AET/PET, the actual evapotranspiration to potential evapotranspiration ratio) remains also included between 65 and 72%.Results are presented in the form of percentage pollen diagrams where samples are arranged according to increasing values of annual precipitation. They indicate that the climatic signal of rainfall is clearly evidenced by distinct pollen associations. Numerical analyses show that annual precipitation is an important parameter explaining the modern distribution of pollen taxa in this region. Pollen taxa markers of high rainfall (Pann >2500mm/yr) are Mallotus type, Elaeocarpus, Syzygium type, Olea dioica, Gnetum ula, and Hopea type, associated with Ixora type and Caryota. Pollen taxa markers of low rainfall (Pann <2500mm/yr) are Melastomataceae/Combretaceae, Maytenus type, Lagerstroemia and Grewia. The proportions of evergreen taxa and of arboreal taxa vary according to rainfall values. Indeed, when rainfall is <2500mm/yr, percentage of arboreal pollen (AP) is <50% and proportion of evergreen taxa is <20%. When rainfall exceeds 2500mm/yr, AP values average 70%, and proportion of evergreen taxa increases from 60 to 90%. Moreover, a good correlation between precipitation and proportion of evergreen taxa (0.85) presumes that precipitation can be estimated from pollen data.