Coupling a Markov Chain and Support Vector Machine for At-Site Downscaling of Daily Precipitation

Abstract

Statistical downscaling is useful for managing scale and resolution problems in outputs from global climate models (GCMs) for climate change impact studies. To improve downscaling of precipitation occurrence, this study proposes a revised regression-based statistical downscaling method that couples a support vector classifier (SVC) and first-order two-state Markov chain to generate the occurrence and a support vector regression (SVR) to simulate the amount. The proposed method is compared to the Statistical Downscaling Model (SDSM) for reproducing the temporal and quantitative distribution of observed precipitation using 10 meteorological indicators. Two types of calibration and validation methods were compared. The first method used sequential split sampling of calibration and validation periods, while the second used odd years for calibration and even years for validation. The proposed coupled approach outperformed the other methods in downscaling daily precipitation in all study periods using both calibration methods. Using odd years for calibration and even years for validation can reduce the influence of possible climate change-induced nonstationary data series. The study shows that it is necessary to combine different types of precipitation state classifiers with a method of regression or distribution to improve the performance of traditional statistical downscaling. These methods were applied to simulate future precipitation change from 2031 to 2100 with the CMIP5 climate variables. The results indicated increasing tendencies in both mean and maximum future precipitation predicted using all the downscaling methods evaluated. However, the proposed method is an atsite statistical downscaling method, and therefore this method will need to be modified for extension into a multisite domain.

Using multi‐model ensembles of CMIP5 global climate models to reproduce observed monthly rainfall and temperature with machine learning methods in Australia

“…The 90% of data sets were selected as training data in our study because we would like to include more data to establish the relationship between GCMs and observed climate data. In previous literatures, the proportion of training data can range from 60 to 95% (Deo and Şahin, ; Vu et al ., ; Hou et al ., ). Therefore, 90% was within the range when we designed the study.…”

“…SVM uses hyperplanes to divide all of the data into different classes optimally. It has a better learning capability and smaller prediction errors than many other methods (Chen et al ., ; Sarhadi et al ., ; Hou et al ., ).…”

Daily precipitation performances of regression-based statistical downscaling models in a basin with mountain and semi-arid climates

“…It is also thought that 40 years of data may represent the actual climatic conditions for the area in question, including less frequent climatic events (Khan et al 2006 ). Again, it has been stated in some studies (Khan et al 2006 , Huang et al 2011 ; Nasseri et al 2013 ; Tavakol-Davani et al 2013 ) that the long training period increases the performance of the models and the potential of the models to catch rare climatic events. Wilby ( 1998 ) also stated that the SD model could perform nonstationary using a long calibration period.…”

“…However, as dynamic downscaling methods require high computing power and design and some specialized knowledge, the statistical downscaling (SD) methods are mainly used in hydrological studies thanks to their cheaper and easier applications (Fowler et al 2007 ; Chen et al 2010 ; Chen et al 2014 ; Ekstrom et al 2015 ). The SD methods are broadly divided into three parts; transfer function (perfect prognosis), weather generator, and weather pattern approach (Chen et al 2010 ; Maraun et al 2010 ; Tavakol-Davani et al 2013 ; Chen et al 2014 ; Hou et al 2017 ). The transfer function method is frequently used because it is easy to apply anywhere and/or anytime.…”

Statistical Applications to Downscale GRACE-Derived Terrestrial Water Storage Data and to Fill Temporal Gaps

“…2020, 12, 533 3 of 29 outputs (dependent variables) using coarse-scale datasets (inputs and outputs) and applies the extracted relationships to produce the dependent variables at a spatial resolution matching that of the fine-scale inputs [24]. A variety of statistical methods have been applied to downscale remote sensing or ground-based data, including Markov chains and support vector machines [25], regression kriging [26], neural networks [27], and stochastic models [28]. Stepwise regression was successfully applied to downscale satellite-based precipitation data (TRMM3B43 products) and average daily precipitation and air temperature data from weather stations [29,30].…”