Convective sources of trajectories traversing the tropical tropopause layer

Transit properties across the tropical tropopause layer are studied using extensive forward and backward Lagrangian diabatic trajectories between cloud tops and the reference surface 380aEuro-K. After dividing the tropical domain into 11 subregions according to the distribution of land and convection, we estimate the contribution of each region to the upward mass flux across the 380aEuro-K surface and to the vertical distribution of convective sources and transit times over the period 2005-2008. The good agreement between forward and backward statistics is the basis of the results presented here. It is found that about 85 % of the tropical parcels at 380 K originate from convective sources throughout the year. From November to April, the sources are dominated by the warm pool which accounts for up to 70 % of the upward flux. During boreal summer, the Asian monsoon region is the largest contributor with similar contributions from the maritime and continental parts of the region; however, the vertical distributions and transit times associated with these two subregions are very different. Convective sources are generally higher over the continental part of the Asian monsoon region, with shorter transit times. We estimate the monthly averaged upward mass flux on the 380aEuro-K surface and show that the contribution from convective outflow accounts for 80 % on average and explains most of its seasonal variations. The largest contributor to the convective flux is the South Asian Pacific region (warm pool) at 39 % throughout the year followed by oceanic regions surrounding continental Asia at 18 % and Africa at 10.8 %. Continental Asian lowlands account for 8 %. The Tibetan Plateau is a minor overall contributor (0.8 %), but transport from convective sources in this region is very efficient due to its central location beneath the Asian upper level anticyclone. The core results are robust to uncertainties in data and methods, but the vertical source distributions and transit times exhibit some sensitivity to the representations of cloud tops and heating rates. The main sensitivity is to the radiative heating rates which vary among reanalyses.