Robust statistical downscaling of rainfall and temperature for reduction of uncertainty in climate change impacts on river discharge
Hydrologic changes are the most significant potential impacts of global climate change. Assessment of possible impacts of climate change on hydrological processes at local or regional scale is considered as the most essential component in devising effective climate change adaptation policies. Howeve...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2016
|
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Hydrologic changes are the most significant potential impacts of global climate change. Assessment of possible impacts of climate change on hydrological processes at local or regional scale is considered as the most essential component in devising effective climate change adaptation policies. However, addressing the uncertainties in future climate projections is the key challenge in climate change impact assessment for effective adaptation planning. The aim of this study is to develop a framework for robust statistical downscaling of climate to assess climate change impacts on river discharge with associated uncertainties. The effectiveness of the developed framework is evaluated through its application in the east coast of peninsular Malaysia, which is considered as the most vulnerable region in peninsular Malaysia to climate change. Eight linear and non-linear transfer function models are used for statistical downscaling of coarse resolution general circulation model (GCM) projections in daily and monthly time scales. A comprehensive framework is also developed for qualitative and quantitative assessment of uncertainty in downscaled climate using various robust statistical approaches. The most suitable models are applied to downscale rainfall and temperature from atmospheric parameters projected by Hadley Centre GCM under A2 and B2 scenarios. The best rainfall-runoff model selected from a set of one lumped conceptual and two state-of-art data driven models is used for future projection of streamflow in seven catchments within study area under different climate change scenarios. The results show that non-linear models based on random forest (RF) and support vector machine (SVM) perform better in downscaling daily climate, whereas linear regression models perform better in downscaling monthly climate. A hydrological model based on Group Methods of Data Handling (GMDH) is found to be the most reliable for assessment of climate change impacts on river flow. The study reveals that climate projections vary considerably with the choice of predictor selection approach, downscaling model, and type and scale of climatic variable being downscaled. Weighted multi-model mean (MM) of downscaled climate reveals 2-3◦ C increase in temperature and 40% increase in rainfall at the end of century in most parts of the region. Projections of streamflow using MM climate reveal river discharge in large catchments in the region will be more variable, particularly during northeast monsoon (NE) season which is likely to cause more floods. However, high uncertainty in climate projections during NE monsoon at later part of the century indicates that the forecasting of more frequent flood, particularly in end of the century is highly uncertain. Analyses of results reveal that robust downscaling of climate and rigorous assessment of model performance can substantially reduce uncertainties arising in different steps of climate downscaling and impact assessment. However, climate projections and impacts on river flow still may vary considerably due to downscaling model and scale of data used. The study suggests that streamflow projections using downscaled climate should be interpreted cautiously considering the uncertainty in projections. |
|---|