Statistical Downscaling Methods


Statistical downscaling methods can be adopted with reasonable confidence as downscaling tools to undertake climate change impact assessment studies for the future. Many studies have documented biases in statistical downscaling methods, including their projections of meteorological variables (rainfall, temperature, and evaporation) corresponding to the climate scenarios. These biases tell you about how accurately and reliable methods in projecting the meteorological variables. In some regions, the performance of the methods can be different. I conduct several assessments in Malaysia using the methods to test the suitability and validity of the methods to be applied in Malaysia.


I applied two statistical downscaling models (SDM), namely Statistical DownScaling Model (SDSM) and Long Ashton Research Station Weather Generator (LARS-WG) to downscale the General Circulation Models (GCMs) output for possible future values of local meteorological variables such as rainfall and temperature. Most of my studies focus on the region of Peninsular Malaysia. I found both models are adequate in the model performance  (Fig. 1) in capturing the present meteorological variables during the calibration and validation period.

model performance

Figure 1: SDSM vs LARS-WG during validation periods (Hassan et al., 2014)

As detailed in Hassan et al. (2014), the trend and pattern of future meteorological variables downscale by the SDSM and LARS-WG are not similar. It is not clear which models give a meaningful result for future scenarios and which results are considered to be more reliable. The example of the general trend of the annual rainfall in the local region (Kurau River catchment) (Hassan et al., 2015) projected by the SDSM model can be illustrated in Figure 2.

rainfall trend_SDSM projected

Figure 2: SDSM vs LARS-WG during validation periods (Hassan et al., 2014)

Challenges and Opportunities

The development and studies in the climate are rapid focused. Many recent models and tools are developed to downscale the GCMs output and those models are claimed to describe well the trend of the meteorological variables corresponding to climate scenarios, as compared to the previous models. These recent advance tell you little about how accurately the model will simulate the response to different boundary conditions.

The dynamical downscaling approaches (DDA) also interested to focus.The progress of DDA involves expensive computational facilities and exclusively for specialized climate research institutions. However, recent support has been introduced to access the DDA for open public. These situations can give an advantage to academicians such as me, to utilise the DDA data.


Hassan, Z., Shamsudin, S., & Harun, S. (2014). Application of SDSM and LARS-WG for simulating and downscaling of rainfall and temperature. Theoretical and Applied Climatology, 116(1-2), 243–257. doi:10.1007/s00704-013-0951-8

Hassan, Z., Shamsudin, S., Harun, S., Malek, M. A., & Hamidon, N. (2015). Suitability of ANN applied as a hydrological model coupled with statistical downscaling model: a case study in the northern area of Peninsular Malaysia. Environmental Earth Sciences, 74(1), 463–477. doi:10.1007/s12665-015-4054-y

Last Update 14th 1 2016

Note: Feeling interested to do a research on the climate change impact (especially on the hydrological response), may contact me through email (