【正文】 22. RnnLet be the daily precipitation amount on day in period. If represents any reasonable daily precipitation value then, count the number of days where:23. CDD*連續(xù)無雨日數(shù)（持續(xù)干燥指數(shù)），閾值為1mmLet be the daily precipitation amount on day in period. Count the largest number of consecutive days where:譯者注：譯者通過對源程序CDD模塊的分析，找到了修改閾值的語句。因為RCLIMDEX自帶的CDD閾值僅是1mm，而農(nóng)業(yè)干旱等實際研究中存在多種降水閾值的設定。如3mm，5mm等。一篇美國農(nóng)業(yè)干旱文獻設為。僅需修改3處，其他操作不變?nèi)缦拢ǎ骸?將prcp=1改為prcp=（或其他因地制宜之干旱指標閾值）　 將PRCP=1mm改為PRCP=（此句僅用作繪圖標題，并不參與計算，不改亦無大礙） 　　這是最關鍵的一句。應將mid[j]1改為 mid[j]24. CWD*Let be the daily precipitation amount on day in period. Count the largest number of consecutive days where:25. R95pTOTLet be the daily precipitation amount on a wet day in periodand letbe the 95th percentile of precipitation on wet days in the 19611990 period. If represents the number of wet days in the period, then:26. R99pLet be the daily precipitation amount on a wet day in periodand letbe the 99th percentile of precipitation on wet days in the 19611990 period. If represents number of wet days in the period, then:27. PRCPTOTLet be the daily precipitation amount on day in period. If represents the number of days in, thenAppendix D: 閾值（門檻值）估計與基期溫度指數(shù)計算Empirical quantile estimation:The quantile of a distribution is defined as , 1p1,where F(x) is the distribution function. Let denote the order statistics of (. sorted values of {X}), and let denote the ith sample quantile definition. The sample quantiles can be generally written as.Hyndman and Fan (1996) suggest a formula to obtain medium unbiased estimate of the quantile by letting and letting , where int(u) is the largest integer not greater than u. The empirical quantile is set to the smallest or largest value in the sample when j1 or j n respectively. That is, quantile estimates corresponding to p1/(n+1) are set to the smallest value in the sample, and those corresponding to pn/(n+1) are set to the largest value in the sample.Bootstrap procedure for the estimation of exceedance rate for the base period:It is not possible to make an exact estimate of the thresholds due to sampling uncertainty. To provide temporally consistent estimate of exceedance rate throughout the base period and outofbase period, we adapt the following procedure (Zhang et al. 2004) to estimate exceedance rate for the base period.a) The 30year base period is divided into one “outofbase” year, the year for which exceedance is to be estimated, and a “baseperiod” consisting the remaining of 29 years from which the thresholds would be estimated.b) A 30year block of data is constructed by using the 29 year “baseperiod” data set and adding an additional year of data from the “baseperiod (., one of the years in the “baseperiod” is repeated). This constructed 30year block is used to estimate thresholds.c) The “outofbase” year is then pared with these thresholds and the exceedance rate for the “outofbase” year is obtained.d) Steps (b) and (c) are repeated for an additional 28 times, by repeating each of the remaining 28 inbase years in turn to construct the 30year block.e) The final index for the “outofbase” year is obtained by averaging the 29 estimates obtained from steps (b), (c) and (d).譯者注：a)30年基期被分成兩部分，第１個值被視作基期外之值估計超出量，剩余的29個值用作新基期估計門檻（大意）。Reference:Hyndman, ., and Y. Fan, 1996: Sample quantiles in statistical packages. The American Statistician, 50, 361367.Zhang, X., G. Hegerl, . Zwiers, and J. Kenyon, 2004: Avoiding inhomogeneity in percentilebased indices of temperature extremes. J. Climate, submitted.Appendix E: R for Windows FAQ24