{"id":2525,"date":"2016-10-06T13:04:57","date_gmt":"2016-10-06T13:04:57","guid":{"rendered":"http:\/\/www.vtei.cz\/?p=2525"},"modified":"2024-07-16T11:46:07","modified_gmt":"2024-07-16T10:46:07","slug":"bias-correction-of-precipitation-and-temperature-from-regional-climate-models-the-impact-on-runoff-modelling","status":"publish","type":"post","link":"https:\/\/www.vtei.cz\/en\/2016\/10\/bias-correction-of-precipitation-and-temperature-from-regional-climate-models-the-impact-on-runoff-modelling\/","title":{"rendered":"Bias correction of precipitation and temperature from regional climate models \u2013 the impact on runoff modelling"},"content":{"rendered":"

<\/i> This article is available in Czech only. For translation or more information on this topic, please contact author.<\/h4>\n

Souhrn<\/h2>\n

Pro odhad vlivu zm\u011bny klimatu na vodn\u00ed re\u017eim v\u00a0krajin\u011b a\u00a0sektory vodn\u00edho hospod\u00e1\u0159stv\u00ed se zpravidla u\u017e\u00edv\u00e1 hydrologick\u00e9ho modelov\u00e1n\u00ed, kde vstupem do hydrologick\u00e9ho modelu jsou sc\u00e9n\u00e1\u0159ov\u00e9 sr\u00e1\u017eky a\u00a0teploty zalo\u017een\u00e9 na simulac\u00edch klimatick\u00fdch model\u016f. Tyto simulace jsou zat\u00ed\u017eeny systematick\u00fdmi chybami, kter\u00e9 lze korigovat pomoc\u00ed \u0159ady dostupn\u00fdch metod. Nicm\u00e9n\u011b se ukazuje, \u017ee shoda rozd\u011blen\u00ed korigovan\u00fdch a\u00a0pozorovan\u00fdch vstupn\u00edch veli\u010din (sr\u00e1\u017eek a\u00a0teploty) nezaru\u010duje shodu rozd\u011blen\u00ed simulovan\u00e9ho odtoku. To je zp\u016fsobeno zejm\u00e9na skute\u010dnost\u00ed, \u017ee b\u011b\u017en\u011b pou\u017e\u00edvan\u00e9 metody korekce systematick\u00fdch chyb klimatick\u00fdch model\u016f neodstra\u0148uj\u00ed chyby v\u00a0\u010dasov\u00e9 struktu\u0159e sr\u00e1\u017eek. Dal\u0161\u00edm probl\u00e9mem je, \u017ee korekce se zpravidla omezuj\u00ed na \u010dasov\u00e9 m\u011b\u0159\u00edtko, v\u00a0n\u011bm\u017e je (hydrologick\u00fd) model provozov\u00e1n\u00a0\u2013 tj.\u00a0v\u011bt\u0161inou denn\u00ed. Ukazuje se, \u017ee p\u0159es uspokojivou korekci veli\u010din v\u00a0denn\u00edm \u010dasov\u00e9m kroku jsou m\u011bs\u00ed\u010dn\u00ed, sezonn\u00ed a\u00a0ro\u010dn\u00ed agregace sr\u00e1\u017eek zat\u00ed\u017eeny podstatnou systematickou chybou, co\u017e n\u00e1sledn\u011b vede k\u00a0chyb\u00e1m v\u00a0dlouhodob\u00e9 hydrologick\u00e9 bilanci a\u00a0variabilit\u011b simulovan\u00e9ho odtoku.<\/p>\n

\"novodomske-raseliniste\"<\/a><\/p>\n

\u00davod<\/h2>\n

Jeliko\u017e jsou simulace klimatick\u00fdch model\u016f zat\u00ed\u017eeny systematick\u00fdmi chybami, nelze simulovan\u00e9 \u010dasov\u00e9 \u0159ady sr\u00e1\u017eek a\u00a0teploty\u00a0pou\u017e\u00edt p\u0159\u00edmo pro hydrologick\u00e9 modelov\u00e1n\u00ed. Relativn\u011b hrub\u00e9 horizont\u00e1ln\u00ed rozli\u0161en\u00ed glob\u00e1ln\u00edch i\u00a0region\u00e1ln\u00edch klimatick\u00fdch model\u016f (RCM) neumo\u017e\u0148uje adekv\u00e1tn\u011b popsat orografii ter\u00e9nu, vznik konvektivn\u00edch sr\u00e1\u017eek\u00a0atp. V\u00a0d\u016fsledku toho je \u0159ada jev\u016f pops\u00e1na pomoc\u00ed empirick\u00fdch nebo semi\u00ad\u2011empirick\u00fdch vztah\u016f, kter\u00e9 jsou \u010dasto zmi\u0148ov\u00e1ny jako jedna z\u00a0dominantn\u00edch p\u0159\u00ed\u010din systematick\u00fdch chyb v\u00a0simulac\u00edch klimatick\u00fdch model\u016f [1\u20133].<\/p>\n

Existuje \u0159ada studi\u00ed, kter\u00e9 se zab\u00fdvaj\u00ed korekc\u00ed chyb klimatick\u00fdch model\u016f. Korekce simulovan\u00fdch sr\u00e1\u017eek a\u00a0teploty\u00a0spo\u010d\u00edv\u00e1 v\u00a0transformaci simulovan\u00fdch \u010dasov\u00fdch \u0159ad takov\u00fdm zp\u016fsobem, kter\u00fd zaru\u010d\u00ed p\u0159ijatelnou shodu vybran\u00fdch statistick\u00fdch charakteristik simulovan\u00fdch a\u00a0pozorovan\u00fdch veli\u010din. Byly vyvinuty r\u016fzn\u00e9 metody korekce systematick\u00fdch chyb, od transformac\u00ed koriguj\u00edc\u00edch\u00a0pr\u016fm\u011br p\u0159es neline\u00e1rn\u00ed transformace koriguj\u00edc\u00ed pr\u016fm\u011br a\u00a0variabilitu a\u017e po pokro\u010dil\u00e9 metody transformuj\u00edc\u00ed cel\u00e9 rozd\u011blen\u00ed pravd\u011bpodobnosti uva\u017eovan\u00fdch veli\u010din, pop\u0159.\u00a0vztahy mezi prom\u011bnn\u00fdmi. Tyto metody p\u0159edstavuje nap\u0159.\u00a0[4\u20136]. \u0158ada aspekt\u016f korekce systematick\u00fdch chyb je v\u0161ak problematick\u00e1, viz\u00a0nap\u0159.\u00a0[7, 8].<\/p>\n

Simulace RCM jsou zpravidla k\u00a0dispozici v\u00a0denn\u00edm \u010dasov\u00e9m kroku. V\u00a0tomto \u010dasov\u00e9m m\u011b\u0159\u00edtku tak\u00e9 prob\u00edh\u00e1 korekce simulovan\u00fdch veli\u010din i\u00a0hodnocen\u00ed jej\u00ed \u00fa\u010dinnosti. Sou\u010dasn\u00e9 metody korekce systematick\u00fdch chyb jsou schopn\u00e9 transformovat simulovan\u00e1 data tak, \u017ee rozd\u011blen\u00ed korigovan\u00fdch veli\u010din se perfektn\u011b shoduje s\u00a0rozd\u011blen\u00edm veli\u010din pozorovan\u00fdch a\u00a0nav\u00edc rozumn\u011b zachov\u00e1v\u00e1 korela\u010dn\u00ed strukturu mezi veli\u010dinami. Shoda rozd\u011blen\u00ed a\u00a0vztah\u016f veli\u010din v\u00a0denn\u00edm kroku v\u0161ak neznamen\u00e1 shodu v\u00a0p\u0159\u00edpad\u011b del\u0161\u00edch \u010di krat\u0161\u00edch \u010dasov\u00fdch m\u011b\u0159\u00edtek, co\u017e je zp\u016fsobeno \u010dasovou strukturou simulovan\u00fdch veli\u010din,\u00a0kter\u00e1 nen\u00ed zpravidla korekc\u00ed ovlivn\u011bna. Toto chov\u00e1n\u00ed bylo pops\u00e1no nap\u0159.\u00a0v\u00a0prac\u00edch Haerter\u00a0aj. [9], Johnson a\u00a0Sharma [10], Ehret\u00a0aj. [7] a\u00a0Addor\u00a0aj. [8], nicm\u00e9n\u011b ve v\u011bt\u0161in\u011b praktick\u00fdch aplikac\u00ed je tento probl\u00e9m \u010dasto p\u0159ehl\u00ed\u017een. Hodnocen\u00ed metod korekce se nav\u00edc \u010dasto omezuje na veli\u010diny simulovan\u00e9 klimatick\u00fdm modelem (nap\u0159.\u00a0sr\u00e1\u017eky a\u00a0teplota) a\u00a0neuva\u017euje v\u00fdstupy z\u00e1visl\u00e9 na t\u011bchto veli\u010din\u00e1ch (nap\u0159.\u00a0odtok).<\/p>\n

V\u00a0p\u0159edkl\u00e1dan\u00e9m \u010dl\u00e1nku jsou proto demonstrov\u00e1ny z\u00e1kladn\u00ed probl\u00e9my korekce systematick\u00fdch chyb p\u0159i pou\u017eit\u00ed v\u00a0hydrologick\u00fdch simulac\u00edch. Ukazuje se, \u017ee rozd\u011blen\u00ed odtoku simulovan\u00e9ho pomoc\u00ed hydrologick\u00e9ho modelu vyu\u017e\u00edvaj\u00edc\u00edho korigovan\u00e9 sr\u00e1\u017eky a\u00a0teplotu neodpov\u00edd\u00e1 rozd\u011blen\u00ed odtoku simulovan\u00e9ho na z\u00e1klad\u011b pozorovan\u00fdch \u010dasov\u00fdch \u0159ad ani v\u00a0\u010dasov\u00e9m kroku, ve kter\u00e9m byla korekce provedena. Pro anal\u00fdzu \u00fa\u010dinnosti metod korekce systematick\u00fdch chyb na r\u016fzn\u00e9 veli\u010diny v\u00a0r\u016fzn\u00fdch \u010dasov\u00fdch m\u011b\u0159\u00edtc\u00edch byl vyvinut bal\u00edk MUSICA pro R software, viz\u00a0https:\/\/github.com\/hanel\/musica.<\/p>\n

Popis oblasti a\u00a0data<\/h2>\n

Studie byla provedena na povod\u00ed Oslavy, kter\u00e9 m\u00e1 rozlohu 861\u2008km2<\/span><\/sup> s\u00a0pr\u016fm\u011brnou v\u00fd\u0161kou 500\u00a0m\u00a0n.\u00a0m. Pr\u016fm\u011brn\u00e1 ro\u010dn\u00ed sr\u00e1\u017eka \u010din\u00ed 594\u2008mm, pr\u016fm\u011brn\u00e1 teplota 7,2\u2008\u00b0C a\u00a0pr\u016fm\u011brn\u00fd pr\u016ftok v\u00a0\u00fast\u00ed je 3,5\u2008m3<\/span><\/sup>\/s. Povod\u00ed je z\u00a0v\u011bt\u0161\u00ed \u010d\u00e1sti neregulovan\u00e9, pouze na horn\u00edm toku Oslavy je v\u00edce\u00fa\u010delov\u00e1 vod\u00e1rensk\u00e1 n\u00e1dr\u017e Mosti\u0161t\u011b. Nevyskytuj\u00ed se zde vy\u0161\u0161\u00ed elevace, kter\u00e9 by znesnad\u0148ovaly pou\u017eit\u00ed RCM.<\/p>\n

Pozorovan\u00e1 data<\/h3>\n

Pro studii byla pou\u017eita hydrometeorologick\u00e1 data (sr\u00e1\u017eky, teplota, pr\u016ftoky) v\u00a0denn\u00edm kroku z\u00a0obdob\u00ed 1970\u20131999. \u010casov\u00e9 \u0159ady sr\u00e1\u017eek a\u00a0teploty poch\u00e1zej\u00ed z\u00a0interpolovan\u00fdch dat do pravideln\u00e9 s\u00edt\u011b 25 \u00d7 25\u2008km [11], pr\u016ftoky pak z\u00a0vodom\u011brn\u00e9 stanice Oslavany.<\/p>\n

Tabulka 1. Pou\u017eit\u00e9 RCM simulace; v\u00a0\u0159\u00e1dc\u00edch jsou uvedeny \u0159\u00eddic\u00ed simulace glob\u00e1ln\u00edch klimatick\u00fdch model\u016f, sloupce odpov\u00eddaj\u00ed region\u00e1ln\u00edm klimatick\u00fdm model\u016fm; v\u00a0jednotliv\u00fdch bu\u0148k\u00e1ch je v\u00a0z\u00e1vork\u00e1ch uveden po\u010det simulac\u00ed podle RCP2.6 (\u201e26\u201c), RCP4.5 (\u201e45\u201c) a\u00a0RCP8.5 (\u201e85\u201c)
\nTable 1. Considered RCM simulations; driving GCM simulations are listed in rows whereas columns correspond to RCMs; a\u00a0number of simulation according to RCP2.6 (\u201e26\u201c),
\nRCP4.5 (\u201e45\u201c) and RCP8.5 (\u201e85\u201c) is listed in brackets in individual cells<\/h5>\n\"hanel-tabulka-1\"<\/a>\n

RCM data<\/h3>\n

Pro vyhodnocen\u00ed \u00fa\u010dinku korekce systematick\u00fdch chyb simulovan\u00fdch \u010dasov\u00fdch \u0159ad sr\u00e1\u017eek a\u00a0teploty na simulovan\u00fd odtok bylo uva\u017eov\u00e1no 52 simulac\u00ed region\u00e1ln\u00edch klimatick\u00fdch model\u016f z\u00a0projektu CORDEX [12]. P\u0159ehled pou\u017eit\u00fdch simulac\u00ed ud\u00e1v\u00e1 tabulka 1<\/em>. K\u00a0dispozici bylo sedm region\u00e1ln\u00edch klimatick\u00fdch model\u016f k\u00a0downscalingu simulac\u00ed deseti glob\u00e1ln\u00edch klimatick\u00fdch model\u016f. Z\u00a0region\u00e1ln\u00edch klimatick\u00fdch model\u016f je nejpo\u010detn\u011bji zastoupen model RCA4 (30 simulac\u00ed) a\u00a0CLM (8 simulac\u00ed). Simulace nej\u010dast\u011bji vyu\u017e\u00edvaj\u00ed koncentra\u010dn\u00ed sc\u00e9n\u00e1\u0159 RCP8.5 p\u0159edpokl\u00e1daj\u00edc\u00ed nejintenzivn\u011bj\u0161\u00ed zvy\u0161ov\u00e1n\u00ed radia\u010dn\u00edho p\u016fsoben\u00ed (26 simulac\u00ed) a\u00a0RCP4.5 (21 simulac\u00ed), m\u00e9n\u011b pak RCP2.6 p\u0159edpokl\u00e1daj\u00edc\u00ed sni\u017eov\u00e1n\u00ed koncentrac\u00ed sklen\u00edkov\u00fdch plyn\u016f (3 simulace). Pro RCP6.0 nebyla dostupn\u00e1 \u017e\u00e1dn\u00e1 simulace. Simulace jsou v\u011bt\u0161inou dostupn\u00e9 pro obdob\u00ed cca\u00a01961\u20132100, n\u011bkter\u00e9 1950\u20132100. Prostorov\u00e9 rozli\u0161en\u00ed je 0,11\u00b0 a\u00a00,44\u00b0, co\u017e odpov\u00edd\u00e1 cca\u00a012 a\u00a050\u2008km. V\u0161echny simulace jsou voln\u011b dostupn\u00e9 prost\u0159ednictv\u00edm Earth System Grid Federation.<\/p>\n

Metodika<\/h2>\n

Pozorovan\u00e9 \u010dasov\u00e9 \u0159ady sr\u00e1\u017eek a\u00a0teploty byly pou\u017eity pro kalibraci hydrologick\u00e9ho modelu Bilan (kap. Hydrologick\u00e9 modelov\u00e1n\u00ed). Simulovan\u00e9 \u010dasov\u00e9 \u0159ady sr\u00e1\u017eek a\u00a0teploty byly korigov\u00e1ny standardn\u00ed a\u00a0kask\u00e1dovou kvantilovou metodou (kap. Korekce systematick\u00fdch chyb). Vybran\u00e9 charakteristiky rozd\u011blen\u00ed korigovan\u00fdch sr\u00e1\u017eek, teploty a\u00a0odpov\u00eddaj\u00edc\u00edho modelovan\u00e9ho odtoku byly porovn\u00e1ny (kap. Vyhodnocen\u00ed) s\u00a0charakteristikami pozorovan\u00fdch veli\u010din (sr\u00e1\u017eek, teploty a\u00a0odpov\u00eddaj\u00edc\u00edho modelovan\u00e9ho odtoku).<\/p>\n\"hanel-1\"<\/a>\n

Obr.\u00a01. Zbytkov\u00e1 chyba (relativn\u00ed pro sr\u00e1\u017eky a\u00a0odtok, absolutn\u00ed pro teplotu [\u00b0C]) po korekci standardn\u00ed kvantilovou metodou pro pr\u016fm\u011br (vlevo), 90% kvantil (uprost\u0159ed) a\u00a0sm\u011brodatnou odchylku (vpravo), pro sr\u00e1\u017eky (naho\u0159e), teplotu (uprost\u0159ed) a\u00a0modelovan\u00fd odtok (dole); barvy ur\u010duj\u00ed ro\u010dn\u00ed obdob\u00ed (\u010derven\u00e1\u00a0\u2013 zima, zelen\u00e1 \u2013 jaro, modr\u00e1\u00a0\u2013 l\u00e9to, sv\u011btle modr\u00e1\u00a0\u2013 podzim); vodorovn\u00e1 osa zna\u010d\u00ed \u010dasov\u00e9 agregace (D1\u00a0\u2013 den, D10\u00a0\u2013 deset dn\u016f, M1\u00a0\u2013 m\u011bs\u00edc, M3\u00a0\u2013 t\u0159i m\u011bs\u00edce, M6\u00a0\u2013 \u0161est m\u011bs\u00edc\u016f, Y1\u00a0\u2013 rok, Y5\u00a0\u2013 p\u011bt let); \u010d\u00e1ry zna\u010d\u00ed pr\u016fm\u011brnou chybu, zelen\u00fd p\u00e1s rozp\u011bt\u00ed 50\u00a0% a\u00a0modr\u00fd p\u00e1s 90\u00a0% ze v\u0161ech model\u016f
\nFig. 1. Residual error (relative for precipitation and runoff, absolute for temperature [\u00b0C]) after the standard quantile correction method for mean (left), 90% quantile (middle) and standard deviation (right), for precipitation (top), temperature (middle) and simulated runoff (bottom); colours indicate seasons (red\u00a0\u2013 winter, green\u00a0\u2013 spring, blue\u00a0\u2013 summer, light blue\u00a0\u2013 autumn); horizontal axis shows time aggregation (D1\u00a0\u2013 a\u00a0day, D10\u00a0\u2013 ten days, M1\u00a0\u2013 a\u00a0month, M3\u00a0\u2013 three months, M6\u00a0\u2013 six months, Y1\u00a0\u2013 a\u00a0year, Y5\u00a0\u2013 five years); the lines represent mean residual error in the climate model ensemble, green (blue) area indicates an envelope of 50 (90)% of climate model simulations<\/h6>\n

Korekce systematick\u00fdch chyb<\/h3>\n

Simulovan\u00e9 \u010dasov\u00e9 \u0159ady sr\u00e1\u017eek a\u00a0teploty byly korigov\u00e1ny pomoc\u00ed standardn\u00ed kvantilov\u00e9 metody popsan\u00e9 nap\u0159.\u00a0[13]. Tato metoda zaru\u010duje, \u017ee rozd\u011blen\u00ed pravd\u011bpodobnosti sr\u00e1\u017eek a\u00a0teploty korigovan\u00fdch dat odpov\u00edd\u00e1 rozd\u011blen\u00ed pravd\u011bpodobnosti pozorovan\u00fdch veli\u010din. Kvantilov\u00e1 metoda byla pou\u017eita v\u00a0denn\u00edm kroku, zvl\u00e1\u0161\u0165 pro jednotliv\u00e9 m\u011bs\u00edce.<\/p>\n

Za \u00fa\u010delem vyhodnocen\u00ed vlivu korigovan\u00e9ho \u010dasov\u00e9ho m\u011b\u0159\u00edtka na modelovan\u00fd pr\u016ftok byla kvantilov\u00e1 metoda aplikov\u00e1na tak\u00e9 iterativn\u011b pro r\u016fzn\u00e1 \u010dasov\u00e1 m\u011b\u0159\u00edtka (konkr\u00e9tn\u011b denn\u00ed, m\u011bs\u00ed\u010dn\u00ed a\u00a0ro\u010dn\u00ed) pomoc\u00ed p\u0159\u00edstupu popsan\u00e9ho Haerterem\u00a0aj. [9]. Podstatou metody je opakovan\u00e1 korekce denn\u00edch \u010dasov\u00fdch \u0159ad na z\u00e1klad\u011b rozd\u011blen\u00ed denn\u00edch hodnot a\u00a0m\u011bs\u00ed\u010dn\u00edch a\u00a0ro\u010dn\u00edch agregac\u00ed. Vzhledem k\u00a0tomu, \u017ee korekce v\u00a0jednom \u010dasov\u00e9m m\u011b\u0159\u00edtku ovliv\u0148uje rozd\u011blen\u00ed v\u00a0jin\u00fdch \u010dasov\u00fdch m\u011b\u0159\u00edtc\u00edch, je postup iterativn\u011b opakov\u00e1n.<\/p>\n

Hydrologick\u00e9 modelov\u00e1n\u00ed<\/h3>\n

Simulace odtoku z\u00a0povod\u00ed Oslavy byla provedena modelem Bilan vyv\u00edjen\u00fdm ve V\u00fdzkumn\u00e9m \u00fastavu vodohospod\u00e1\u0159sk\u00e9m (V\u00daV). Bilan [14] je konceptu\u00e1ln\u00ed model hydrologick\u00e9 bilance, kter\u00fd je v\u00a0denn\u00edm kroku \u0159\u00edzen \u0161esti parametry. Sr\u00e1\u017eky jsou transformov\u00e1ny na odtok pomoc\u00ed soustavy line\u00e1rn\u00edch a\u00a0neline\u00e1rn\u00edch n\u00e1dr\u017e\u00ed. Kalibrace parametr\u016f modelu prob\u00edh\u00e1 na pozorovan\u00fdch \u010dasov\u00fdch \u0159ad\u00e1ch (sr\u00e1\u017eky, teplota a\u00a0odtok), kdy se sleduje nejlep\u0161\u00ed shoda mezi pozorovan\u00fdm a\u00a0modelovan\u00fdm odtokem. Nakalibrovan\u00fdm modelem se d\u00e1le generuje odtok na z\u00e1klad\u011b korigovan\u00fdch sr\u00e1\u017eek a\u00a0teploty vych\u00e1zej\u00edc\u00edch z\u00a0RCM. V\u00edce o\u00a0modelu lze nal\u00e9zt na bilan.vuv.cz.<\/p>\n

Vyhodnocen\u00ed<\/h3>\n

K\u00a0vyhodnocen\u00ed zbytkov\u00e9 chyby v\u00a0korigovan\u00fdch \u010dasov\u00fdch \u0159ad\u00e1ch sr\u00e1\u017eek, teploty a\u00a0modelovan\u00e9ho odtoku byl pou\u017eit bal\u00edk MUSICA pro prost\u0159ed\u00ed R [15]. Bal\u00edk umo\u017e\u0148uje pohodln\u00e9 porovn\u00e1n\u00ed libovoln\u00fdch charakteristik rozd\u011blen\u00ed hodnocen\u00fdch veli\u010din pro r\u016fzn\u00e1 \u010dasov\u00e1 m\u011b\u0159\u00edtka, p\u0159i\u010dem\u017e tato m\u011b\u0159\u00edtka je mo\u017eno u\u017eivatelsky definovat. V\u00a0tomto \u010dl\u00e1nku jsou d\u00e1le uva\u017eov\u00e1na denn\u00ed (D1), 10denn\u00ed (D10), m\u011bs\u00ed\u010dn\u00ed (M1), 3 a\u00a06m\u011bs\u00ed\u010dn\u00ed (M3 a\u00a0M6) a\u00a0ro\u010dn\u00ed a\u00a0p\u011btilet\u00e9 (Y1 a\u00a0Y5) \u010dasov\u00e9 agregace. Pro krat\u0161\u00ed ne\u017e p\u016flro\u010dn\u00ed agregace jsou v\u00fdsledky prezentov\u00e1ny jako pr\u016fm\u011bry pro jednotliv\u00e9 sezony (MAM\u00a0\u2013 b\u0159ezen, duben, kv\u011bten; JJA\u00a0\u2013 \u010derven, \u010dervenec, srpen; SON\u00a0\u2013 z\u00e1\u0159\u00ed, \u0159\u00edjen, listopad; DJF\u00a0\u2013 prosinec, leden, \u00fanor). Prim\u00e1rn\u011b se hodnot\u00ed chyba korigovan\u00fdch veli\u010din, kter\u00e1 je d\u00e1le ozna\u010dov\u00e1na jako \u201ezbytkov\u00e1 chyba\u201c.<\/p>\n\"hanel-2\"<\/a>\n

Obr.\u00a02. Zbytkov\u00e1 chyba (relativn\u00ed pro sr\u00e1\u017eky a\u00a0odtok, absolutn\u00ed pro teplotu [\u00b0C]) po korekci kask\u00e1dovou kvantilovou metodou pro pr\u016fm\u011br (vlevo), 90% kvantil (uprost\u0159ed) a\u00a0sm\u011brodatnou odchylku (vpravo), pro sr\u00e1\u017eky (naho\u0159e), teplotu (uprost\u0159ed) a\u00a0modelovan\u00fd odtok (dole); barvy ur\u010duj\u00ed ro\u010dn\u00ed obdob\u00ed (\u010derven\u00e1\u00a0\u2013 zima, zelen\u00e1\u00a0\u2013 jaro, modr\u00e1\u00a0\u2013 l\u00e9to, sv\u011btle modr\u00e1\u00a0\u2013 podzim); vodorovn\u00e1 osa zna\u010d\u00ed \u010dasov\u00e9 agregace (D1\u00a0\u2013 den, D10\u00a0\u2013 deset dn\u016f, M1\u00a0\u2013 m\u011bs\u00edc, M3\u00a0\u2013 t\u0159i m\u011bs\u00edce, M6\u00a0\u2013 \u0161est m\u011bs\u00edc\u016f, Y1\u00a0\u2013 rok, Y5\u00a0\u2013 p\u011bt let); \u010d\u00e1ry zna\u010d\u00ed pr\u016fm\u011brnou chybu, zelen\u00fd p\u00e1s rozp\u011bt\u00ed 50\u00a0% a\u00a0modr\u00fd p\u00e1s 90\u00a0% ze v\u0161ech model\u016f
\nFig. 2. Residual error (relative for precipitation and runoff, absolute for temperature [\u00b0C]) after the cascade quantile correction method for mean (left), 90% quantile (middle) and standard deviation (right), for precipitation (top), temperature (middle) and simulated runoff (bottom); colours indicate seasons (red\u00a0\u2013 winter, green\u00a0\u2013 spring, blue\u00a0\u2013 summer, light blue\u00a0\u2013 autumn); horizontal axis shows time aggregation (D1\u00a0\u2013 a\u00a0day, D10\u00a0\u2013 ten days, M1\u00a0\u2013 a\u00a0month, M3\u00a0\u2013 three months, M6\u00a0\u2013 six months, Y1\u00a0\u2013 a\u00a0year, Y5\u00a0\u2013 five years); the lines represent mean residual error in the climate model ensemble, green (blue) area indicates an envelope of 50 (90)% of climate model simulations<\/h6>\n

V\u00fdsledky<\/h2>\n

Z\u00a0obr.\u00a01<\/em> a\u00a02<\/em> je patrn\u00e9, \u017ee pro v\u0161echny t\u0159i prezentovan\u00e9 charakteristiky (pr\u016fm\u011br, 90% kvantil a\u00a0sm\u011brodatn\u00e1 odchylka) jsou zbytkov\u00e9 chyby velmi mal\u00e9 u\u00a0sr\u00e1\u017eek i\u00a0teploty, zejm\u00e9na v\u00a0D1 m\u011b\u0159\u00edtku, pro kter\u00e9 byly korekce kalibrov\u00e1ny. Nicm\u00e9n\u011b v\u00a0p\u0159\u00edpad\u011b pou\u017eit\u00ed t\u011bchto sr\u00e1\u017eek a\u00a0teplot pro modelov\u00e1n\u00ed odtoku doch\u00e1z\u00ed k\u00a0\u201ezes\u00edlen\u00ed\u201c zbytkov\u00fdch chyb nad \u00fanosnou mez. Uspokojiv\u00e9 v\u00fdsledky se nedostavily ani pro m\u011b\u0159\u00edtko D1, pro kter\u00e9 byly korekce kalibrov\u00e1ny. Extr\u00e9mn\u00edm p\u0159\u00edpadem je podzimn\u00ed sezona (sv\u011btle mod\u0159e)\u00a0\u2013 v\u00a0absolutn\u00edm vyj\u00e1d\u0159en\u00ed se v\u0161ak jedn\u00e1 o\u00a0velmi mal\u00e9 hodnoty.<\/p>\n\"hanel-3\"<\/a>\n

Obr.\u00a03. Korelace mezi\u00a0sr\u00e1\u017ekou a\u00a0modelovan\u00fdm odtokem (naho\u0159e), mezi sr\u00e1\u017ekou a\u00a0teplotou (uprost\u0159ed) a\u00a0mezi teplotou a\u00a0modelovan\u00fdm odtokem (dole) pro jednotliv\u00e1 ro\u010dn\u00ed obdob\u00ed (DJF\u00a0\u2013 zima, MAM\u00a0\u2013 jaro, JJA\u00a0\u2013 l\u00e9to, SON\u00a0\u2013 podzim); \u010derven\u011b jsou zobrazena korigovan\u00e1 data standardn\u00ed kvantilovou metodou, mod\u0159e pozorovan\u00e1 data a\u00a0zelen\u011b simulovan\u00e1 data; vodorovn\u00e1 osa zna\u010d\u00ed \u010dasov\u00e1 m\u011b\u0159\u00edtka (D1\u00a0\u2013 den, D10\u00a0\u2013 deset dn\u016f, M1\u00a0\u2013 m\u011bs\u00edc, M3\u00a0\u2013 t\u0159i m\u011bs\u00edce, M6\u00a0\u2013 \u0161est m\u011bs\u00edc\u016f, Y1\u00a0\u2013 rok, Y5\u00a0\u2013 p\u011bt let); \u010d\u00e1ry zna\u010d\u00ed pr\u016fm\u011brnou korelaci, zelen\u00fd p\u00e1s rozp\u011bt\u00ed 50\u00a0% a\u00a0modr\u00fd p\u00e1s 90\u00a0% ze v\u0161ech model\u016f
\nFig. 3. Correlation between precipitation and modelled runoff (above), between precipitation and temperature (middle) and between temperature and modelled runoff (below) for seasons of year (DJF\u00a0\u2013 winter, MAM\u00a0\u2013 spring, JJA\u00a0\u2013 summer, SON\u00a0\u2013 autumn); data corrected by the standard quantile method are in red, observed in blue and simulated in green; horizontal axis shows time aggregation (D1\u00a0\u2013 a\u00a0day, D10\u00a0\u2013 ten days, M1\u00a0\u2013 a\u00a0month, M3\u00a0\u2013 three months, M6\u00a0\u2013 six months, Y1\u00a0\u2013 a\u00a0year, Y5\u00a0\u2013 five years); the lines represent mean correlation coefficient, green (blue) area indicates an envelope of 50 (90)% of climate model simulations<\/h6>\n

Korekce kask\u00e1dovou kvantilovou metodou (obr.\u00a02<\/em>) p\u0159in\u00e1\u0161\u00ed v\u00fdrazn\u011b lep\u0161\u00ed v\u00fdsledky. Je vid\u011bt, \u017ee rozp\u011bt\u00ed chyb u\u00a0sr\u00e1\u017eek a\u00a0teploty je v\u00a0m\u011b\u0159\u00edtc\u00edch M1 a\u00a0Y1 podstatn\u011b ni\u017e\u0161\u00ed ne\u017e v\u00a0p\u0159\u00edpad\u011b korekce standardn\u00ed kvantilovou metodou (obr.\u00a01<\/em>). Bohu\u017eel u\u00a0agregac\u00ed D10, M3 a\u00a0M6 se ji\u017e tak v\u00fdrazn\u00e9 zlep\u0161en\u00ed nevyskytuje, jeliko\u017e tyto nebyly zahrnuty do kask\u00e1dov\u00e9 korekce. Je zaj\u00edmav\u00e9, \u017ee u\u00a0modelovan\u00e9ho odtoku doch\u00e1z\u00ed k\u00a0podstatn\u00e9mu sn\u00ed\u017een\u00ed zbytkov\u00e9 chyby pro prezentovan\u00e9 charakteristiky ve v\u0161ech \u010dasov\u00fdch m\u011b\u0159\u00edtc\u00edch.<\/p>\n

Obr.\u00a03<\/em> a\u00a04<\/em> ukazuj\u00ed m\u00edru korelace mezi veli\u010dinami p\u0159es r\u016fzn\u00e1 \u010dasov\u00e1 m\u011b\u0159\u00edtka ve \u010dty\u0159ech sezon\u00e1ch. Lze pozorovat, \u017ee korelace pro odtok a\u00a0sr\u00e1\u017eku je pozitivn\u00ed, naopak pro teplotu a\u00a0sr\u00e1\u017eku a\u00a0odtok a\u00a0teplotu je korelace sp\u00ed\u0161e negativn\u00ed. Ve v\u0161ech p\u0159\u00edpadech m\u00edra korelace roste s\u00a0\u010dasov\u00fdm m\u011b\u0159\u00edtkem. D\u00e1le je vid\u011bt, \u017ee z\u00e1kladn\u00ed pr\u016fb\u011bh korelace mezi sledovan\u00fdmi veli\u010dinami z\u016fst\u00e1v\u00e1 zachov\u00e1n pro simulovan\u00e1 i\u00a0korigovan\u00e1 data a\u00a0pou\u017eit\u00ed metody korekce systematick\u00fdch chyb m\u00e1 mal\u00fd vliv na korela\u010dn\u00ed strukturu. Nicm\u00e9n\u011b doch\u00e1z\u00ed ke zlep\u0161en\u00ed v\u00a0p\u0159\u00edpad\u011b pou\u017eit\u00ed kask\u00e1dov\u00e9 korekce, zejm\u00e9na u\u00a0korelace mezi sr\u00e1\u017ekou a\u00a0odtokem, kde lze pozorovat t\u011bsn\u011bj\u0161\u00ed vztah mezi korigovan\u00fdmi a\u00a0pozorovan\u00fdmi daty.<\/p>\n\"hanel-4\"<\/a>\n

Obr.\u00a04. Korelace mezi\u00a0sr\u00e1\u017ekou a\u00a0modelovan\u00fdm odtokem (naho\u0159e), mezi sr\u00e1\u017ekou a\u00a0teplotou (uprost\u0159ed) a\u00a0mezi teplotou a\u00a0modelovan\u00fdm odtokem (dole) pro jednotliv\u00e1 ro\u010dn\u00ed obdob\u00ed (DJF\u00a0\u2013 zima, MAM\u00a0\u2013 jaro, JJA\u00a0\u2013 l\u00e9to, SON\u00a0\u2013 podzim); \u010derven\u011b jsou zobrazena korigovan\u00e1 data kask\u00e1dovou kvantilovou metodou, mod\u0159e pozorovan\u00e1 data a\u00a0zelen\u011b simulovan\u00e1 data; vodorovn\u00e1 osa zna\u010d\u00ed \u010dasov\u00e1 m\u011b\u0159\u00edtka (D1\u00a0\u2013 den, D10\u00a0\u2013 deset dn\u016f, M1\u00a0\u2013 m\u011bs\u00edc, M3\u00a0\u2013 t\u0159i m\u011bs\u00edce, M6\u00a0\u2013 \u0161est m\u011bs\u00edc\u016f, Y1\u00a0\u2013 rok, Y5\u00a0\u2013 p\u011bt let); \u010d\u00e1ry zna\u010d\u00ed pr\u016fm\u011brnou korelaci, zelen\u00fd p\u00e1s rozp\u011bt\u00ed 50\u00a0% a\u00a0modr\u00fd p\u00e1s 90\u00a0% ze v\u0161ech model\u016f
\nFig. 4. Correlation between precipitation and modelled runoff (above), between precipitation and temperature (middle) and between temperature and modelled runoff (below) for seasons of year (DJF\u00a0\u2013 winter, MAM\u00a0\u2013 spring, JJA\u00a0\u2013 summer, SON\u00a0\u2013 autumn); data corrected by the cascade quantile method are in red, observed in blue and simulated in green; horizontal axis shows time aggregation (D1\u00a0\u2013 a\u00a0day, D10\u00a0\u2013 ten days, M1\u00a0\u2013 a\u00a0month, M3\u00a0\u2013 three months, M6\u00a0\u2013 six months, Y1\u00a0\u2013 a\u00a0year, Y5\u00a0\u2013 five years); the lines represent mean correlation coefficient, green (blue) area indicates an envelope of 50 (90)% of climate model simulations<\/h6>\n

Diskuse a\u00a0z\u00e1v\u011br<\/h2>\n

Z\u00a0prezentovan\u00fdch v\u00fdsledk\u016f je z\u0159ejm\u00e9, \u017ee korekce sr\u00e1\u017eek a\u00a0teploty pomoc\u00ed standardn\u00ed kvantilov\u00e9 metody nen\u00ed vhodn\u00e1 pro rutinn\u00ed vyu\u017eit\u00ed v\u00a0hydrologick\u00fdch simulac\u00edch, jeliko\u017e v\u00a0n\u011bkter\u00fdch ro\u010dn\u00edch obdob\u00edch vede ke zna\u010dn\u00e9 zbytkov\u00e9 chyb\u011b modelovan\u00e9ho odtoku ve v\u0161ech \u010dasov\u00fdch m\u011b\u0159\u00edtc\u00edch krat\u0161\u00edch ne\u017e rok, v\u010detn\u011b denn\u00edch hodnot, na kter\u00fdch byla korekce kalibrov\u00e1na. Podobn\u00e9 v\u00fdsledky prezentuje i\u00a0Teng\u00a0aj. [16]. Relativn\u011b nejv\u011bt\u0161\u00ed zbytkov\u00e9 chyby se vyskytuj\u00ed v\u00a0obdob\u00ed n\u00edzk\u00fdch pr\u016ftok\u016f (a\u00a0v\u00a0absolutn\u00edm vyj\u00e1d\u0159en\u00ed se zpravidla jedn\u00e1 o\u00a0n\u00edzk\u00e9 hodnoty).<\/p>\n

Dal\u0161\u00ed v\u00fdsledky (kter\u00e9 nejsou kv\u016fli rozsahu zahrnuty v\u00a0tomto \u010dl\u00e1nku) ukazuj\u00ed, \u017ee podobn\u00e9 v\u00fdsledky lze o\u010dek\u00e1vat i\u00a0u\u00a0jin\u00fdch povod\u00ed a\u00a0p\u0159i pou\u017eit\u00ed jin\u00fdch hydrologick\u00fdch model\u016f. V\u00fdchodiskem m\u016f\u017ee b\u00fdt vyu\u017eit\u00ed nejnov\u011bj\u0161\u00edch (pom\u011brn\u011b komplexn\u00edch) metod popsan\u00fdch Mehrotrou a\u00a0Sharmou [17], kter\u00e9 umo\u017e\u0148uj\u00ed korigovat\u00a0rozd\u011blen\u00ed pravd\u011bpodobnosti a\u00a0korela\u010dn\u00ed a\u00a0autokorela\u010dn\u00ed strukturu pro veli\u010diny v\u00a0r\u016fzn\u00fdch \u010dasov\u00fdch m\u011b\u0159\u00edtc\u00edch.<\/p>\n

\u00dasp\u011b\u0161nost t\u011bchto metod p\u0159i hydrologick\u00fdch simulac\u00edch nen\u00ed nicm\u00e9n\u011b zat\u00edm zn\u00e1ma. Druhou mo\u017enost\u00ed je vyu\u017eit\u00ed kombinac\u00ed jednoduch\u00fdch metod (nap\u0159.\u00a0p\u0159\u00edr\u016fstkov\u00e1 metoda, pop\u0159.\u00a0aplikovan\u00e1 v\u00a0r\u016fzn\u00fdch \u010dasov\u00fdch m\u011b\u0159\u00edtc\u00edch) v\u00a0kombinaci se stochastick\u00fdmi metodami umo\u017e\u0148uj\u00edc\u00edmi generov\u00e1n\u00ed dlouhodob\u00e9 variability [18].<\/p>\n

Pod\u011bkov\u00e1n\u00ed<\/h3>\n

Tento \u010dl\u00e1nek vznikl v\u00a0r\u00e1mci \u0159e\u0161en\u00ed projektu \u201eMo\u017enosti kompenzace negativn\u00edch dopad\u016f klimatick\u00e9 zm\u011bny na z\u00e1sobov\u00e1n\u00ed vodou a\u00a0ekosyst\u00e9my vyu\u017eit\u00edm lokalit vhodn\u00fdch pro akumulaci povrchov\u00fdch vod\u201c (TA04020501), kter\u00fd je spolufinancov\u00e1n Technologickou agenturou \u010cesk\u00e9 republiky, a\u00a0projektu \u201eP\u016fdn\u00ed a\u00a0hydrologick\u00e9 sucho v\u00a0m\u011bn\u00edc\u00edm se klimatu\u201c (1616549S) financovan\u00e9m Grantovou agenturou \u010cesk\u00e9 republiky.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"

Hydrological modelling is often used for assessment of climate change impacts on water resources. Inputs into the hydrological model are represented by precipitation and temperature based on simulations of climate models. <\/p>\n","protected":false},"author":8,"featured_media":2475,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[86],"tags":[584,586,585,587],"coauthors":[34,29],"class_list":["post-2525","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-hydraulics-hydrology-and-hydrogeology","tag-bias-correction","tag-climate-scenarios","tag-quantile-method","tag-runoff-modelling"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/posts\/2525","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/comments?post=2525"}],"version-history":[{"count":2,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/posts\/2525\/revisions"}],"predecessor-version":[{"id":30354,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/posts\/2525\/revisions\/30354"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/media\/2475"}],"wp:attachment":[{"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/media?parent=2525"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/categories?post=2525"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/tags?post=2525"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/coauthors?post=2525"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}