• 검색 결과가 없습니다.

Evaluation of Short-Term Drought Using Daily Standardized Precipitation Index and ROC Analysis

N/A
N/A
Protected

Academic year: 2021

Share "Evaluation of Short-Term Drought Using Daily Standardized Precipitation Index and ROC Analysis"

Copied!
10
0
0

로드 중.... (전체 텍스트 보기)

전체 글

(1)

** ⦽᧲ݡ⦺Ʊ ݡ⦺ᬱ Õᖅ⪹ĞŖ⦺ŝ ᕾᔍŝᱶ ([email protected])

Received April 8, 2013/ revised May 8, 2013/ accepted June 11, 2013

Copyright ⵑ 2013 by the Korean Society of Civil Engineers

 ǣŠ––’ǣȀȀ†šǤ†‘‹Ǥ‘”‰ȀͳͲǤͳʹ͸ͷʹȀ•…‡ǤʹͲͳ͵Ǥ͵͵ǤͷǤͳͺͷͳ

™™™Ǥ•…‡Œ‘—”ƒŽǤ‘”Ǥ”

ⴺ#ᢦⳂ#SPI⯾#ROC ⍂⛛ⴂ#ⴲⱧ㬚#ᢦᏮᆾℂⴖ#㦇ᆾ

କ஺ઽ ȵ৉෹૳ ȵ׌೾૽ ȵੲ୍෮

Yoo, Ji Young*, Song, Hoyong**, Kim, Tae-Woong***, Ahn, Jae-Hyun****

Evaluation of Short-Term Drought Using Daily Standardized Precipitation Index and ROC Analysis

ABSTRACT

The Standardized Precipitation Index (SPI) is widely applied to evaluate for meteorological droughts. However, the SPI is limited to capture a drought event with a short duration, expecially shorter than one month. In this study, we proposed a daily SPI (DSPI) as a way to overcome the limitation of the monthly SPI for drought monitoring. In order to objectively assess the ability of the drought reproduction of the DSPI, we performed a receiver operating characteristic (ROC) analysis using the quantified drought records from official reports, newspapers, etc. The results of ROC analysis showed that the DSPI has an ability to reproduce short-term drought compared with other indices. It also showed that the main cause of historical droughts was the shortage of rainfall accumulated during the time period less than 90 days compared with the rainfall of normal years.

Key words : Short-Term drought, Daily standardized precipitation index, ROC analysis

Ⅹಾ

SPI(Standardized Precipitation Index)۵ĥᔑᱩ₉ae݉⦹ʑভྙᨱʑᔢ⦺ᱢaྥᮥ⠪a⦹۵ߑŲჵ᭥⦹íᔍᬊࡹŁᯩ݅. ə్ӹᬵ݉

᭥ʑၹᮝಽᔑᱶࡹʑভྙᨱ1ݍᯕԕ᮹Ṉᮡḡᗮʑeᮥaḡ۵aྥᔍᔢᮥ⡍₊⦹ḡ༜⦹۵݉ᱱᮥaḡŁᯩ݅. ᅙᩑǍᨱᕽ۵ᬵ݉᭥ಽĥ ᔑࡹ۵SPI᮹݉ᱱᮥɚᅖ⦹ʑ᭥⦽ႊᦩᮝಽᯝ݉᭥SPI(DSPI)ෝᱽᦩ⦹ᩡ݅. ੱ⦽, DSPI᮹aྥᰍ⩥܆ಆᮥ~šᱢᮝಽ⠪a⦹ʑ᭥⧕ŝ Ñ᮹aྥၽᔾʑಾᮥᯕᬊ⦹ᩍROC ᇥᕾᮥᙹ⧪⦹ᩡ݅. ᯕෝၵ┶ᮝಽᬑญӹ௝᮹݉ʑaྥᮥᰍ⩥⦹۵ߑᱢᱩ⦽ḡᗮʑeᮥᱽ᜽⦹ᩡᮝ໑, aྥ᮹ၽᔾᮁྕෝđᱶ⦹۵ߑ⪽ᬊ⦹۵3aḡ॒ɪ(ᅕ☖, ᝍ⦽, ɚ⦽aྥ)ᄥDSPI᮹aྥᩩᅕ܆ಆᮥá☁⦹ᩡ݅. əđŝ3aḡ॒ɪᨱݡ⦽

ᰍ⩥܆ಆᯕᬑᙹ⦽aྥḡᗮʑeᮥđᱶ⦹ᩡ݅. ੱ⦽, ŝÑ2000֥ᯕ⬥ᨱၽᔾ⦽ᬑญӹ௝ݡᇡᇥ᮹aྥᮡ90ᯝ(3}ᬵ) ᯕԕ᮹٥ᱢvᙹప ᯕ⠪֥᮹ᙹᵡᨱእ⦹ᩍᇡ᳒⦹ʑভྙᨱၽᔾ⦽ʑᔢ⦺ᱢ⊂໕᮹݉ʑaྥᯙäᮝಽӹ┡ԍ݅.

áᔪᨕ ݉ʑaྥ, ᯝ݉᭥SPI, ROC ᇥᕾ

1. ᕽು

⩥ᰍ ᱥ ᖙĥᱢᮝಽ ʑ⬥ᄡ⪵ಽ ᯙ⦽ ݡȽ༉ aྥ᮹ ၽᔾ a܆ᖒᯕ ׳ᦥḡŁ ᯩᮝ໑, ᬑญӹ௝ࠥ ᩩ᫙۵ ᦥܩ݅. ᯕ۵ ᱥḡǍᱢ

᪉ӽ⪵⩥ᔢ᮹ᩢ⨆ᮝಽᬑญӹ௝۵vᙹప᮹qᗭ᪡޵ᇩᨕ᷾ၽపࠥⓍí᷾a⦹Łᯩᮝ໑, đǎaྥŝšಉࡽ⦝⧕ࠥ޵ᬒ⍅ḩ

ƒ–‡”‰‹‡‡”‹‰ սėॡ

(2)

äᯕ݅(Choi et al., 2008). ੱ⦽, aྥᮡə✚ᖒᔢ݅ෙᯱᩑᰍ⧕᪡

۵ݍญḥ⧪ᗮࠥa۱ญအಽ᜽·Ŗeᱢᮝಽᱶ⪶⦹í❱݉⦹ʑ

ᛞḡ ᦫ݅.

ᯝၹᱢᮝಽaྥᮡⓍí4aḡᬱᯙᮝಽǍᇥ⦹໑, ᯕ۵vᙹప ᮹ᇡ᳒ᮥᬱᯙᮝಽ⦹۵ʑᔢ⦺ᱢᯙaྥ, ḡ⢽ᙹၰḡ⦹ᙹ᮹

ᇡ᳒ᮥᬱᯙᮝಽ⦹۵ᙹྙ⦺ᱢaྥ, ☁᧲ᙹᇥ᮹ᇡ᳒ᮥᬱᯙᮝಽ

⦹۵׮ᨦᱢaྥ, ᬊᙹၰᱥಆŖɪ᮹ᱽ⦽ᮥᬱᯙᮝಽ⦹۵ᔍ⫭Ğ ᱽ⦺ᱢ aྥᯕ݅. ᯕ్⦽ aྥᮡ ᵝಽ aྥḡᙹᨱ ᮹⧕ ᔢ⫊ᮥ

❱݉⦹໑, ⩥ᰍʭḡ݅᧲⦽aྥḡᙹॅᯕ}ၽࡹᨩ݅. ᬑญӹ௝ࠥ

aྥᨱ ݡ⦽ ฯᮡ ᩑǍॅᯕ ḡᗮᱢᮝಽ ḥ⧪ࡹŁ ᯩᮝ໑(Yoo, 2006; Kim and Yoo, 2006; Yoo et al., 2008; Choi and Kim, 2010; Kim and Lee, 2011; Kim et al., 2011; Ahn et al., 2012),

⢽ᵡvᙹḡᙹ(Standardized Precipitation Index, SPI), ❭ນaྥ

ḡᙹ(Palmer Drought Severity Index, PDSI), ḡ⢽ᙹŖɪḡᙹ (Surface Water Supply Index, SWSI) ॒ᯕݡ⢽ᱢᮝಽ⪽ᬊࡹŁ

ᯩ݅.

ᯕ⃹ౝaྥᮥᱶ᮹⦹Ł❱݉⦹ʑ᭥⦽༊ᱢᮝಽ⪽ᬊࡹ۵݅᧲

⦽aྥḡᙹᵲᨱᕽSPI۵݅ෙḡᙹॅŝእƱ⦹ᩍaྥḡᙹෝ

ĥᔑ⦹۵ŝᱶᨱᕽᬵvᙹప ᯱഭอᮥᯕᬊ⦹ʑভྙᨱእƱᱢ

ᯱഭǍ⇶ᨱ ݡ⦽ ᱽ᧞᳑Õᯕ ᱢ݅. ੱ⦽ ݅᧲⦽ ᜽e⃺ࠥ(time scale)ෝŁಅ⦹ᩍĥᔑ⧁ᙹᯩʑভྙᨱ݉ʑၰᰆʑᱢᯙvᙹ

ᇡ᳒పᮥ⠪a⧁ᙹᯩ۵ᰆᱱᯕᯩ݅. ə్ӹSPI۵ᬵ݉᭥ಽ

ĥᔑࡹʑভྙᨱṈᮡʑe࠺ᦩ᮹ᙹྙ⦺ᱢᄡ࠺ᖒᮥŁಅ⦹ḡ

༜⦹۵⦽ĥaᯩ݅. ✚⯩1}ᬵᯕԕᨱ݉ʑaྥᔍᔢᯕၽᔾ⦹Ł

⧕ᗭࡹ۵ Ğᬑෝ ⡍₊⦹ḡ ༜⦹۵ Ğᬑa ᳦᳦ ၽᔾ⦽݅. ᩩෝ

ॅᨕ, 5ᬵ1ᯝⓑvᙹaၽᔾ⦽⬥, 5ᬵ2ᯝᇡ░6ᬵ29ᯝʭḡ

vᙹaԕญḡᦫ۵݅໕ᝍ⦽aྥ(severe drought)ᯕၽᔾ⧁ᙹ

ᯩᮝӹ, 6ᬵ30ᯝᨱੱ݅᜽ⓑvᙹaၽᔾ⦽݅໕ᬵ݉᭥ʑၹ᮹

aྥḡᙹᨱᕽ۵ ⧕ݚ aྥ ᔢ⫊ᮥ ⢽⇽ࡹḡ ༜⦹í ࡽ݅.

ᅙᩑǍᨱᕽ۵ʑ᳕᮹ᬵ݉᭥SPIaw۵ᱢᬊᔢ᮹⦽ĥෝɚᅖ⦹

Ł݉ʑaྥᮥ⠪a⦹ʑ᭥⦹ᩍᯝ݉᭥᮹٥ᱢvᙹపᮥʑၹᮝಽ

vᙹᇡ᳒ᇥᮥĥᔑ⦹ᩍᯝ݉᭥SPIෝ}ၽ⦹ᩡ݅. ੱ⦽, ݅᧲⦽

᜽e⃺ࠥᄥᯝ݉᭥SPIᨱݡ⦽ᱢᬊᔢ᮹⬉ᮉᖒᮥá☁⦹ʑ᭥⧕

ROC(Receiver Operating Characteristic) ᇥᕾᮥ ᝅ᜽⦹ᩡ݅.

ROC ༉⩶ᮡᵝಽʑᔢᇥ᧝ᨱᕽ⪶ශᩩᅕ᮹ᱶᖒᱢá᷾ᨱ⪽ᬊࡹ۵

ʑჶᯕ݅(Mason, 1982). Kim and Lee (2011)۵ᝅᱽaྥᔍಡ᪡

aྥḡᙹ᪡᮹ᱶపᱢᯙ⠪aෝ᭥⧕☖ĥᱢʑჶᯙROC ᇥᕾᮥ

ᱢᬊ⦽ၵᯩ݅. ᯕ్⦽ROC ᇥᕾᮥ᭥⧕ᕽᅙᩑǍᨱᕽ۵2000֥

ᯕ⬥(2000֥, 2001֥, 2008֥, 2009֥, 2012֥) ᬑญӹ௝ᨱᕽᝅᱽ

ၽᔾ⧩޹ ݉ʑaྥ ʑಾᮥ ᩍ్ aḡ aྥšಉ ᅕŁᕽ ၰ ᨙು

ᅕࠥᯱഭ॒ᮥ⪽ᬊ⦹ᩍᙹḲ⦹ᩡᮝ໑, ݅᧲⦽᜽e⃺ࠥෝaḡ۵

ᯝ݉᭥SPI᪡እƱᇥᕾ⦹ᩡ݅. ᯕ۵݉ʑaྥᮥ⠪a⧁ᙹᯩ۵

aྥḡᙹෝ}ၽ⦹Ł⪽ᬊ⦹ʑ᭥⦽ʑⅩᩑǍಽᕽᯝ݉᭥aྥḡᙹ

ෝ ʑၹᮝಽ ⦽ ⬉ᮉᱢᯙ aྥ ༉ܩ░ยᮥ a܆⍡ ⧁ äᯕ݅.

2. ᯝ݉᭥SPIෝᯕᬊ⦽aྥ✚ᖒᇥᕾ

2.1 ԧࢇଭ୨ଭࢫଵۚ଍SPI Թڄ

vᙹ᮹ ᇡ᳒ᮡ aྥ᮹ ᵝࡽ ᫵ᯙᯕ௝ ⧁ ᙹ ᯩᮝ໑, aྥ᮹

ᱶపᱢ ⠪aᨱ ⬉ŝᱢᮝಽ ᯕᬊࡹŁ ᯩ݅. ᷪ ⠪Ɂᱢᯙ ᙹᵡᮥ

ᖅᱶ⦹Ł, ᯕෝʑᵡᮝಽaྥᮥᱶ᮹⦹Łaྥᔍᔢ᮹ḡᗮʑe, ᝍࠥ, ၽᔾኩ॒ࠥᮥᱶప⪵⦽⬥, ᜽ĥᩕᇥᕾჶᮥᯕᬊ⦹ᩍaྥ᮹

✚ᖒᮥ ᇥᕾ⦹۵ äᯕ݅.

ᯝၹᱢᮝಽaྥ᮹݉ʑၰᰆʑᱢᯙ✚ᖒᮥᅖ⧊ᱢᮝಽ⧕ᕾ⦹

ʑ᭥⦹ᩍᬵᄥvᙹప᜽ĥᩕᮥ݅᧲⦽᜽e⃺ࠥ(3, 6, 9, 12}ᬵ

॒)᮹٥avᙹ᜽ĥᩕᮥǍᖒ⦹Ł, b᜽e⃺ࠥᄥvᙹᇡ᳒పᮥ

ĥᔑ⦹ᩍᬵ݉᭥SPI(Monthly SPI, MSPI)ෝᔑᱶ⦽݅(McKee et al., 1993). ə్ӹMSPIෝᯕᬊ⦹ᩍaྥ⠪aෝᙹ⧪⧁Ğᬑ, Ṉᮡḡᗮʑeᮥaḡ۵aྥ⩥ᔢᮥᖅ໦⦹۵ߑᨱ۵⦽ĥaᯩ݅.

ᯕᨱᅙᩑǍᨱᕽ۵ᯝ݉᭥SPIෝ}ၽ⦹ᩍᬵ݉᭥SPIaaḡ۵

ᱢᬊᔢ᮹ ⦽ĥᱱᮥ ᅕ᪥⧁ ᙹ ᯩ۵ ႊᦩᮥ ᱽ᜽⦹Łᯱ ⦽݅.

ᯝ݉᭥SPI᮹}ֱᮡᬵ݉᭥SPI᪡}ֱᱢᮝಽ۵࠺ᯝ⦹໑, SPI᮹ĥᔑᮥ᭥⧕٥avᙹ᜽ĥᩕᮥǍ⇶⧁Ğᬑᨱᯝvᙹప (daily precipitation)ᮥ݅᧲⦽᜽e⃺ࠥෝŁಅ⦹ᩍᯝᄥ٥av ᙹ᜽ĥᩕ(daily cumulative precipitation time series)ᮥǍᖒ⦽

݅. ᯝᄥ٥avᙹ᮹⪶ශᇥ⡍⧉ᙹෝ⇵ᱶ⦹ᩍ, ٥a⪶ශsᨱ⧕ݚ

⦹۵⢽ᵡᱶȽᄡపᮥ⇵ᱶ⦹໕, ᯝ݉᭥SPI(Daily SPI, DSPI)a

ࡽ݅. ᅙ ᩑǍᨱᕽ۵ Gamma ᇥ⡍᮹ ⪶ශၡࠥ⧉ᙹ(Eq. (1))᪡

٥ᱢᇥ⡍⧉ᙹ(Eq. (2))ෝ ᯕᬊ⦹ᩍ ᬑญӹ௝ 228} ⧪ᱶǍᩎᄥ

MSPI ၰDSPIෝ༉ࢱᔑᱶ⦹ᩡ݅. DSPIෝᔑᱶ⦹۵ŝᱶᨱᕽ

Gamma ᇥ⡍᮹ᱢ⧊ࠥáᱶᮥᙹ⧪⦽đŝ, 228}⧪ᱶǍᩎᄥ

DSPIෝ ĥᔑ⦹۵ ߑ ᱢᬊࡹᨕḥ ⅾ 365}᮹ Gamma ᇥ⡍⧉ᙹ

ᵲݡᇡᇥᯕᱢ⧊ࠥáᱶ(Chi-Square Test᪡Kolmogorov-Smirnov Test)ᮥ ☖ŝ⦹۵ äᮝಽ ⪶ᯙࡹᨩ݅.

ƄÞƖßá ć ķġÞĸß Î Þ ć ķ Ɩ ß

ĸ àÎ

ŒŸ— Þ à ćķ Ɩ ß

(1)

ŸÞƖß á

õ×Ɩ

ƄÞƖßƂƖá

õ×Ɩ

ć ķġÞĸß Î

Þ ćķ Ɩ ß

ĸ àÎ

ŒŸ—Þà ć ķ Ɩ ß ƂƖ

(2)

ᩍʑᕽ,

ķ

۵⇶⃺ๅ}ᄡᙹ(scale parameter),

ĸ

۵⩶ᔢๅ}ᄡ ᙹ(shape parameter),

ġÞĸß

۵qษ⧉ᙹ(Gamma function)ᯕ݅.

(3)

Table 1. Definition of the SPI Classes and Corresponding Event Probabilities

SPI interval SPI classes Occurrence probability (%) More than 2.00 Extreme wet 2.3

1.99 to 1.50 Severe wet 4.4 1.49 to 1.00 Moderate wet 9.2 0.99 to -0.99 Normal 68.2 -1.00 to -1.49 Moderate drought 9.2 -1.50 to -1.99 Severe drought 4.4 Less than -2.00 Extreme drought 2.3

Fig. 1. Thiessen Polygons to Transform Point Rainfalls() Into Fig. 2. Daily SPI and Monthly SPI for Gyeongsangnamdo (Time Table 1ᮡ᭥᪡zᮡŝᱶᮥ☖⧕ᔑᱶࡽDSPI᮹॒ɪᇥඹʑᵡ

ᮝಽ aྥ᮹ ᝍࠥෝ ❱݉⦹۵ ʑᵡᯕ ࡽ݅.

2.2 ෘ୨֜લ࣢ଵۚ଍SPI ॺ୨ࢫୡ૳নՑഠ

ᯝၹᱢᮝಽḡᩎᱢaྥšญෝ᭥⦹ᩍ⧪ᱶǍᩎᨱ঑ෙḡᯱℕ aaྥݡ᮲ၰ⦝⧕ᅖǍෝᙹ⧪⦹Łᯩᮝӹ, ᝅᱽaྥᮥ❱݉⦹۵ ߑ⪽ᬊࡹ۵aྥḡᙹ᮹ݡᇡᇥᮡvᬑš⊂ḡᱱᮥʑᵡᮝಽĥᔑ

ࡹŁ ᯩ݅. ə్အಽ ᅕ݅ ⬉ᮉᱢᯙ aྥšญෝ ᭥⧕ᕽ۵ ᝅᱽ

⧪ᱶǍᩎᄥvᙹ✚ᖒᮥၹᩢ⦹۵⧪ᱶǍᩎᄥaྥḡᙹ᮹ᔑᱶᯕ

⦥᫵⦹݅. ᯕෝ ᭥⧕ᕽ ᅙ ᩑǍᨱᕽ۵ ᬑญӹ௝ ʑᔢℎ ᔑ⦹᮹

57}vᬑš⊂ᗭᄥvᙹపᮥFig. 1ŝzᯕ❑ᖝ฾ᮥᯕᬊ⦹ᩍ

⧪ᱶǍᩎᄥ໕ᱢ⠪ɁᯝvᙹపᮝಽǍ⇶⦽⬥, ᯕෝʑၹᮝಽ⧪ᱶ Ǎᩎᄥ SPIෝ ĥᔑ⦹ᩡ݅.

ᬑญӹ௝۵vᬱᩢ࠺, vᬱᩢᕽ, ᕽᬙĞʑ, ∊ℎᇢࠥ, ∊ℎԉࠥ, Ğᔢᇢࠥ, Ğᔢԉࠥ, ᱥ௝ᇢࠥ, ᱥ௝ԉࠥḡᩎᮝಽⅾ9}᮹ݡǭᩎ, ᖙᇡᱢᮝಽ۵ⅾ228}᜽ǑǍಽǍᇥࢁᙹᯩ݅. ᅙᩑǍᨱᕽ۵

228} ⧪ᱶǍᩎᄥ aྥ᮹ ⠪aෝ ༊ᱢᮝಽ ᯝ vᙹపᮥ 6aḡ

᜽e⃺ࠥ(30ᯝ, 60ᯝ, 90ᯝ, 120ᯝ, 150ᯝ, 180ᯝ)᮹ᯝᄥ٥av ᙹ᜽ĥᩕᮥǍᖒ⦹Ł, Gamma ᇥ⡍᮹⪶ශᇥ⡍⧉ᙹ᮹٥a⪶ශs ᨱ⧕ݚ⦹۵⢽ᵡᱶȽᄡపᮥ⇵ᱶ⦹ᩍ⧪ᱶǍᩎᄥDSPI(SPI_D30, SPI_D60, SPI_D90, SPI_D120, SPI_D150, SPI_D180)ෝĥᔑ

⦹ᩡ݅. ੱ⦽᭥᮹6aḡ᜽e⃺ࠥᨱᔢ᮲⦹۵MSPI(SPI_M1, SPI_M2, SPI_M3, SPI_M4, SPI_M5, SPI_M6)ෝ⧪ᱶǍᩎᄥಽ

ĥᔑ⦹ᩡ݅. ⧪ᱶǍᩎᄥDSPIෝᬵᄥಽ⠪Ɂ⦽⬥MSPI᮹እƱ⦽

đŝ, DSPI᮹ ⠪Ɂŝ MSPI۵ ݡᇡᇥ 9} ݡǭᩎᨱᕽ እ᜘⦽

ᄡ࠺ᖒᮥaḡ۵äᮝಽӹ┡ԍ݅. ၹ໕, ࢱaྥḡᙹ᮹↽ݡၰ

↽ᗭ⊹۵MSPIᨱእ⦹ᩍDSPIᨱᕽ༉ࢱⓍíӹ┡ӹ, DSPIa

ɚ⦽᜖ᮅၰɚ⦽aྥᮥӹ┡ԕ۵ߑᯩᨕᅕ݅⬉ŝᱢᯙäᮝಽ

⪶ᯙࡽ݅(Fig. 2 ₙŁ).

✚⯩ࢱaྥḡᙹᨱݡ⦽↽ᗭ⊹(ɚ⦽aྥ)ෝá☁⦹໕, 6aḡ

᜽e⃺ࠥᨱ঑ෙDSPIᨱᕽ༉ࢱⓑᱩݡsᮥӹ┡ԕᨩ݅. ᩩෝ

ॅᨕ, SPI_D30᪡SPI_M1ᮥእƱ⧩ᮥĞᬑᨱ۵∊ℎԉࠥḡᩎᨱ ᕽaᰆⓑ₉ᯕ(0.956)ෝӹ┡ԕ۵äᮝಽ⪶ᯙࡹᨩᮝ໑, SPI_D60

᪡ SPI_M2ෝእƱ⧩ᮥĞᬑᨱ۵∊ℎᇢࠥḡᩎᨱᕽaᰆⓑ₉ᯕ (0.885), SPI_D90᪡SPI_M3ᮥእƱ⧩ᮥĞᬑᨱ۵ᱥ௝ᇢࠥḡᩎ ᨱᕽaᰆⓑ₉ᯕ(0.920), SPI_D120᪡SPI_M4, SPI_D150᪡

SPI_M5, SPI_D180᪡SPI_M6ෝእƱ⧩ᮥĞᬑᨱ۵༉ࢱᕽᬙĞ ʑ ḡᩎᨱᕽ ⓑ ₉ᯕ(0.676, 0.676, 0.545)ෝ ӹ┡ԙ݅.

ੱ⦽ࢱaྥḡᙹᨱݡ⦽↽ݡ⊹(ɚ⦽᜖ᮅ)ෝá☁⦹໕, 6aḡ

᜽e⃺ࠥᨱ ঑ෙ DSPIᨱᕽ ༉ࢱ ⓑ ᱩݡsᮥ ӹ┡ԕᨩᮝ໑, SPI_D30᪡SPI_M1, SPI_D60᪡SPI_M2ෝእƱ⧩ᮥĞᬑᨱ۵

(4)

Table 2. Correlation Coefficients Between Minimum Value of Daily SPI and Monthly SPI SPI_D30

(SPI_M1)

SPI_D60 (SPI_M2)

SPI_D90 (SPI_M3)

SPI_D120 (SPI_M4)

SPI_D150 (SPI_M5)

SPI_D180 (SPI_M6)

Gangwon Youngdong 0.556 0.768 0.827 0.866 0.860 0.881

Gangwon Youngseo 0.522 0.794 0.851 0.891 0.896 0.917

Seoul Gyeonggi 0.534 0.762 0.830 0.859 0.860 0.869

Chungcheongbukdo 0.546 0.772 0.840 0.874 0.880 0.890

Chungcheongnamdo 0.548 0.778 0.849 0.881 0.888 0.897

Gyeongsangbukdo 0.528 0.766 0.840 0.885 0.890 0.902

Gyeongsangnamdo 0.545 0.788 0.851 0.890 0.902 0.918

Jeollabukdo 0.539 0.785 0.854 0.880 0.887 0.902

Jeollanamdo 0.532 0.775 0.837 0.873 0.876 0.904

(a) May 2012 (b) June 2012

Fig. 3. Daily SPI and Monthly SPI for Seoul·Gyeonggi Region Ğᔢԉࠥḡᩎᨱᕽaᰆⓑ₉ᯕ(1.073, 1.110)ෝӹ┡ԕ۵äᮝಽ

⪶ᯙࡹᨩᮝ໑, SPI_D90᪡SPI_M3, SPI_D120᪡SPI_M4ෝእ Ʊ⧩ᮥĞᬑᨱ۵vᬱᩢ࠺ḡᩎᨱᕽaᰆⓑ₉ᯕ(0.704, 0.653), SPI_D150᪡SPI_M5ෝእƱ⧩ᮥĞᬑᨱ۵Ğᔢԉࠥḡᩎᨱᕽ

aᰆⓑ₉ᯕ(0.603), SPI_D180᪡SPI_M6ᮥእƱ⧩ᮥĞᬑᨱ۵

ᱥ௝ԉࠥ ḡᩎᨱᕽ aᰆ ⓑ ₉ᯕ(0.529)ෝ ӹ┡ԙ݅.

Table 2۵ࢱaḡSPI᮹ᔢšĥᙹෝᱶญ⦽äᮝಽ, ᩍʑᕽ

MSPI᪡እƱࡹᨕḥDSPI۵ᬵᄥ↽ᘀs(ᮭ᮹s)ᮥᯕᬊ⦹ᩡ݅.

ə đŝ, ✚⯩ 90ᯝ ᯕ⦹ ʑe ࠺ᦩ᮹ ٥ᱢ vᙹపᮥ ᯕᬊ⦹ᩍ

ĥᔑ⦽SPI_D30(vs SPI_M1), SPI_D60(vs SPI_M2), SPI_D90 (vs SPI_M3)ᮡԏᮡᔢšᖒᮥӹ┡ԕᨩᮝӹ, 90ᯝᮥⅩŝ⦹۵

ʑe࠺ᦩ᮹٥ᱢvᙹపᮥᯕᬊ⦹ᩍĥᔑ⦽SPI_D120(vs SPI_M4), SPI_D150(vs SPI_M5), SPI_D180(vs SPI_M6)ᮡᔢݡᱢᮝಽ

ᔢšᱶࠥaᱱ₉Ⓧíӹ┡ӹ۵äᮥ⪶ᯙ⧁ᙹᯩ݅. ᯕ۵ᰆʑe᮹

٥ᱢ vᙹపᮥ ᯕᬊ⦹ᩍ ᰆʑ aྥᮥ ⠪a⧁ Ğᬑᨱ۵ ၹऽ᜽

ࢱaḡSPIෝǍᇥ⦹ᩍᔍᬊ⧕᧝⧁⦥᫵۵ᨧḡอ, 3}ᬵၙอ᮹

݉ʑaྥᮥ⠪a⧁Ğᬑᨱ۵ᯝ݉᭥SPIෝၹऽ᜽á☁⧕᧝⧁

⦥᫵a ᯩᮥ äᯕ݅.

ᩩෝॅᨕ, ḡӽ1973֥4ᬵvᬱࠥv෪᜽aྥᮥSPI_M1 (0.542)ᮥᯕᬊ⦹ᩍ⠪a⦽݅໕, ⢽1᮹Ǎᇥᨱ঑௝ᱶᔢ(Normal) ᔢ┽௝Ł❱݉ࡽ݅. ၹ໕, ࠺ᯝ᜽ᱱᯙ1973֥4ᬵ᮹ᯝᄥSPI_D30 ɚ⦽(Extreme) aྥᔢ┽ᯕ໑, 4ᬵ24ᯝᨱ۵4ᬵᵲ↽ݡs(0.721) ᯕĥᔑࡹᨕᱶᔢ(Normal) ᔢ┽ಽ⪶ᯙࡹᨩ݅. ᯕ۵4ᬵᵲᙽᨱ

ԕฑvᙹపᮝಽᯙ⧕4ᬵⅩʭḡḡᗮࡹᨩ޹aྥᔢ┽a⧕iࡹᨕ

MSPI۵ᱶᔢ(Normal) ᔢ┽ᯙäᮝಽ⢽⩥ࡽäᯕӹ, ᝅᱽaྥ᮹

᜽᯲ၰ⧕i᮹᜽ᱱᮥᱶ⪶⦹í❱݉⦹ʑ᭥⧕ᕽ۵DSPIෝᯕᬊ⦽

aྥ ༉ܩ░ย᮹ ⦥᫵ᖒᮥ ᰍ⪶ᯙ ⧁ ᙹ ᯩ݅.

ᅙ ᩑǍᨱᕽ ᱽᦩ⦽ DSPIෝ ⪽ᬊ⦹۵ aྥ⧕ᕾ᮹ ᱢᬊᖒᮥ

ᬑญӹ௝ᨱᕽᝅᱽಽၽᔾ⦽2012֥ᅥaྥᮥݡᔢᮝಽá☁⦹ᩡ

݅. 2012֥5ᬵ1ᯝᇡ░6ᬵ27ᯝʑe࠺ᦩ, ᱥǎᱢᮝಽԕฑ

vᙹపᮡ⠪֥(248mm)᮹28.7%(71.2mm)ᨱᇩŝ⦽ᙹᵡᯕᨩᮝ

(5)

(a) June 23 (b) June 24 (c) June 25

(d) June 26 (e) June 27 (f) June 28

(g) June 29 (h) June 30

໑, ᯕಽᯙ⧕ᅥaྥᯕၽᔾ⦽ၵᯩ݅. ✚⯩, ᕽᬙḡᩎᨱᕽ۵

࠺ᯝ⦽ʑe࠺ᦩᱥǎᨱᕽaᰆᱢᮡvᙹప(10.6mm)ᯕԕಙ݅Ł

ᅕŁࡹᨩ݅(Ahn, et al., 2012). Fig. 3ᮡ2012֥ᅥaྥᯕၽᔾ⦽

5ᬵŝ6ᬵᨱݡ⦽ᕽᬙĞʑḡᩎ᮹MSPI ၰDSPIෝӹ┡ԕ໑, ᩍʑᕽၶᜅ⥭೐ᮡ⧕ݚᬵ᮹DSPIᯕŁ, ၶᜅ⥭೐ԕ⢽ʑࡽᱱᮡ

⧕ݚ ᬵ᮹ MSPIᯕ݅.

Fig. 3(a)᪡Fig. 3(b)᮹đŝෝእƱ⦹ʑᨱᦿᕽvᙹၽᔾ✚ᖒ

ᮥⓍíࢱaḡಽǍᇥ⧕᧝⦽݅. ᷪ, ⦽ݍԕԕvᙹaԕญḡ

ᦫ۵ Ğᬑ᪡ ⧕ݚ ᬵ ԕ ✚ᱶ ʑe ࠺ᦩ ∊ᇥ⦽ vᙹa ԕญ۵

Ğᬑಽӹڽ݅. ᝅᱽḡӽ2012֥5ᬵᮡᱥᯱ᮹Ğᬑᯕ໑, 6ᬵᮡ

⬥ᯱ᮹Ğᬑಽᕽ2012֥6ᬵ29ᯝŝ30ᯝᨱÑℱ30mmⴇ110mm ᮹vᙹaၽᔾ⦹ᩡ݅. ᷪ, ᱥᯱ᮹Ğᬑᨱ۵MSPI᪡DSPI۵༉ࢱ

ɚ⦽ aྥᔢ┽ಽ ⢽⩥ࡹŁ ᯩᮝӹ(Fig. 3(a)), ⬥ᯱ᮹ Ğᬑᨱ۵

DSPIa⢽⩥⦹۵ɚ⦽aྥᔢ┽ෝMSPI۵ɚ⦽aྥᔢ┽ෝ⢽⩥

(6)

Table 3. ROC Model

Previous drought records Drought Non-drought Drought Index

(SPI)

Drought Hit (H) False (F) Non-drought Missing (M) Negative hit (N)

Fig. 5. ROC Curve

⦹۵ ߑ⦽ĥa ᯩ۵äᮝಽ ⪶ᯙࡽ݅. ᯕ۵ MSPIෝ ᔑᱶ⦹۵

ߑ ᯩᨕ, 6ᬵ᮹ ᜖ᮅʑe(2012֥ 6ᬵ 29ᯝŝ 30ᯝ: 30mmⴇ 110mm vᙹၽᔾ) ࠺ᦩᨱԕฑᯝvᙹపᯕ⡍⧉ࡽᬵvᙹపᮥ

ᯕᬊ⦹ᩍMSPIaĥᔑࡽäᯕʑভྙᯕ݅. ə్အಽ⧪ᱶǍᩎᄥ

DSPI۵aྥᮥḥ݉⦹Ł⠪a⦹۵ߑᯩᨕ, ʑ᳕ᬵ݉᭥aྥḡᙹ

ෝᯕᬊ⦽aྥ⠪aႊჶŝእƱ⦹ᩍᅕ݅ᔢᖙ⦽᜽e⃺ࠥ(ᯝ݉᭥) ᮹aྥ༉ܩ░ยᯕa܆⦹݅. ੱ⦽, DSPIෝ⪽ᬊ⦽aྥ⠪aෝ

☖⧕Õ᳑⦽ᔢ┽a᜖ᮅ⦽ᔢ┽ಽᄡ⪵ࡹ۵aྥ⧕ᗭŝᱶ᮹༉ܩ

░ยᯕ a܆⦹݅(Fig. 4 ₙŁ). ᯕ۵ ʑ᳕᮹ MSPIෝ ⪽ᬊ⦹ᩍ

aྥᮥšญၰq᜽ෝᙹ⧪⦹۵ߑᯩᨕၽᔾ⦹۵ᱢᬊᔢ᮹⦽ĥᱱ

ᮥ ɚᅖ⦹ʑ ᭥⦽ ႊᦩᮝಽ DSPI᮹ ⪽ᬊᖒᮥ ᷾໦⧁ ᙹ ᯩ۵

⦹ӹ᮹ ~šᱢᯙ ɝÑಽ ❱݉ࡽ݅.

3. ROCᇥᕾᮥ☖⦽݉ʑaྥᰍ⩥܆ಆ⠪a

ᅙᩑǍᨱᕽ۵DSPI᮹aྥᰍ⩥܆ಆᮥᅕ݅~šᱢᮝಽ⠪a⦹

ʑ᭥⧕ᕽROC(Receiver Operating Characteristics) ᇥᕾᮥᙹ⧪

⦹ᩡ݅. ᯕ۵ ᬑญӹ௝ ݡǭᩎᄥ ݉ʑaྥᮥ ⠪a⦹۵ ߑ ᯩᨕ

ŝÑᨱᝅᱽၽᔾ⦽aྥʑಾᮥʑၹᮝಽDSPI᮹aྥᰍ⩥܆ಆᮥ

⠪a⦹۵ äᯕ݅.

3.1 ROC ंজԹڄࢫࡦ෴ড୨

ROC ᇥᕾᮡ ᵝಽ ʑᔢᇥ᧝ᨱᕽ ⪶ශᩩᅕ᮹ ᱶᖒᱢ á᷾ᨱ

⪽ᬊࡹ۵ʑჶᮝಽ, Kim and Lee (2011)۵ᝅᱽaྥᔍಡ᪡aྥ

ḡᙹ᪡᮹ ᱶపᱢᯙ ⠪aෝ ᭥⧕ ROC ᇥᕾᮥ ᱢᬊ⦽ ၵ ᯩ݅.

ᅙᩑǍᨱᕽ۵aྥḡᙹෝ⪽ᬊ⦽ᝅᱽaྥ᮹ᰍ⩥܆ಆᮥ⠪a⦹

ʑ᭥⧕ᕽTable 3ŝzᯕROC ༉⩶ᮥᖅᱶ⦹ᩡ݅. 2000֥ᯕ⬥ᨱ

ʑಾࡽaྥၽᔾ⩥⫊ᮥ᳑ᔍ⦹ᩍ“š⊂đŝ(Observed value)”ಽ

ᖅᱶ⦹Ł, ⧕ݚ᜽ᱱᮥݡᔢᮝಽĥᔑࡽaྥḡᙹෝ“ᩩ⊂đŝ (Prediction value)”ಽᇥඹ⦹ᩍ“aྥ(ⷂ)”ŝ“aྥ(×)”ᨱ᮹⧕ᕽ

bb 2aḡᦊ ᇥඹ⦹ᩡ݅.

ᅙ ᩑǍᨱᕽ ᖅᱶ⦽ ༉⩶ᮡ ᝅᱽ aྥᯕ ၽᔾ⦽݅Ł ʑಾࡽ

Ğᬑᨱaྥḡᙹ(SPI)ᨱᕽࠥaྥᯕၽᔾ⧩݅Łӹ┡ӹ໕“ᖒŖ (Hit, H)”, ə౨ḡᦫ݅໕“᯹༜ࡽĞŁ(Missing, M)”ಽӹ┡ԕᨩ

݅. ၹ໕, ᝅᱽᔍÕᯕၽᔾ⦹ḡᦫᮥĞᬑᩩ⊂đŝᨱᕽᔍÕᯕ

ၽᔾ⦽݅໕“ᝅ➉(False, F)”, ə౨ḡᦫ݅໕“ᮭ᮹ᖒŖ(Negative hit, N)”ᮝಽӹ┡ԕᨩ݅. ᯕᵲᖒŖ(H)ŝᮭ᮹ᖒŖ(N)᮹Ğᬑ۵

bbₙ᮹sᮝಽ❱݉⦹Ł, ᝅ➉᪡᯹༜ࡽĞŁ۵ÑḴ᮹sᮝಽ

đᱶ⦽݅.

Table 3ŝ zᯕ ᇥඹࡽ 4aḡ᮹ Ǎᇥᮥ ☖⧕ Eqs. (3) and (4)ෝᯕᬊ⦹ᩍᱢᵲශ(Hit Rate, HR)ŝእᱢᵲශ(False Alram Rate, FAR)ᮥbbĥᔑ⦹໑, Fig. 5᪡zᮡROC ᳭⢽ĥԕᨱ

⦹ӹ᮹ᱱᮝಽ⢽⩥⧉ᮝಽ៉ROC łᖁᮥǍᖒ⧁ᙹᯩ݅. ROC s᮹ჵ᭥۵0ᨱᕽ1 ᔍᯕ᮹sᮥaḡ໑, łᖁᦥ௹᮹໕ᱢ(Area Under Curve, AUC)ᯕᔑᱶࡽ݅. ੱ⦽᪥ᄞ⦽ᩩ⊂ᯝভHR=1, FAR=0 ᯕအಽROC łᖁᯕ᳭⊂ᔢ݉ᨱaʭᯕ᭥⊹⧁ᙹಾaྥ

ḡᙹ᮹ ᰍ⩥܆ಆ ᬑᙹ⦹݅Ł ❱݉⧁ ᙹ ᯩ݅(Fawcett, 2006).

¡« á ¡îÞ¡ â¦ß

(3)

Ÿš« á Ÿîޟ â§ß

(4)

3.2 ਓ୪րՋԧࢇ׆ߧ৤ுࢫۚ׆ԧࢇඌԧ

ROC ᇥᕾᮥᙹ⧪⦹ʑ᭥⧕ᕽ۵aྥḡᙹ᮹ᰍ⩥܆ಆᮥ❱݉⧕

ᵥɝÑෝᙹḲ⧕᧝⦹໑, ᅙᩑǍᨱᕽ۵ᩍ్aḡaྥšಉᅕŁᕽ (aྥʑಾ᳑ᔍᅕŁᕽ(MLIT, 2001; 2005), ┽႒᜽aྥ႒ᕽ(Tae Baek City, 2009)) ၰᨙುᅕࠥᯱഭ॒ᮥᙹḲ⦹ᩍ2000֥ᯕ⬥᮹

ɚ⦽݉ʑaྥ(2000֥(7Õ), 2001֥(45Õ), 2008֥(45Õ), 2009

֥(13Õ), 2012֥(3Õ)) ʑಾᮥ ᙹḲ⦹ᩡ݅. Table 4۵ ᙹḲࡽ

ᯱഭᵲ2009֥aྥᔢ⫊ᨱݡ⦽ᯱഭෝᩩ᜽ಽᅕᩍᵡ݅. ੱ⦽

DSPIෝᯕᬊ⦽aྥᔢ┽᮹ᮁྕෝ❱݉⦹۵ߑʑᵡᯕࡹ۵ᱩ݉ᱱ

(7)

Table 4. The Collection of Previous Records for Extreme Drought in 2009

Year Area Starting date Finishing date Contents Duration(Day)

2009

Hongcheon 01-01 01-31 Rainfall(0.4mm) 31

Incheon 01-01 01-31 Rainfall(1.2mm) 31

Inje 01-01 01-31 Rainfall(1.7mm) 31

Taebaek 01-06 01-14 Restriction on water supply: 1,2 step

(3ton/day, Decrease 530%) 9 Taebaek 01-15 04-02 Restriction on water supply : 3step(3ton/day Decrease 50%) 87 The west coast of

Chungcheong 01-18 01-18 Daejun(8), Geumsan(7.5), Boryeong(7), Yesan(4), Cheonan(1.5) 1

Gangwon, Youngseo 03-03 03-03 Gangwon 13cm(Snow) 1

Korea south·central area 03-03 03-03 Heavy snow in Korea south·central area 1 Korea 03-20 03-21 Seoul(41.5), Jinju(100.5), Daegu(11.5), Chuncheon(40.5),

Taebaek(48.8), 2

Busan 08'12-03 01-07 Rainfall(9.5mm: 20% of rainfall compare to last year) 35

Busan 08'12-25 01-18 Drought warning 24

Table 5. Cut-off Value of ROC Curve Degree DSPI

Moderate Dry Severe Dry Extreme Dry

-1.49 to -1.00 -1.99 to -1.50 Less than -2.00

SPI_D30 SPI_D30(MD) SPI_D30(SD) SPI_D30(ED)

SPI_D60 SPI_D60(MD) SPI_D60(SD) SPI_D60(ED)

SPI_D90 SPI_D90(MD) SPI_D90(SD) SPI_D90(ED)

SPI_D120 SPI_D120(MD) SPI_D120(SD) SPI_D120(ED)

SPI_D150 SPI_D150(MD) SPI_D150(SD) SPI_D150(ED)

SPI_D180 SPI_D180(MD) SPI_D180(SD) SPI_D180(ED)

(cut-off value)ᮡTable 5᪡zᯕ3aḡaྥ॒ɪᮝಽđᱶ⦹ᩡ݅.

݅᧲⦽᜽e⃺ࠥෝŁಅ⦽DSPI᮹ᱢᵲශ(Hit Rate, HR)ŝእᱢ ᵲශ(False Alram Rate, FAR)ᮥbbĥᔑ⦹Ł, ROC łᖁᦥ௹᮹

໕ᱢ(AUC)ᮥ ĥᔑ⦹ᩡ݅. Figs. 6(a), 6(b) and 6(c)۵ 3aḡ

aྥ᮹ ॒ɪᄥ aྥḡᙹ᮹ ᱢᵲශᮥ ӹ┡ԕ໑, b əฝ ᔢ݉ᨱ۵

baྥḡᙹ(SPI_D30, SPI_D60, SPI_D90, SPI_D120, SPI_D150, SPI_D180)ᨱݡ⦽bb᮹ᱢᵲශᯕ⢽ʑࡹᨕᯩŁFig. 6(d)۵

3aḡ aྥ ॒ɪᄥ AUCෝ ӹ┡ԙ݅.

ᯝၹᱢᮝಽROC łᖁᮡ᳭⊂ᔢ݉ᨱaʭᯕ᭥⊹⧁ᙹಾ, ੱ۵

ᱢᵲශsᯕ1.0ᨱaʭᬙᙹಾaྥḡᙹ᮹ᰍ⩥܆ಆᬑᙹ⦹݅Ł

⧕ᕾࡽ݅. ᅙᩑǍᨱᕽ۵ᅕ☖aྥᨱᕽ۵ᱢᵲශ0.76ŝ0.74ಽ

⪶ᯙࡹ۵SPI_D30᪡SPI_D150, ᝍ⦽aྥᨱᕽ۵ᱢᵲශ0.83ŝ

0.81ಽ⪶ᯙࡹ۵SPI_D90ŝSPI_D60, ɚ⦽aྥᨱᕽ۵ᱢᵲශ

0.68ಽ⪶ᯙࡹ۵SPI_D30aDSPIෝᯕᬊ⦹ᩍaྥᮥᰍ⩥⦹۵

ߑ ᯩᨕ ᅕ݅ ⬉ᮉᱢᯙ äᮝಽ ❱݉ࡽ݅.

ᯕ⃹ౝ, ᝅᱽ݉ʑaྥ᮹⠪aෝ༊ᱢᮝಽ⦽݅໕, ʑ᳕ᨱ3}ᬵ

ᯕ࠺⠪Ɂvᙹపᮥʑၹᮝಽᔑᱶ⦽SPI_M3ŝ޵ᇩᨕSPI_D30

᪡ SPI_D60ŝ zᮡ ᅕ݅ Ṉᮡ ʑe᮹ ٥ᱢvᙹపᮥ ᯕᬊ⦹ᩍ

ĥᔑࡽDSPIෝ⇵aಽá☁⧁⦥᫵aᯩ݅. ၹ໕, ᩩᅕ᮹ᱢᵲශᯕ

ԏíӹ┡ӽaྥḡᙹෝᯕᬊ⦹ᩍ݉ʑaྥᮥ ⠪a⧁Ğᬑᨱ۵

ᔢݡᱢᮝಽaྥ❱݉᮹ᝁ഑ࠥaᱡ⦹ࢁa܆ᖒᯕᯩᮝအಽ, ݉ʑ aྥ ⠪aෝ ᭥⧕ aྥḡᙹෝ ᖁᱶ⧁ Ğᬑᨱ۵ ᖙᝍ⦽ ᵝ᮹a

⦥᫵⦹݅.

3.3 କตԧࢇ஺৤(EDI)ࠜଲ૳෉DSPIଭୡ૳নඌԧ

ᦿᱩᨱᕽᔑᱶ⦽DSPI᮹ᱢᬊᖒᮥá☁⦹ʑ᭥⧕ᕽʑ᳕ᨱ

}ၽࡽ aྥḡᙹ ᵲ ᮁᯝ⦽ ᯝ ݉᭥ aྥḡᙹᯙ ᮁ⬉aྥḡᙹ (Effective Drought Index, EDI)᪡እƱ⦹ᩡ݅. Byun and Wilhite (1999)ᨱ঑෕໕, ᯝၹᱢᮝಽEDI۵✚ᱶᯝ᮹“ᮁ⬉٥ᱢvᙹప (EP)”ᮥ ĥᔑ⦽ ⬥, ə ԁ᮹ ʑ⬥ ⠪Ɂᯙ “ᮁ⬉vᙹvᙹప”ŝ

እƱ⦹ᩍᇡ᳒ప⪚ᮡⅩŝపᮥĥᔑ⦽⬥ᯕෝ⢽ᵡ⪵⦽݅. ᩍʑᕽ,

“ᮁ⬉٥ᱢvᙹప”ᯕ௡ ᯕᱥ 365ᯝ(⪚ᮡ ə ᯕᔢ) ࠺ᦩ ٥ᱢࡽ

(8)

(a) Moderate Dry (MD) (b) Severe Dry (SD)

(c) Extreme Dry (ED) (d) AUC for Each Degree of Dryness Fig. 6. The results of ROC Analysis (HR, FAR, AUC)

Table 6. The Results of ROC Score for Daily SPI and EDI in 9 Regions Degree

Region

Moderate Dry(MD) Severe Dry(SD) Extreme Dry(ED) SPI_D30 SPI_D60 SPI_D90 EDI SPI_D30 SPI_D60 SPI_D90 EDI SPI_D30 SPI_D60 SPI_D90 EDI South Korea 0.761 0.723 0.530 0.663 0.729 0.807 0.825 0.648 0.622 0.546 0.578 0.496 Gangwon Youngdong 0.589 0.559 0.532 0.531 0.812 0.674 0.689 0.571 0.618 0.588 0.522 0.479 Gangwon Youngseo 0.621 0.516 0.577 0.581 0.656 0.519 0.506 0.498 0.660 0.541 0.593 0.512 Seoul Gyeonggi 0.621 0.512 0.531 0.502 0.576 0.511 0.620 0.507 0.601 0.582 0.588 0.502 Chungcheongbukdo 0.653 0.516 0.443 0.453 0.574 0.498 0.667 0.449 0.857 0.727 0.654 0.887 Chungcheongnamdo 0.925 0.699 0.707 0.571 0.876 0.893 0.767 0.977 0.947 0.886 0.775 1.000 Gyeongsangbukdo 0.837 0.590 0.646 0.492 0.842 0.794 0.710 0.408 1.000 0.966 0.855 0.886 Gyeongsangnamdo 0.606 0.539 0.524 0.581 0.748 0.630 0.503 0.465 0.797 0.649 0.582 0.491 Jeollabukdo 0.680 0.556 0.528 0.577 0.838 0.547 0.521 0.459 0.746 0.775 0.566 0.490 Jeollanamdo 0.745 0.775 0.567 0.724 0.764 0.825 0.820 0.685 0.589 0.508 0.551 0.488

(9)

vᙹపᯕ໑, 365ᯝ࠺ᦩvᙹప⧊ᔑᮥǍ⧁ভ᪅௹ᱥvᙹపᮡ

ᯕၙฯᯕᗭᝅࡹᨩᮭᮥaᵲᯙᯱෝ☖⧕ၹᩢ⦹Ł࠺᜽ᨱ↽ɝ

vᙹపᮡฯᮡᇡᇥԉᦥᯩᮭᮥŁಅ⦹ᩍ}ၽࡽaྥḡᙹᯕ݅.

ᅙᩑǍᨱᕽ۵ᇡĞݡ⦺ƱႊᰍʑᔢᩑǍᝅ(http://atmos.pknu.

ac.kr/~mdr/)ᨱᕽᱽŖ⦹۵EDI ĥᔑᗭ⥥✙ᭉᨕෝ⪽ᬊ⦹ᩡᮝ໑, SPI_D30, SPI_D60, SPI_D90ŝ ⧪ᱶǍᩎᄥ EDIෝ ⪽ᬊ⦹ᩍ

ᬑญӹ௝2000֥ᯕ⬥ᨱၽᔾ⦽ɚ⦽݉ʑaྥ(2000֥, 2001֥, 2008֥, 2009֥, 2012֥) ʑಾᮥၵ┶ᮝಽROC ᇥᕾᮥᙹ⧪⦹ᩡ

݅. ə đŝ ᱥǎᱢᮝಽ۵ ᅕ☖ aྥ(MD)ᮥ ᰍ⩥⧁ Ğᬑᨱ۵

SPI_D30᮹ᱢᵲශ(0.761)ᯕ׳íӹ┡ԍᮝ໑, ၹ໕SPI_D90᮹

ᱢᵲශ(0.530)ᯕaᰆԏᮡäᮝಽ⪶ᯙࡹᨩ݅. ੱ⦽ᝍ⦽aྥ

(SD)ᨱᕽ۵SPI_D90᮹ᱢᵲශ(0.825)ᯕaᰆ׳ŁEDI᮹ᱢᵲශ (0.648)ᮡ aᰆ ԏᦹᮝ໑, ɚᝍ⦽ aྥ(ED)ᨱᕽ۵ SPI_D30᮹

ᱢᵲශ(0.622)ᯕ aᰆ ׳Ł EDI᮹ ᱢᵲශ(0.496)ᮡ aᰆ ԏᮡ

äᮝಽ ⪶ᯙࡹᨩ݅(Table 6 ₙŁ).

4. đು

ᅙᩑǍᨱᕽ۵ᬑญӹ௝᮹⧪ᱶǍᩎᮥ9}᮹ݡǭᩎᮝಽǍᇥ⦹

ᩍʑᔢ⦺ᱢaྥ᮹}ֱᮥᱢᬊ⦹ᩍ, ʑ᳕᮹ᬵ݉᭥ʑၹᮝಽ

ĥᔑࡹᨕḥSPIᨱݡ⦽ᱢᬊᔢ᮹⦽ĥෝɚᅖ⧁ᙹᯩ۵ݡᦩᮝಽ

ᯝ݉᭥SPIෝᱽᦩ⦹ᩡ݅. ੱ⦽6}ᬵၙอ᮹݉ʑaྥᮥ⠪a⧁

Ğᬑᯝ݉᭥SPI᮹⦥᫵ᖒၰ⪽ᬊࠥෝ~šᱢᮝಽá☁⦹Łᯱ, ᝅᱽ ŝÑᨱ ၽᔾ⦽ aྥᔍಡෝ ᙹḲ⦹ᩍ እƱ·ᇥᕾ⦹ᩡ݅. ə

đŝ, ݉ʑᱢᯙ⊂໕ᨱᕽʑ᳕᮹ᬵ݉᭥ʑၹ᮹aྥḡᙹෝᯕᬊ⦹

ᩍaྥᮥ❱݉⧁Ğᬑᨱၽᔾa܆⦽ᱢᬊᔢ᮹⦽ĥෝᅙᩑǍᨱᕽ

ᱽᦩ⦽ᯝ݉᭥SPIෝ⪽ᬊ⦹ᩍɚᅖᯕa܆⦹ᩡ݅. ᯕ۵ᝅᱽಽ

ၽᔾ⦽aྥᔍᔢ᮹ḥ⧪ ᱶࠥෝᩑᗮᱢᮝಽᯝᄥ༉ܩ░ย⦹ᩍ

aྥᮥqḡ⦽⬥, aྥᯕ⧕iࡹ۵ᱥŝᱶᮥᯝ݉᭥ಽq᜽⧁

ᙹ ᯩí ࡹᨩ݅.

޵ᇩᨕᯝ݉᭥SPI᮹┡ݚᖒᮥá☁⦹ʑ᭥⧕ᕽ2000֥ᯕ⬥ᨱ

ၽᔾ⦽ ŝÑ aྥʑಾᮥ ᙹḲ⦽ ⬥, ᯕෝ ʑၹᮝಽ ᬑญӹ௝᮹

ݡǭᩎᄥ ݉ʑaྥᨱ ݡ⦹ᩍ ݅᧲⦽ ᜽e⃺ࠥᨱ ঑ෙ ᯝ ݉᭥

SPI᮹aྥᰍ⩥܆ಆᮥ⠪a⦹ᩡ݅. ᅙᩑǍᨱᕽ۵ROC ༉⩶ᮥ

⪽ᬊ⦹ᩍaྥᰍ⩥᮹ᱢᵲශᮥĥᔑ⦹ᩡᮝ໑, əđŝaྥၽᔾ᮹

ᮁྕෝđᱶ⦹۵ᱩ݉ᱱ᮹ᄡ⪵ᨱ঑௝aྥᮥᰍ⩥⦹۵ߑᯩᨕ

⬉ᮉᖒᯕ׳ᮡSPI᮹᜽e⃺ࠥෝၽč⧁ᙹᯩᨩ݅. ᷪ, ᯝ݉᭥

SPIෝ⪽ᬊ⦹ᩍ6}ᬵၙอ᮹݉ʑaྥᮥ⠪a⧁Ğᬑ, ໑⋁࠺ᦩ᮹

٥ᱢvᙹపᮥ Łಅ⦹ᩍ ĥᔑࡹᨕḥ ᯝ ݉᭥ SPIෝ ᅖ⧊ᱢᮝಽ

⪽ᬊ⧕᧝ᝅᱽၽᔾ⦹۵aྥᮥḥ݉⧉ᨱᯩᨕᝁ഑ᖒᮥ⨆ᔢ᜽┍

ᙹ ᯩ۵aᨱ ݡ⦽ ❱݉ᯕ a܆⧕ ḥ݅.

✚⯩, ݅᧲⦽᜽e⃺ࠥᨱ঑ෙᯝ݉᭥SPI ᵲᨱᕽSPI_D30,

SPI_D60, SPI_D90 ॒᮹ᙽᕽಽɚ⦽aྥᨱݡ⦽ᰍ⩥܆ಆᯕ

ᬑᙹ⦽äᮝಽ⪶ᯙࡹᨩ݅. ᩍʑᕽɚ⦽aྥᮡ30ᯝ, 60ᯝ, 90ᯝŝ

zᯕݡᇡᇥ3}ᬵၙอ᮹᜽e࠺ᦩᨱ٥ᱢࡽvᙹపᯕ⠪֥᮹

ᙹᵡŝእƱ⦹ᩍⓍíᇡ᳒⧉ᮝಽᯙ⧕ၽᔾ⦽aྥᯕ௝Ł⧕ᕾࡹ

ᨕḥ݅. ݅᜽ั⧕, ʑ᳕᮹3}ᬵ, 6}ᬵ॒᮹٥ᱢvᙹపᮥʑၹᮝ ಽᔑᱶࡽᬵ݉᭥SPIෝᯕᬊ⦹ᩍᝅᱽ݉ʑᱢᯙvᙹᇡ᳒ᮝಽ

ᯙ⧕ၽᔾ⦽aྥᮥḥ݉⦹Ł⠪a⦹۵ߑᨱ۵ᝅḩᱢᯙ⦽ĥᱱᯕ

ᯩ۵ äᮝಽ ❱݉ࡽ݅.

đುᱢᮝಽᯝ݉᭥SPIෝʑၹᮝಽḡᗮᱢᯙaྥ༉ܩ░ยᮥ

☖⧕aྥ᮹ᱶࠥෝᅕ݅⬉ŝᱢᮝಽ⢽⩥⧕ᵥᙹᯩ۵٥ᱢvᙹᯝ ᙹෝᔍᱥᨱຝᱡđᱶ⦽⬥, ᯕෝʑၹᮝಽĥᔑ⦽aྥḡᙹෝ

ᖁᱶ⦹ᩍ ⪽ᬊ⧕᧝ ⦽݅. ੱ⦽ ၙ௹ aྥᰍ⧕ෝ ݡእ⦹ʑ ᭥⦽

ႊᦩᮝಽɚ⦽aྥၽᔾၰᰆʑᱢᮝಽḡᗮࡹ۵aྥᮥq᜽⦹Ł

šญ⦹ʑ᭥⧕ᕽ۵ᯝ݉᭥aྥḡᙹෝ ⪽ᬊ⦽aྥ༉ܩ░ยᮡ

⦥ᙹᱢᯝ äᯕ௝Ł ❱݉ࡽ݅.

qᔍ᮹ɡ

ᅙ ᩑǍ۵ ᗭႊႊᰍℎ᮹ ᯱᩑᰍ⧕ᱡqʑᚁ}ၽᔍᨦ[NEMA-

ᯱᩑ-2011-40] ḡᬱᮝಽ ᙹ⧪ࡹᨩ᜖ܩ݅. ᯕᨱ qᔍऽพܩ݅.

References

Ahn, J.-H., Hwang, P.-S. and Kang, S.-W. (2012). “The situation and corresponding in the spring drought, 2012.” Magazine of Korea Water Resources Association, Vol. 45, No. 9, pp. 84-91 (in Korean).

Byun, H. and Wilhite, D. A. (1999). “Objective quantification of drought severity and duration.” Journal of Climate, Vol. 12, pp.

2747-2756.

Choi, D. and Kim, S. (2010). “Revisiting Horton index using a conceptual soil water balance model.” Journal of Civil Engineering, Vol. 30, No. 5B, pp. 471-477 (in Korean).

Choi, G., Kwon, W.-T., Boo, K.-O. and Cha, Y.-M. (2008). “Recent spatial and temporal changes in means and extreme events of temperature and precipitation across the Republic of Korea.”

Journal of the Korean Geographical Society, Vol. 43, No. 5, pp.681-700 (in Korean).

Fawcett, T. (2006). “An introduction to ROC analysis.” Pattern Recognition Letters, No. 27, pp. 861-874.

Kim, G. and Lee, J. (2011). “Evaluation of drought indices using the drought records.” Journal of Korea Water Resources Association, Vol. 44, No. 8, pp. 639-652 (in Korean).

Kim, D.-H. and Yoo, C. (2006). “Analysis of spatial distribution of droughts in Korea through drought severity-duration-frequency analysis.” Journal of Civil Engineering, pp. 1597-1600 (in Korean).

Kim, D.-H., Yoo, C. and Kim, T.-W. (2011). “Application of spatial

(10)

EOF and multivariate time series model for evaluating agricultural drought vulnerability in Korea.” Advances in Water Resources, Vol. 34, No. 3, pp. 340-350.

Mason I. (1982). “A model for assessment of weather forecasts.”

Australian Meteorological Magazine, Vol. 30, pp. 291-303.

McKee T. B., Nolan J. D. and John K. (1993). “The relationship of drought frequency and duration to time scales.” Eighth Conference on Applied Climatology, American Meteorological Society, Anaheim, California, pp. 179-184.

Ministry of Land, Infrastructure and Transport (2001). Drought

records survey report in the 2001 (in Korean).

Ministry of Land, Infrastructure and Transport (2005). Drought management monitoring system reports (in Korean).

Tae Baek City (2009). The Drought Basic Books (in Korean).

Yoo, C. (2006). “Long term analysis of wet and dry years in Seoul, Korea.” Journal of Hydrology, Vol. 318, pp. 24-36.

Yoo, C., Kim, D., Kim, T.-W. and Hwang, K.-M. (2008).

“Quantification of drought using a rectangular pulses poisson

process model.” Journal of Hydrology, Vol. 355, pp. 34-48.

수치

Table 1. Definition of the SPI Classes and Corresponding Event  Probabilities
Fig. 3. Daily SPI and Monthly SPI for Seoul·Gyeonggi RegionĞᔢԉࠥḡᩎᨱᕽaᰆⓑ₉ᯕ(1.073, 1.110)ෝӹ┡ԕ۵äᮝಽ⪶ᯙࡹᨩᮝ໑, SPI_D90᪡SPI_M3, SPI_D120᪡SPI_M4ෝእƱ⧩ᮥĞᬑᨱ۵vᬱᩢ࠺ḡᩎᨱᕽaᰆⓑ₉ᯕ(0.704, 0.653), SPI_D150᪡SPI_M5ෝእƱ⧩ᮥĞᬑᨱ۵Ğᔢԉࠥḡᩎᨱᕽaᰆⓑ₉ᯕ(0.603), SPI_D180᪡SPI_M
Table 3. ROC Model
Table 4. The Collection of Previous Records for Extreme Drought in 2009
+2

참조

관련 문서

본 연구는 심층학습 기법인 Long Short Term Memory (LSTM)를 이용하여 지하수위를 예측 후 표준지하수위지수(Standardized Groundwater level Index, SGI)를 산정함으로써

‘Standard Precipitation Index’ (SPI) that is widely used to express the level of meteorological drought intensity has a limit of not being able to consider the hydrological changes

Key words : ensemble model, demand forecasting, short-term forecasting, urban water demand 주제어 : 앙상블 모형, 수요 예측, 단기 예측, 도시 물수요.. Evaluation of

Therefore, short-term application of 1mg·L −1 of ABA in the nutrient solution stimu- lated the root growth and drought tolerance of tomato seedlings by delaying the start time

The hot spot analysis was used to explore the extreme drought conditions and showed an increasing trend in the middle and northeastern parts of South Korea.. Overall, this

In this study, we investigated the applicability of drought assessment using the Evaporative Stress Index (ESI) based on Moderate Resolution Imaging

Projection of temporal trends on drought characteristics using the Standardized Precipitation Evapotranspiration Index (SPEI) in South Korea.. Drought hazard assessment in the

First, the changes of drought occurrence were analyzed by applying the Standardized Precipitation Evapotranspiration Index (SPEI) for meteorological data on 54 meteorological