SPATIOTEMPORAL CHARACTERISTICS
OF EXTREME RAINFALL EVENTS OVER JAVA ISLAND, INDONESIA
Case: East Java Province
COVER
Thesis
submitted to the Double Degree M.Sc. Program, Gadjah Mada University and Faculty of Geo-Information Science and Earth Observation, University of Twente
in partial fulfillment of the requirement for the degree of Master of Science in Geo-Information for Spatial Planning and Risk Management
By:
S U P A R I
GMU: 10/307098/PMU/06742 ITC: 27678
Supervisor:
1. Prof. Dr. HA. Sudibyakto M.S (GMU) 2. Dr. Ir. Janneke Ettema (ITC)
3. Dr. Edvin Aldrian M.Sc (BMKG)
THE GRADUATE SCHOOL GADJAH MADA UNIVERSITY
FACULTY OF GEO-INFORMATION AND EARTH OBSERVATION UNIVERSITY OF TWENTE
2012
APPROVAL SHEET
ABSTRACT
Extreme rainfall event is one of natural events frequently generating serious impact to many sectors. To date, its characteristic is expected being changing due global climate change. The study was aimed to identify the spatio- temporal characteristics of extreme rainfall events over Java Island, Indonesia by focusing analysis to East Java Province.
Some extreme indices calculated as annual series, were generated from rainfall record within period of 1981 – 2010. The maximum number of consecutive wet days, number of days where daily rainfall is more than or equal to 20 mm, 50 mm and 90th percentile were chosen to represent the frequency of extreme rainfall events. Meanwhile the highest 1-day rainfall amount, the highest 5-day rainfall amount, annual total and daily rainfall intensity were selected to represent the intensity of the events. A set of quality control procedures including duplicated data check, spatial outliers check, missing value check and homogeneity test was applied prior the analysis. The spatial characteristic of those events was identified by mapping climatological mean of indices while temporal characteristic was assessed using the non-parametric Mann-Kendal test.
The quality control procedures selected 84 stations as high quality data from total of 461 rainfall stations. The spatial pattern of extreme rainfall events over East Java Province is generally characterized by low frequency and intensity in the coastal area, and high frequency and intensity in the mountainous area. The dominant finding from trend assessment is not-significant trend. However, the consistently significant trend was observed in some districts. Rain stations in District of Ponorogo, Ngawi, Bojonegoro, Gresik and Sumenep showed significant negative trend for almost all indices whereas significant positive trend was found in District of Surabaya and Pasuruan.
Key words: spatio-temporal characteristics, extreme rainfall events, hydro- meteorological disaster, threshold, Java Island
INTISARI
Hujan ekstrim merupakan salah satu fenomena alam yang seringkali menyebabkan dampak negatif pada berbagai sektor. Saat ini frekuensi dan intensitas hujan ekstrim diduga telah mengalami perubahan sebagai akibat dari perubahan iklim. Penelitian ini bertujuan untuk mengetahui karakteristik dari hujan ekstrim di Pulau Jawa, khususnya di Propinsi Jawa Timur.
Analisis dilakukan dengan menghitung indeks-indeks hujan ekstrim, berdasarkan data hujan harian pada periode tahun 1981 – 2010. Indeks-indeks tersebut mengandung informasi baik tentang intensitas maupun frequency hujan ekstrim. Data hujan harian diuji kualitasnya sebelum digunakan untuk menghitung indeks ekstrim. Pengecekan kualitas data meliputi pengecekan kode standard, pengecekan data ganda, pengecekan data pencilan, pengecekan data kosong dan pengecekan terhadap homogenitas data. Karakteristik keruangan dari hujan ekstrim dinilai dengan cara memetakan rata-rata klimatologis dari masing-masing indeks, sedangkan karakterisktik terkait perubahan terhadap waktu diuji dengan metode Mann-Kendal test.
Pengujian kualitas data menyisakan 84 pos hujan sebagai data yang bagus dan layak untuk dianalisis. Karakteristik hujan ekstrim di Propinsi Jawa Timur umumnya bisa dikenali dengan ciri-ciri frequency dan intensitas yang rendah di wilayah pantai dan tinggi di wilayah pegunungan. Secara umum hujan ekstrim tidak mengalami perubahan yang sangat nyata, namun pada beberapa kabupaten, teramati adanya pos-pos hujan yang menunjukkan perubahan yang nyata dan terjadi tidak hanya pada satu indeks tapi konsisten pada beberapa indeks.
Penelitian ini menemukan bahwa pos hujan di Kabupaten Ponorogo, Ngawi, Bojonegoro, Gresik dan Sumenep menunjukkan penurunan frequensi dan intesitas hujan ekstrim. Sebaliknya, pos hujan di Kabupaten dan Kota Surabaya dan Pasuruan menunjukkan peningkatan frequensi dan intensitas hujan ekstrim.
ACKNOWLEDGMENTS
The Master project detailed in this thesis was supervised by Prof. Dr. HA.
Sudibyakto, M.S (Gadjah Mada University), Dr. Ir. Janneke Ettema (Univ. of Twente) and Dr. Edvin Aldrian M.Sc (BMKG). Their advices and comments were very much appreciated. Special thanks are extended to Division of Climate Data Analysis - Head office of BMKG, BMKG’s regional office of Malang and Semarang from which the rainfall data were collected.
I would also like to thank to Indonesian People for providing the scholarship of this master study through national budget given for Pusbindiklatren Bappenas and Pusdiklat BMKG, and to my institution for permitting me to pursue this master degree.
For my parents as well as my parents in law, my deepest regards to you for giving valuable supports. Your blessing enabled me to finish this hard work. For Rey, my wife and Humam-Hanif, my sons this is dedicated to you.
i
Table of Contents
COVER __________________________________________________________ 0 APPROVAL LETTER ______________________________________________ 0 ABSTRACT ______________________________________________________ 2 ACKNOWLEDGMENTS ___________________________________________ 3 Table of Contents __________________________________________________ i List of Figures ____________________________________________________ iv List of Tables_____________________________________________________ vi Abbreviations ___________________________________________________ vii I. INTRODUCTION ____________________________________________ 1 1.1. Background _______________________________________________ 1 1.2. Problem Statement __________________________________________ 2 1.3. Research Objective __________________________________________ 5 1.4. Research Questions _________________________________________ 5 1.5. Research Hypothesis ________________________________________ 6 1.6. Research Benefits ___________________________________________ 7 1.7. Research Limitation _________________________________________ 7 II. LITERATURE REVIEW_______________________________________ 8 2.1. Climate in Indonesia _________________________________________ 8
2.1.1. Rainfall Process and Cloud Formation __________________________________ 9 2.1.2. Rainfall Classification in Indonesia ___________________________________ 10
2.2. Extreme Value Analysis _____________________________________ 12
2.2.1. Extreme Rainfall Event _____________________________________________ 13 2.2.2. Indices of Extreme Rainfall Events ___________________________________ 14 2.2.3. Spatial Analysis for Rainfall Data ____________________________________ 14 2.2.4. Temporal Trend Analysis ___________________________________________ 15
2.3. Research on Extreme Rainfall Events over the World ______________ 16
2.3.1. America_________________________________________________________ 16 2.3.2. Africa __________________________________________________________ 16 2.3.3. Asia ____________________________________________________________ 17 2.3.4. Europe __________________________________________________________ 18 2.3.5. Australia ________________________________________________________ 18
ii
III. STUDY AREA AND DATA ___________________________________ 20 3.1. Study Area _______________________________________________ 20
3.1.1. Java Island ______________________________________________________ 20 3.1.2. East Java Province ________________________________________________ 21
3.2. Data ____________________________________________________ 22 IV. RESEARCH METHODS _____________________________________ 26 4.1. Research Framework _______________________________________ 26 4.2. Method for Quality Control __________________________________ 27
4.2.1. Checking for duplicated data ________________________________________ 27 4.2.2. Checking for outliers and missing values _______________________________ 28 4.2.3. Homogeneity test _________________________________________________ 29
4.3. Method for Identification of Extreme Indices ____________________ 33
4.3.1. Fix Threshold ____________________________________________________ 33 4.3.2. Site Specific Threshold _____________________________________________ 34
4.4. Method for Spatial Analysis __________________________________ 37 4.5. Method for Temporal Trend Analysis __________________________ 38 4.6. Method for Severity Analysis _________________________________ 40 V. SCREENING DAILY RAINFALL DATA ________________________ 43 5.1. Converting into Standard Format ______________________________ 43 5.2. Duplicate Data Check _______________________________________ 45
5.2.1. Procedures on Checking Duplicated Data_______________________________ 45 5.2.2. Result of Duplicated Data Check _____________________________________ 49
5.3. Outlier Check _____________________________________________ 53 5.4. Missing Value Check _______________________________________ 58 5.5. Homogeneity Test _________________________________________ 62 VI. RESULT ___________________________________________________ 67 6.1. Spatial Characteristic of Extreme Rainfall Events _________________ 67
6.1.1. Fix Threshold ____________________________________________________ 67 6.1.2. Site Specific Threshold _____________________________________________ 70 6.1.3. Climatological Mean of Annual Indices ________________________________ 77 6.1.4. Topography Effect ________________________________________________ 88
6.2. Temporal Trend of Extreme Rainfall Events _____________________ 90
6.2.1. Result of the Assessment ___________________________________________ 90 6.2.2. Spatial Pattern of Detected Trend _____________________________________ 97
iii
6.3. Severity Analysis _________________________________________ 109 VII. DISCUSSION _____________________________________________ 113 VIII. CONCLUSION AND RECOMMENDATION ____________________ 119 8.1. Conclusion ______________________________________________ 119 8.2. Recommendation _________________________________________ 120 IX. REFERENCES _____________________________________________ 122 APPENDICES __________________________________________________ 127
iv
List of Figures
Figure 1-1. Frequency of floods and landslides over Java Island _____________ 4 Figure 2-1. Geographic location of Indonesia ____________________________ 8 Figure 2-2. The illustration of three types of rainfall ______________________ 11 Figure 2-3. Illustration of two ways on analyzing extreme values ___________ 13 Figure 3-1. Java Island (red box) among the other islands in Indonesian archipelagos______________________________________________________ 20 Figure 3-2. The three dominant rainfall regions in Indonesia _______________ 21 Figure 3-3. Topographic feature of East Java Province ____________________ 24 Figure 3-4. Distribution of 2580 rain gauges over Java Island ______________ 25 Figure 4.1. Research Framework _____________________________________ 26 Figure 4-2. Unadjusted and adjusted time series _________________________ 30 Figure 4-3. Probability distribution function of Gumbel, Frechet and Weibull __ 35 Figure 4-4. Trend of rainfall being larger than 95th percentile over Europe ____ 38 Figure 4-5. Thiessen polygons of Australia _____________________________ 41 Figure 4-6. The procedure to generate severity index _____________________ 42 Figure 5-1. The example of digital data stored by Malang Climatological Station __________________________________________________________ 44 Figure 5-2. The example of digital data stored by Semarang Climatological Station __________________________________________________________ 44 Figure 5-3. The example of standard format of rainfall series _______________ 45 Figure 5-4. The example of detection process for duplicated data ___________ 46 Figure 5-5. The example of coincident similar total value _________________ 46 Figure 5-6. The example of duplication where not all daily data are copied ____ 47 Figure 5-7. The duplication case at Station Maelang, District of Banyuwangi __ 47 Figure 5-8. A technique to identify original-copied data ___________________ 48 Figure 5-9. Number of months containing duplicated data _________________ 49 Figure 5-10. The process of rainfall data collecting in Indonesia ____________ 51 Figure 5-11. Spatial distribution of registered gauges _____________________ 52 Figure 5-12. Spatial correlation function of daily rainfall over study area _____ 53 Figure 5-13. An example of replacing outliers by statistical threshold ________ 57 Figure 5-14. Scatter plot of original series of Station Gombal ______________ 57 Figure 5-15. Scatter plot of corrected series of Station Gombal _____________ 58 Figure 5-16. Percentage of missing values average from 10 districts _________ 60 Figure 5-17. Spatial distribution of selected gauges based on missing value criteria __________________________________________________________ 61 Figure 5-18. An interface of tool of homogeneity test on xlstat Add-Ins package _________________________________________________________ 62 Figure 5-19. A shift in series of Wagir rain station, District of Malang _______ 63 Figure 5-20. A break in series of Dam Sembah rain station, District of Jember _ 64 Figure 5-21. Spatial distribution of useful series _________________________ 65
v
Figure 5-22. Spatial distribution of high quality rainfall data passing all quality control procedures _________________________________________________ 66 Figure 6-1. Histogram and box plot of rainfall correspond to disaster ________ 67 Figure 6-2. The affected area due to hydro-meteorological disaster __________ 69 Figure 6-3. Box plot of threshold based on 90th percentile, 1-yr, 5-yr and 25-yr return period _____________________________________________________ 71 Figure 6-4. Spatial distribution of threshold based on 90th percentile _________ 73 Figure 6-5. As Figure 6-4 but for rainfall with 1-year return period, R1yr _____ 74 Figure 6-6. As Figure 6-4 but for rainfall with 5-year return period, R5yr _____ 75 Figure 6-7. As Figure 6-4 but for rainfall with 25-year return period, 25yr ____ 76 Figure 6-8. The climatological mean of R20mm _________________________ 80 Figure 6-9. As Figure 6-8 but for R50mm ______________________________ 81 Figure 6-10. As Figure 6-8 but for R90p _______________________________ 82 Figure 6-11. As Figure 6-8 but for CWD _______________________________ 83 Figure 6-12. As Figure 6-8 but for RX1d _______________________________ 84 Figure 6-13. As Figure 6-8 but for RX5d _______________________________ 85 Figure 6-14. As Figure 6-8 but for RTOT ______________________________ 86 Figure 6-15. As Figure 6-8 but for SDII _______________________________ 87 Figure 6-16. The scatter plot of index CWD versus log of elevation __________ 89 Figure 6-17. The example of significant temporal change of R20mm _________ 91 Figure 6-18. The example of significant temporal change of R50mm _________ 92 Figure 6-19. The example of significant temporal change of R90p ___________ 93 Figure 6-20. The example of significant temporal change of CWD __________ 93 Figure 6-21. The example of significant temporal change of RX1d __________ 95 Figure 6-22. The example of significant temporal change of RX5d __________ 95 Figure 6-23. The example of significant temporal change of RTOT __________ 96 Figure 6-24. The example of significant temporal change of SDII ___________ 96 Figure 6-25. Spatial pattern of detected trend for R20mm __________________ 99 Figure 6-26. As Figure 6-25 but for R50mm ___________________________ 100 Figure 6-27. As Figure 6-25 but for R90p _____________________________ 101 Figure 6-28. As Figure 6-25 but for CWD _____________________________ 102 Figure 6-29. As Figure 6-25 but for RX1d _____________________________ 105 Figure 6-30. As Figure 6-25 but for RX5d _____________________________ 106 Figure 6-31. As Figure 6-25 but for RTOT ____________________________ 107 Figure 6-32. As Figure 6-25 but for SDII _____________________________ 108 Figure 6-33. Thiessen polygons created for regional analysis ______________ 111 Figure 6-34. The severity map for extreme rainfall events ________________ 112 Figure 7-1. Spatial distribution of stations showing consistently significant trend___________________________________________________________ 118
vi
List of Tables
Table 1-1. Areal daily rainfall observed in the Bengawan Solo Watershed during 2007 Java flood ___________________________________________________ 4 Table 1-2. List of research questions __________________________________ 6 Table 2-1. European trends per decade _______________________________ 18 Table 2-2. Numbers of stations in Australia with positive and negative trend _ 19 Table 3-1. The detail required data___________________________________ 23 Table 4-1. Correspondence between GEV and three basic extreme value distribution _____________________________________________________ 36 Table 4-2. Detail extreme rainfall indices _____________________________ 37 Table 5-1. The detail result of cases of duplicated data check for East Java Province________________________________________________________ 50 Table 5-2. Summary of spatial outliers check in respect of district __________ 56 Table 5-3. Summary of MVs check with regard to district ________________ 59 Table 6-1. List of disasters occurred in the last ten year __________________ 68 Table 6-2. Summary of trend assessment for all frequency indicators _______ 94 Table 6-3. Summary of trend assessment for all intensity indicators_________ 97 Table 7-1. Contingency table showing inter-index relation _______________ 115 Table 7-2. List of stations which are consistently increasing ______________ 116 Table 7-3. List of stations which are consistently decreasing _____________ 116
vii
Abbreviations
APN Asia Pacific Network
BBWS Balai Besar Wilayah Sungai, Regional Office for Watershed Management, Ministry of Public Work
BMKG Badan Meteorologi Klimatologi dan Geofisika, National Agency for Meteorology, Climatology and Geophysics BNPB Badan Nasional Penanggulangan Bencana, National
Agency for Disaster Management
BP DAS Balai Pengelola Daerah Aliran Sungai, Watershed Management Authority, Ministry of Forestry
BPS Badan Pusat Statistik, Statistics Indonesia
GEV Generalized Extreme Value
GHCN Global Historical Climatology Network IPCC Intergovernmental Panel on Climate Change
IQR Inter-Quartile Range
ITCZ Inter Tropical Convergence Zone
MV Missing Value
NCDC National Climatic Data Centre - USA
SNHT Standard Normal Homogeneity Test
SRTM Shuttle Radar Topography Mission
TSE The Theil-Sen Estimator
WMO World Meteorological Organization
I. INTRODUCTION
1.1. Background
Extreme rainfall events are among the most devastating weather phenomena since they are frequently followed by flash floods and sometimes accompanied by severe weather such as lightning, hail, strong surface winds, and intense vertical wind shear (Jones et al. 2004). Consequently, they generate large economic, social and environmental impact (Manton et al. 2001; Carvalho et al.
2002; Jones et al. 2004). In rural area, the extreme rainfall events can damage crops and livestock. While in urban area, these events often cause flood problem due to inadequate drainage system to accommodate a sudden large amount of rainfall (Carvalho et al. 2002). In global perspective, these events are also supposed being responsible for rapidly rising costs of losses since the 1970s (Rosenzweig et al. 2007).
Extremes of climate are an expression of the natural variability, actually (Trenberth et al. 2007). However, these events become serious issue to date because their frequency is expected being change. It is confirmed by the Intergovernmental Panel on Climate Change (IPCC) that human influences on climate lead to change in frequency and intensity of extreme weather events (Trenberth et al. 2007). Some extremes are expected to become more frequent, more widespread and/or more intense. It is logic then, when the demand for information of extreme weather is growing (WMO, 2009).
Observational studies over some regions suggested evidence of change in climate extremes. Using daily rainfall data from 1931 – 1996, Kunkel et al.
(1999), examined the trend of extreme rainfall events over the Conterminous United States and Canada and found an indication of increasing trend in the number of 7-day, 1-yr events. Even, some climate divisions have experienced increases of 50% – 100%. Zang et al. (2001), found an upward trends in the number of extreme rainfall events for the spring over eastern Canada when they examined the characteristic of extreme rainfall events using site specific threshold
over whole country. Study on this event in the Alpine Region by Frei and Schar (2001), also confirmed an increasing trend for autumn and winter season.
However, some other studies in tropical region reveal inconsistent result.
For example in Peninsular Malaysia, Suhaila et al. (2010) found that almost all stations in the eastern region show a decreasing trend of frequency of extreme rainfall events during southwest monsoon period. Nevertheless, the western region even shows the contrast result, an increasing trend. Atsamon et al. (2009) also found different trend between two regions in Thailand. On the Andaman Sea, they characterized an overall decrease while on the Gulf of Thailand they detected an increasing trend in magnitude and frequency of more intense rainfall events.
Regarding the statement at WMO guidelines on analysis of extremes (WMO, 2009) that the sustainability of economic development and living conditions depends on our ability to manage the risks associated with extreme events, the study of extreme rainfall events in Indonesia is urgent. The present study focus on analyzing the spatial and temporal characteristics of extreme rainfall events using GIS tools. The result of this research is expected being able to provide crucial inputs to manage the risk as mentioned before.
1.2. Problem Statement
Changing of probability of extreme rainfall events implies seriously to many sectors such as engineering, regional planning and other activities which traditionally assumed that climate is stationary (Suppiah and Hennessy, 1998).
This assumption states that climate is variable, but the variation tends to be constant meaning it occurs around unchanging mean state (WMO, 2009).
As described in previous sub chapter, the different changing trend was found in distinct region. This confirms us that regional scale perhaps has different response to the global climate which is identified changing. So far to the author’s knowledge, there is no regional scale study on extreme rainfall events in Indonesia. Manton et al. (2000) have studied extreme rainfall events over Indonesia but for the large region i.e. Asia Pacific Region. They found different trend. There is an increasing trend for extreme rainfall in Fiji and French
Polynesia. However, Solomon Island, Philippines, New Zealand, Malaysia and Japan show the decrease trend. The others country even show no significant trend.
Over Indonesia, they concluded that the trend of extreme rainfall events is not significant. Unfortunately, they only used six rainfall stations which are Pangkalpinang, Jakarta, Balikpapan, Manado, Ambon and Palu. Those stations represent western, central and eastern region of Indonesia. Those six stations are inadequate certainly to display climatic condition of the whole country. The regional scale study using more rainfall stations is therefore needed to figure out the actual trend.
Relating to disasters triggered by extreme rainfall event (maximum daily), Java Island is chosen as an ideal area for this study. There were high frequency of severe disaster events here related to extreme weather event such as flood and landslide event. National Agency for Disaster Management, BNPB recorded that more than 1,000 occurrences of floods and landslides strike the Island with various intensity within 2002 - 2008. The frequency of those events for each province is shown in the figure 1-1 where West and Central Java take place as first and second province with frequency of those disasters being more than 300 events.
The recent example of those disasters is flood which occurred in the end of 2007. Expanding from Central to East Java, it caused hundreds of casualties and damaged thousands houses. The flood was triggered by heavy rainfall event with intensity more than 100 mm/day taking place simultaneously and intensively in December 25, 2007 (Hidayat et al. 2008). The record from three regions showed that rainfall which occurred during this disaster has return period ranging from 40 up to more than 100 year (Table 1-1). The maximum areal daily rainfall was more than 100 mm. The similar disaster reoccurred at the beginning of 2009 (BNPB, 2011).
Figure 1-1. Frequency of floods and landslides over Java Island: 2002 – 2008 (Source: BNPB, 2011)
Table 1-1. Areal daily rainfall observed in the Bengawan Solo Watershed during 2007 Java flood (Source: Nippon Koeico. Ltd cited in Hidayat, 2008)
PARAMETER UPPER SOLO BASIN MADIUN BASIN WONOGIRI DAM
Intensity 134 mm/day 141 mm/day 128 mm/day
Return Period 55 year 500 year 40 year
Broadening our understanding of extreme rainfall events is especially relevant for Java Island because it is the most populous Island in Indonesia, with more than 120 million people living there. For national perspective, Java is the centre of economic activity, national government system and agricultural product.
We need to examine then whether the extreme rainfall events in java regional scale is changing or not. The changing of these events either in frequency or intensity will affect on formulating the policy in many fields such as agricultural (agricultural product management) and infrastructural sectors (construction management). Without studying this subject, we will never realize whether our strategies at those sectors are still supported by recent extreme climate condition or not.
1.3. Research Objective
The major objective of this study is to analyze pattern of extreme rainfall events within last three decades (1981 - 2010) both spatially and temporally in the context of hazard study. By documenting its spatio-temporal characteristic, we can recognize well the potential places where and when the extreme rainfall events occur. The output will be displayed on the maps and curves, for example map of annual probability of rainfall with intensity more than or equal to 50 mm and map of distribution of daily rainfall intensity with 25 year return period. The minor objectives are as follows:
1. To define threshold value of extreme rainfall events for general application.
Fix threshold defined by BMKG (National Agency for Meteorology, Climatology and Geophysics) was evaluated by correlating it to historical data of disaster. Site specific threshold was calculated for each station based on statistical parameter of rainfall data.
2. To characterize spatial characteristics of extreme rainfall events over study area by mapping its indices for various intensity and recurrence interval.
3. To detect possible temporal trend of extreme rainfall events.
4. To identify districts recording high frequency of extreme rainfall events.
Based on its frequency, the extreme rainfall events were classified in to the category of less severe, moderate and more severe.
1.4. Research Questions
To address those all objectives, here some questions are formulated as shown in the table below:
Table 1-2. List of research questions
Research objectives Research questions To define threshold for
extreme rainfall events
How do we evaluate extreme threshold defined by BMKG?
How do we define the threshold of Extreme Rainfall Events using site specific threshold?
What is the amount of rainfall at different return period?
To document spatial characteristics of extreme rainfall events
What is the spatial characteristic of Extreme Rainfall Events over Java Island?
What is the effect of topography to extreme rainfall events?
To characterize temporal trend
What is the temporal trend of Extreme Rainfall Events within last three decades?
How significant is the trend?
To identify the district with severe extreme rainfall events
How to map the severity of extreme rainfall events?
Where is the most severe extreme rainfall events present?
1.5. Research Hypothesis
1. The site specific threshold will be more appropriate to express extreme rainfall in certain area. The amount of rainfall varies spatially in respect of return period.
2. The spatial pattern of extreme rainfall events is affected by topographical feature.
3. There is significant trend of extreme rainfall events.
1.6. Research Benefits
By producing some maps which display indices of rainfall extreme events, this research will be beneficial for some stakeholders for example:
1. Disaster management authorities. They can utilize the research output to identify which area should be prioritized more related to risk of extreme rainfall events.
2. Public work authorities. They can use the output to evaluate whether the existing infrastructure still appropriate considering recent extreme climate condition or not.
3. Agricultural management authorities. They can use the output to define the best agricultural product regarding extreme rainfall characteristic in certain area.
4. For general society. The output will give better understanding about various return period and intensity of extreme rainfall events so that they are able to cope with the possible effects.
1.7. Research Limitation
The research will be limited to analyze the characteristics of extreme rainfall events without studying more to the related disaster. It means that the individual extreme rainfall event will not be analyzed. Considering the available digital data of daily rainfall, the record used for the study is within 1981 – 2010.
The trend identified from the study is expected being able to figure out the change of extreme rainfall in the last three decades.
The extreme events analyzed in the study refer to maximum extremes, not minimum extremes since the minimum extreme in daily resolution give no impact to the human life. The severity of extreme rainfall events was identified based on its frequency. A recommendation to the districts was designed with regard to severity level of extreme rainfall events only, without looking at the environmental condition of those districts.
II. LITERATURE REVIEW
2.1. Climate in Indonesia
Climate is, in general, an expression of weather average (Petterssen, 1958).
The main energy source for climate dynamic is solar energy. Climate in the world is variable both spatial and temporally as a result of difference respond of earth surface on receiving solar energy.
The archipelagos of Indonesia are located between Asian Continent and Australian Continent stretching out along the equator line, in east-west direction (Figure 2-1). Every place in Indonesia receives solar energy in similar amount approximately. It is logic then if the spatial variation of temperature and pressure is quite limited. The variation of those two variables is only on vertical manner due to altitude variation.
Figure 2-1. Geographic location of Indonesia - yellow box (a) and general atmospheric circulation over Indonesia (b). The two continents and two oceans control air mass movement there causing variability of climate (Source: google earth and www.climate4you.com).
As a tropical country, Indonesia receives abundant incoming solar radiation along the year so the solar energy is surplus here. Indonesia has also plentiful source of water vapor because it is located between two oceans which are
Indian Ocean and Pacific Ocean. It is not surprising that the climate here is very both dynamics and humid. Two big continents (Asia and Australia) flanking it also persuades the climate pattern in Indonesia by intervening general atmospheric circulation over Indonesia (Figure 2-1).
A clear illustration of climate in Indonesia is the different weather condition between dry and wet season. In the dry season, the weather will be sunny, humid and less rainfall whilst in the wet season it will be cloudy, humid and much rainfall. Rainfall varies notably with respect to its frequency, duration, intensity and spatial pattern, a common characteristic of rainfall (Barrett and Martin, 1981) but in Indonesia its variation is more complex. Therefore climate regime classification in Indonesia is primarily based on rainfall variation.
2.1.1. Rainfall Process and Cloud Formation
Rainfall process refers to a cycle where the parcel or sample of air undergoes a process to be moist, grow to be a cloud cell and produce rainfall finally. The factors governing the occurrence of rainfall are the motion of cloud air and its aerosol properties, which determine the concentration, initial size distribution and nature of cloud properties (Mason, 1971 cited in Barrett and Martin, 1981).
Rainfall process determines the characteristic of rainfall that it produces. It describes the mechanism which stimulates cooling and condensation process by which the moist air starts to become cloud droplets. Principally, cloud will form if there is a parcel of moist air lifted. Petterssen (1958) described in his book that the rainfall process is started when the moist air ascends and cools by expansion. As it cools, the relative humidity will increase. When the process continues, the air will be saturated and cloud droplets form.
These droplets do not freeze until the temperature is far below freezing point (less than -28 C). In this step, the cloud has already formed. As soon as some of cloud elements have outgrown the others, the larger ones will fall through the cloud and further growth will result from collisions among them. When the
cloud droplets are large enough they will fall as rain droplets since the lifting force is less than gravity.
Rainfall probability is a function of cloud thickness, base and top temperature (Barrett and Martin, 1981). In tropical region, the clouds will not produce rain until their thickness reaches more than 6000 feet. They are almost produce rain when their thickness achieves more than 12000 feet (Petterssen 1958).
2.1.2. Rainfall Classification in Indonesia
The most common approach to classify rainfall is based on its forming mechanism or its process. Based on the mechanism how the moist air is lifted, there are three rainfall types found in Indonesia i.e. convectional rainfall, orographic rainfall and convergence rainfall. The illustration is presented in the Figure 2-2 whereas the description is summarized in the following paragraph.
1. Convectional Rainfall.
Convectional rainfall most commonly results from the air which having been warmed by conduction from a heated land surface, expands and rises in a cold and dense surrounding air (Monkhouse, 1959; Linsley Jr et al. 1982). The local heating starts the whole process and is therefore known as a trigger effect (Figure 2-2). However, when the moist air ascends, it can carry on even when the heating process ends.
In the equator region, Indonesia for instance, convectional rainfall occurs throughout the year because of constant high temperature and humidity (Monkhouse, 1959). It is characterized by spotty rainfall with intensity is ranging from light shower up to rainstorm (Linsley Jr et al. 1982).
2. Orographic Rainfall.
The orographic rainfall occurs when moist air is forced to climb the side of mountain range. It is commonly found in the area where hills lie parallel to the coast over which moist air are blown by wind from the sea (Monkhouse, 1959).
In Indonesia, we can find the suitable area for producing orographic rainfall, like the mountainous area stretching out along Sumatra Island and Java Island. When the warm moist air from sea climbs the mountain range, it is cooled and condensed. The cloud forms then and develops intensively as long as the supply of warm moist air is still continuous (Figure 2-2).
If the mountainous area is high enough, the cloud producing rainfall often forms in front of the summit only (windward sides) and remains the relative dry area in the leeward sides known as “rain shadow area”.
Figure 2-2. The illustration of three types of rainfall i.e. convectional rainfall (left top), orographic rainfall (right top) and convergence rainfall (bottom). In nature, the mechanism process is interrelated and the produced rainfall cannot be identified as being of one type exactly (Source:
http://splashman.phoenix.wikispaces.net).
3. Convergence Rainfall.
Convergence rainfall is produced from the cloud resulting from the convergence of two moist air masses. When those two air masses converge, as a fluid, they will ascend to find less dense area. The ascending causes the air masses being cooled and condensed and the clouds form finally.
The inter-tropical convergence zone (ITCZ) is a perfect example for the area where convergence rainfall occurs. Over Indonesia, the ITCZ often forms as a convergence of tropical maritime air masses from Asia and Australia, lying some hundreds of miles, ranging from Indian Ocean until Pacific Ocean (Monkhouse, 1959). A long line of massive cumulonimbus clouds with torrential rain and thunderstorm are common appearance of the ITCZ.
2.2. Extreme Value Analysis
Extreme value analysis is a statistical analysis based on what we know as Extreme Value Theory. It is the branch of statistics which describes the behavior of the extreme data observations (Gilli, 2003; Naveau et al. 2005). The purpose of extreme value analysis is to estimate what future extreme levels of a process might be expected (Coles SG and RSJ Sparks, 2001) and what the likely recurrence of these events is, based on a historical series of observations (Murphy, 1997).
Generally there are two ways of identifying extremes in real data. The first approach considers the maximum (or minimum) of the variable taking in successive periods, for example months or years. These selected observations constitute the extreme events, also called block maxima (or per-period maxima).
The second approach focuses on the appearance of values exceeding a given threshold. Figure 2-3 displays the difference of those two approaches. The block maxima method is the traditional method used to analyze data with seasonality as for instance hydrological data. However, threshold methods use data more efficiently and, for that reason, seem to become the choice method in recent applications (Gilli, 2003).
Figure 2-3. Illustration of two ways on analyzing extreme values. Left panel, block maxima method, displays the observations X2, X5, X7 and X11 representing these block maxima for four periods with three observations per period . Right panel shows threshold method, by which X1, X2, X7, X8, X9 and X11 are categorized as extremes since they exceed a threshold µ (Source: Gilli, 2003).
2.2.1. Extreme Rainfall Event
Extreme value analysis is applied by scientists to study climate extremes as for instance extreme of temperature and rainfall. In line with increasing attention to climate change issue, study of extreme climate events has also been more interesting because their characteristics can be used to indicate the change in climate.
The two approaches of analyzing extremes, as mentioned in previous sub chapter, are operated. In the context of extreme rainfall events researches, identifying annual series of maximum daily rainfall is example of block maxima method while calculating frequency of rain day with rainfall more than 20 mm is case of threshold method.
Extreme rainfall events are defined as 24-hour accumulative rainfall exceeding a certain threshold. There are some different ways by which meteorologists determine this threshold. Goswami and Ramesh (2007) used a daily rainfall depth of 250 mm as a threshold on analyzing vulnerability of Indian Region due to extreme rainfall events. Bodini and Cossu (2010) used a daily rainfall depth above 95th percentile over certain period on assessing vulnerability of Central-East Sardinia to extreme rainfall events. It sounds as a site dependent threshold.
Fu et al. (2010), which adopted the methodology used by Kunkel et al.
(1999, 2003), used also a site dependent threshold which is defined by recurrence interval when they analyzed long-term temporal variation of extreme rainfall events in Australia. Hernandez et al. (2009) registered that there are at least 23 different ways used in professional literatures to define threshold or indices explaining extreme rainfall events. World meteorological organization has published a set of standard indices expressing extreme rainfall events (WMO, 2009) and standard procedure how to investigate that.
2.2.2. Indices of Extreme Rainfall Events
There are many different ways by which scientists define the indices of extreme rainfall. Hernandez et al. (2009) registered that there are at least 23 different ways used in professional literatures throughout the world to define threshold or indices explaining extreme rainfall events. WMO, (2009) through their guidelines on analysis of extremes in a changing climate define 11 extreme rainfall indices.
The readers interested to the detail indices are recommended to refer the guidelines directly. The other ways to define indices can be found such as in the study of Kunkel et al. (1999), Hernandez et al. (2009) and Bodini and Cossu (2010). All indices are calculated for annual number or annual value. The rainfall value of 1 mm is commonly applied to define rain event or rain day.
2.2.3. Spatial Analysis for Rainfall Data
Rainfall, as the other natural phenomena, is a kind of regionalized variable.
It varies in space and time. Local atmospheric condition and topographical factors affect the spatial distribution of rainfall (Subyani, 2004). However, the rainfall is measured in point base using rain gauges network. The density of rain gauges often depends on the accessibility to location. On the flat area we can find the dense rain gauges but in the complex terrain, a sparse rain gauge network is common situation.
On the other hand, areal rainfall value is more essential for any applications such as hydrological model and weather prediction rather than point value. Interpolation technique is hence applied to get areal value. The basic principle of interpolation is the assumption that at short distance the values are more similar than at further distance (Meijerink et al. 1994).
There are many choices of interpolation techniques for rainfall data ranging from simple approaches such as thiessen polygons and inverse distance weighted to more complex approaches such as krigging and genetic algorithm (Subyani, 2004; Haberlandt, 2007).
2.2.4. Temporal Trend Analysis
The method to identify temporal trend of time series data is a lot but the most frequently used by meteorologist is Man-Kendall test. The use of this test on detecting trend is shown for instance in Zang et al. (2001), Fu et al. (2010) and Penalba and Robleda (2010). Principally, the Man-Kendall test examines the observation by calculating a gap between one observation data with earlier one.
The data surely should be arranged in time order. The next data are calculated respectively. The null hypothesis is that the total of those gaps will be 0 (zero) meaning that there is no change in the series.
Mason et al. (1999) used an alternative method to calculate the significance of change so called a re-sampling method. It is a method being free from distributional assumptions. A series of data is divided into two successive periods.
Those two successive periods should be balance in term of long of data period.
The beta- and beta –P distribution are fitted then for each of those two periods.
The change of those periods can be assessed then by comparing those beta- and beta –P parameters. The limitation of this method is that the results can be sensitive to the a priori definition of the sub-period and to the value of n (number of selected extremes).
2.3. Research on Extreme Rainfall Events over the World
Observational studies of climate extremes are focused on characterizing the possible change of the extreme events as it is shown by some models. Over the world, there are some studies to examine the change of extreme rainfall events.
The summary of those study based on its region is given in the next paragraphs.
2.3.1. America
In America continent, Kunkel et al. (1999) have investigated trend of extreme rainfall events particularly over United States and Canada. The data involved are 1295 rainfall stations with 66 year observation data (1931 – 1996) for United Sates and 63 rainfall stations with 43 year observation data (1951 – 1993) for Canada.
They designed a procedure to define extreme rainfall events. Event durations of 1, 3, and 7 days were examined. Two precipitation total thresholds were used to screen events for use in the analysis, defined by recurrence intervals of 1 and 5 year. For each station, the annual number of events for each duration and recurrence interval was identified.
The linear trend analysis indicates that there has been a significant increase in the number of 7-day, 1-yr events over the period of 1931 – 1996 over United States. Some climate divisions have even experienced increases of 50% – 100%.
While over Canada, an upward trend is not significant.
2.3.2. Africa
South Africa is the region which was selected by Mason et al. (1999) on detecting the change of extreme rainfall events. 60 year observational data (1931 – 1990) from 314 rainfall station is involved. They did not test the homogeneity data due to lack of metadata, but the rainfall stations selected are only un- relocated rain gauges. For each station, they calculated the intensity of 5, 10, 20, 30 and 50 year recurrence interval of extreme rainfall events.
The method they applied was different. They divided those 60 year data in to two time windows e.g. 1931 – 1960 and 1961 – 1990. The null hypothesis for those time windows is that there is no difference in the intensity of extreme rainfall events between first and second window. The significance of changes was then assessed by comparing the difference in the intensity of those events with the difference expected under the null hypothesis.
Over much of the country, they found a significant evidence of increases in the intensity of high rainfall events between 1931 – 1960 and 1961 – 1990.
Percentage increases in intensities are largest for the most extreme rainfall events.
Similar patterns of change are evident for the different return periods analyzed, but the percentage changes are even larger for the more extreme rainfall events.
2.3.3. Asia
Rainfall extreme events over Asia particularly Southeast Asia have been studied by Manton et al. (2001) under the Asia Pacific Network (APN) for Global Change Research. Using relatively short data period (1961 – 1998) they examine 91 rainfall stations over Southeast Asian and South Pacific. The criteria they used to calculate the extreme indices are:
1. Frequency of daily rainfall exceeding the 1961 – 1990 mean 99th percentile (extreme frequency).
2. Average intensity of events greater than or equal to the 99th percentile each year, i.e. in the four wettest events (extreme intensity).
3. Percentage of annual total rainfall from events greater than or equal to the 99th percentile, i.e. received in the four wettest events (extreme proportion).
4. Frequency of days with at least 2 mm of rain (rain days).
They found that the number of rain days, annual total rainfall and frequency of extreme rainfall events have decreased within the study period at the majority of stations. The decreasing trend of the number of rain days is significant whereas that of frequency and intensity is not significant.
2.3.4. Europe
Study of Klein Tank and Konnen (2003) is outlined here to figure the trend of daily rainfall extremes in Europe. Using seven indices of climate extremes for precipitation which are agreed internationally i.e. highest 1-day (RX1d), highest 5- day (consecutive, RX5d), heavy rainfall day (R10mm), very heavy rainfall day (R20mm), moderate wet days (R75%), very wet days (R95%) and rainfall fraction due to very wet days (R95%tot, see www.knmi.nl/samenw/eca), they tested 151 rainfall stations over whole of Europe.
The result revealed that, averaged over Europe, six out of those seven indices significantly increase between 1946 – 1999 but the spatial pattern is not really coherence (Table 2-1).
Table 2-1. European trends per decade (with 95% confidence intervals in brackets) in the indices of extreme precipitation for the periods 1946–1999. Values significant at the 5% level (t test) are set bold face (Source: Klein Tank and Konnen, 2003).
2.3.5. Australia
Study on extreme rainfall events in Australia presented here is work of Fu et al. (2010). They used 97 years observation record from 191 rainfall stations to investigate temporal changes in the number of extreme rainfall events by closely following the method of Kunkel et al. (1999).
From the data series, event duration of 1, 5 10 and 30 days were examined.
The new series which are identified from first step was screened then using recurrence interval of 1, 5 and 20 year. They conclude that more than half of stations show negative trend but they are mostly not significant. Their result is
shown in the table below. The figure of global trend of extreme rainfall events over other regions including Japan, Russia and Brazil can be found in Easterling et al. (2000).
Table 2-2. Numbers of stations with positive and negative trend in the numbers of extreme events during the period 1910–2006, and numbers of stations for which these trend are statistically significant at one-sided = 0.05 (Source: Fu et al. 2010).
III. STUDY AREA AND DATA
3.1. Study Area
3.1.1. Java Island
Situated between 105 2’ – 114 6’ E and 5 8’ – 8 8’ S (Figure 3-1), Java is one of the big islands in Indonesia whose area is 126,700 km². Its topography is characterized by low land whose elevation is less than 30 meter in the coastal area and mountainous area whose elevation could reach up to more than 3,500 meter.
Java is the densely populated island in Indonesia and is the centre of national economic activity. Sixty percent out of total populations in Indonesia inhabit this island. The Statistics Indonesia (BPS) reported that the population of Java Island in 2005 was 128.5 million inhabitants, distributed in six provinces i.e. Banten, Jakarta, West Java, Central Java, Yogyakarta and East Java Province.
Figure 3-1. Java Island (red box) among the other islands in Indonesian archipelagos (Source:
data processing).
The climate of Java is mainly controlled by monsoon system. There are two monsoon systems influencing this area. The northwest (NW) monsoon is active from November to March (NDJFM) and the southeast (SE) monsoon is working from May to September (MJJAS) (Aldrian and Susanto, 2003). The characteristic
of those two monsoons is significantly different. The northwest monsoon is wet and implies much rainfall while the southeast monsoon is dry and is responsible for less rainfall period over Java. Consequently, there is significant difference of rainfall amount between dry and wet season. Figure 3-2 present general climate of Indonesia which is divided in to three regions namely region A, B and C. Region A is mainly controlled by monsoon system with one rainy season and one dry season, including Java Island. Region B is characterized by double rainy seasons and Region C is classified as local climate.
Figure 3-2. Left panel shows the three dominant rainfall regions in. Right panel shows the annual cycles of rainfall in the region of climate “A”. Solid line indicates rainfall average, dashed lines indicates one standard deviation (σ) above and below of it (Source: Aldrian and Susanto, 2003).
3.1.2. East Java Province
East Java Province was selected for the case of analysis of extreme rainfall analysis even though all data have been prepared for Java Island entirely. East Java Province comprises of two main islands i.e. the eastern part of Java Island and Madura Island. Administratively, the province is divided in to 29 districts and 9 municipalities.
For the current study, 29 districts and 1 municipality only were considered since the others municipalities are too small and located in the centre part of districts for instance Municipality of Madiun is located in the centre part of Madiun District . For effective analysis, those small municipalities therefore were supposed as single region with related districts (see Figure 3-3).
