Statistical prediction of non‐Gaussian climate extremes in urban areas based on the first‐order difference method. (7th March 2018)

• Record Type: Journal Article
• Authors: Qian, Cheng; Zhou, Wen; Yang, Xiu‐Qun; Chan, Johnny C. L.
• Is Part Of: International journal of climatology. Volume 38:Number 6(2018)
• Page Start: 2889
• Page End: 2898
• Publication Date: 2018-03-07
• Subjects: climate extremes -- first‐order difference -- non‐Gaussian -- seasonal prediction -- urbanization effect; Climatology -- Periodicals; Climat -- Périodiques; Climatologie -- Périodiques; 551.605
• DOI: 10.1002/joc.5464
• Languages: English
• ISSNs: 0899-8418

Abstract

Abstract : Prediction of climate extremes is challenging, especially for non‐Gaussian extremes in urban areas where the majority of people live, since the Gaussian assumption used in linear regression is violated and the urbanization effect needs to be considered. In this study, the first‐order difference method is introduced to take these difficulties into account. Statistical prediction of the non‐Gaussian annual occurrence of hot days in downtown Hong Kong, which is highly urbanized, is used to illustrate this method. With the help of the first‐order difference of the annual occurrences, which follows a Gaussian distribution, the difference series is used as the predictant to find predictors and to construct a prediction model by using traditional linear regression. The difference is first predicted and is then added to the observed value at the preceding time to obtain the predicted annual occurrence. The historical urbanization effect is thus obtained directly from the observations at the preceding time. The prediction results are found desirable. The broad application potential and conditions in which this method should be used are also discussed. Abstract : Time series of annual occurrence of hot days (blue curve) in Hong Kong for the period 1947–2016 (a) and its normality tests (b, c). In (a), the linear trend estimated by the WS2001 (red line) and the 11‐year running mean (black curve) are also plotted. In (b), the blue bars are the target data under testing and the red line is the fitted Gaussian distribution. The p > .05 in the Jarque–Bera test indicates that the target data are normally distributed. In the quantile–quantile plot (c), red circles indicate the distribution of the target data and the black solid lines represent the Gaussian distribution, with the 95% confidence intervals shown as black dashed lines. The approximate linearity of the circles suggests that the target data are normally distributed. (d), (e) and (f) are the same as (a), (b) and (c), respectively, but for the year‐to‐year increment of hot days.