Trend analysis of statistically downscaled precipitation for tropical
semi-arid climate
Abstract
The coarse resolution climatic data extracted from the global climate
models (GCM) cannot be utilised straightaway for research works on
climate change impact analysis. Thus, the downscaling technique is used
to attain a higher resolution scenario from the GCM. In the current
study, the Canadian Centre for Climate Modelling and Analysis
(CCCma)-GCM is used to predict monthly precipitation using the
statistical downscaling technique for the tropical semi-arid region of
Eastern Gujarat for the period 2019-2099. Geo-potential height (h500)
and mean sea level pressure (MSLP) are chosen as explanatory variables
for the downscaling model. The model is developed using the principal
component analysis (PCA) - multiple linear regression (MLR) combined
approach. The model is applied to predict rainfall for three
representative concentration pathway (RCP) scenarios, RCP2.6, RCP 4.5,
and RCP 8.5. The best-suited scenario for the study area is selected
using robust statistical indicators such as the Nash-Sutcliffe
efficiency and the root mean square error. Several parametric and
nonparametric tests are conducted to analyse rainfall trends in the
Eastern Gujarat region. The outcomes showed that the precipitation
scenario produced for RCP4.5 replicates the climatology of the region
suitably. The trend investigation of the predicted rainfall showed that
the significance of the seasonal trend is independent of the
significance of monthly trends. Trend analysis of downscaled
precipitation series can report multiple change points for a region.
Further, the annual rainfall increases tremendously in the tropical
semi-arid regions over the 21st century. This study will provide
insights into sustainable water resource management and development.