The end of the summer monsoon season over the Indochina Peninsula (ICP), which extends from August to October, is preceded by a change in the zonal wind from westerly to easterly in September. Analysis of gauge-based precipitation shows that, over the past four decades, a positive trend in precipitation in this region is primarily due to increases in the later part of the monsoon season, with the highest wetting trend occurring in mid–September with a rate of 1.7 mm/(day·decade). This precipitation trend is correlated (95% significance level) with a trend in the date of the zonal wind direction reversal, which is becoming later. A comparison between the 1979-1988 and 2008-2017 means shows that the precipitation from September 11 to October 10 is 1.8 mm/day greater in the later decade, while the wind change reversal is 15.4 days later.
To understand the longer summer monsoon season over the ICP, we examine the reasons for and trends in the zonal wind reversal. Surface temperature from several observational datasets are examined, including monthly temperature from the GISS Surface Temperature Analysis, daily land temperatures from the Berkeley Earth Surface Temperature product, and daily SSTs from HadISST. JRA–55 and ERA–Interim reanalyses are used to evaluate the geopotential height and wind fields. A physical model for the zonal wind response to the meridional temperature gradient is derived based on the hypsometric and geostrophic wind equations. On average, the zonal wind changes by 2.25 m/s in association with a change in the meridional land/sea temperature difference of 1 K according to the physical model. We validate this relation by using daily data for September to October in two reanalyses. The results are 2.13 and 2.16 m/(s·K) for JRA–55 and ERA–Interim, respectively. If the meridional land-sea temperature gradient controls the zonal wind through the geostrophic balance, the trend of the monsoon retreat date will be related to the warming contrast between the ICP and the ocean to the south on the inter-annual time scale. Linear regression shows that the average response of the summer monsoon retreat date to land-sea temperature contrast is 23.5 day/K in both reanalyses (99% significance level).
The ability of five CMIP5 models (CCSM4, CNRM-CM5, GFDL-CM3, MIROC5 and MRI-CGCM3) to capture the relationships among the meridional land/sea temperature contrast, the zonal wind direction, and the monsoon retreat date is evaluated. Regressions between the daily zonal wind and the land-sea temperature difference during September and October show that four of the five models are able to simulate the relationship with coefficients around 2.25 m/(s·K). However, on the inter-annual time scale, none of the models reproduce the observed relationship between the monsoon retreat date and the land-sea temperature difference in full-forcing historical runs. This is also the case in versions of the models with prescribed SSTs (AMIP runs). This failure occurs because the models do not capture the stronger warming trends over the ICP compared with the adjacent ocean. Only CCSM4 produces a significant correlation between monsoon retreat date and land–sea temperature contrast in RCP8.5 runs.