Title ASSESSMENT of HYDRO-METEOROLOGICAL

ASSESSMENT OF HYDRO-METEOROLOGICAL Title DROUGHTS RELATED TO ENSO IN LOMBOK AND SUMATRA ISLANDS, INDONESIA( Dissertation_全文 )

Author(s) Karlina

Citation 京都大学

Issue Date 2018-03-26

URL https://doi.org/10.14989/doctor.k21058

Right 許諾条件により本文は2019-03-26に公開

Type Thesis or Dissertation

Textversion ETD

Kyoto University ASSESSMENT OF HYDRO-METEOROLOGICAL DROUGHTS RELATED TO ENSO IN LOMBOK AND SUMATRA ISLANDS, INDONESIA

KARLINA

2018

ASSESSMENT OF HYDRO-METEOROLOGICAL DROUGHTS RELATED TO ENSO IN LOMBOK AND SUMATRA ISLANDS, INDONESIA

KARLINA

A dissertation Submitted in partial fulfillment of the requirements for the Degree of Doctor of Engineering

Department of Civil and Earth Resources Engineering Kyoto University, Japan

2018

Acknowledgements All praise is due to Allah S.W.T. the Almighty, for giving me the blessing, the strength, the chance, the patience, and endurance to complete this study. Without His blessed, this study which is entitled “Assessment of Hydro-Meteorological Droughts Related to ENSO in Lombok and Sumatra Islands, Indonesia” would never be finished. The research was done in the Laboratory of Innovative Disaster Prevention Technology and Policy Research, Disaster Prevention Research Institute, Kyoto University under the supervision of Professor Kaoru TAKARA. I would like to thank everyone for support me until I could finish this dissertation.

First, I would like to express my sincere gratitude to Professor Kaoru TAKARA who gave me a valuable chance to stay in his laboratory for three years. His continuous support and encouragement have been a great inspiration for me to finish this study. He, who always give challenge during my Ph.D. life, has always been the greatest source of strength and courage for me to try some new experiences to broaden my research and also develop my self-value and capability.

Secondly, I wish to express my most profound appreciation to Assoc. Prof. Takahiro SAYAMA, who has been continuously giving me guidance and valuable input during my Ph.D. His continuous support, encouragement, advice, and patients have been a high motivation for me to finish this research. Through so many discussions with him, I learned a lot how to build up logic for my study and finish it.

I would like to thank Ms. Sono INOUE and Ms. Kaori SAIDERA for their supports of administrative tasks and kindness before and during my stay in Japan. Through their assistance, I could finish my study activities without any meaningful problem.

I would also like to acknowledge the GSS Program Faculty Members, especially Assoc. Prof. Shinpei Kudo as my mentor for always gives me support, advice, guidance, and valuable insight for finishing my GSS Activity. I thank Assoc. Prof. Minako Jen Yoshikawa, Assoc. Prof. Kumiko Kondo, Assoc. Prof. Mika Shimizu, Assoc. Prof. Masanori Katsuyama, Assoc. Prof. Makoto Nishi, Assoc. Prof. Florance Lahournat, and Assoc. Prof. Nobuyuki Ito for their kindness during GSS activities. I would also like to thank Ms. Maki Katsuyama and all GSS staff members for their kind assistance in GSS administrative papers.

I sincerely express my gratitude to Mr. Anang Fariansyah from River Basin Organization of Nusa Tenggara 1 for assisting in the collection of data used in this study. Without his help, this study would not have been possible.

I express my acknowledgment to former lab members: Dr. Apip (who always gives support during fieldworks in Indonesia), Dr. Sahu (for a valuable time in discussing my research work), Dr. Pedro Chaffe, Dr. Bounhieng Vilaysane, Dr. Maochuan Hu, Dr. Josko Troveli, Dr. Hendy Setiawan (for always gives advice, courage, support, and insight in our discussions), Dr. Han Xue (for always gives positive insight), Dr. Chong Khai Lin (for lovely spending time together in and outside of laboratory life), Mr. Shusuke Takahashi (for being my tutor during my first 6 months stay in Japan), Mr. Takuma Ushiro and Mr. Tsukasa Goto. I would also like to thank all lab members: Mr. Pham Van Tien, Mr. Adnan Arutyuniv, Mr. Toma Stoyanov, Mr. Nguyen Duc Ha, Mr. Kodai Yamamoto, Mr. Ryosuke Kobayashi, Mr. Shintaro Miyake, Mr. Yoshito Sugawara, Mr. Koji Matsumoto, Ms. Saeka Togashi, Mr. Try Sophal and Mr. Steven Ly. Lovely thanks to Ms. Shi Yongxue who always be a great friend during my time. Also, thanks to Ms. Eva Mia Siska for jointly strengthening each other in completing this dissertation. Lastly, I would send my most profound thank to my dearest friends: Ms. Widha Kusumaningdyah and Ms. Salma Intifadha for spending wonderful time together in Japan and supporting each other in completing the study.

My most profound gratitude belongs to my family: my mother, my father, my mother-in-law, my father-in-law, my brother, my sister-in-law, my brother-in-law and my nephews for always sending me love and encouraging me to finish my study. My utmost thanks and sincere gratitude go to my beloved husband who always is patients and strong for me. His continuous support has been the greatest energy for me to accomplish this achievement.

Finally, I am grateful to acknowledge the financial support received from ASEAN University Network / Southeast Asia Engineering Education Development Network (AUN/SEED-Net) and Japan International Cooperation Agency (JICA) through AUN/SEED-Net scholarship to perform my Ph.D. study during three years (2015-2018).

Kyoto, November 2017

Abstract

Indonesia is prone to drought disaster. However, despite the long historical record of drought in Indonesia, it is still challenging to understand the disaster completely. It is difficult to detect and monitor the drought event. Lack of physical evidence of the disaster is one of the reasons. The impact of a drought is less visible and spread over a large geographical extent, different from other natural hazards. Drought also a creeping phenomenon which makes the onset and termination difficult to determine. The effect of drought accumulate slowly and may continue after sometimes, could be months, seasons even years. Also, the existed of the various definition of drought adds confusion about drought, whether it exists and if it is severe or not. For these reasons, it is still difficult task to quantify the drought impacts and to define the disaster relief strategy.

Unlike meteorological drought which is routinely monitored, there is still lack of hydrological drought monitoring system. The study about the relation of drought and ENSO also focus only on the meteorological drought. There is the necessity of improving the understanding of the relationships between climatological and hydrological parameters to develop measures to reduce the impacts of droughts. The understanding at the local scale is critical due to the heterogeneity in spatiotemporal hydro-meteorological variability.

This dissertation attempted to assess the integration between meteorological and hydrological drought with regards to hydro-meteorological droughts forecasting. The two kinds of drought, hydrological and meteorological drought analysis were performed in this study. This study focuses on the development of drought forecasting method. This study also aims for better understandings of historical hydro-meteorological drought in relation to the climate phenomenon such as ENSO. This study focused on the seasonal streamflow correlation with ENSO and its relation with drought forecasting system. The specific objectives of the study are to understand the effect of the climate phenomenon, ENSO specifically, to the hydro- meteorological drought condition in Lombok Island and Sumatra Island, to develop the method of drought forecasting system based on ENSO indices, and to build hydrological drought forecasting system based on the low flow characteristics.

The hydrological drought mainly focuses on the analysis of streamflow drought in Lombok Island. The discharge index was introduced for two catchment analysis; the Jangkok Catchment and the Babak Catchment. The streamflow used for the Jangkok Catchment is the discharge

iii

data in the Bug Bug Station, while Lantandaya Station represented the Babak Catchment. The calculation of the discharge index mainly use the monthly discharge data where the difference of monthly average with monthly data was compared with the standard deviation.

The impact of ENSO to the hydrological drought was examined by using Pearson Correlation. The hydrological drought analysis in Lombok Island shows the ENSO impact on discharge in Bug Bug station is more significant than in Lantandaya station. The strongest ENSO-discharge correlation in Bug Bug station happens during Sep-Oct-Nov (SON) season, and the weakest one happens during Jun-Jul-Aug (JJA) season. Furthermore, the lag correlation between JJA ENSO and SON discharge in Bug Bug Station shows the possibility to use ENSO for discharge drought prediction.

The investigation of the impacts of ENSO and local SST to rainfall at the Batanghari River Basin in Sumatra, Indonesia was presented to discuss general characteristics of rainfall patterns in the region as well as the possibility of using those indicators for seasonal rainfall predictions. The monthly, seasonal, and annual rainfall shows no significant trend, except for the decreasing trend in MAM season. The Pearson’s Correlation between monthly rainfall and ENSO shows a significant influence of ENSO on the rainfall from July to October. Moreover, local SST shows significant correlation with rainfall in July to September. In addition, the three-month lag correlation between ENSO to rainfall was found to be statistically significant in SON. Also, local SST in Mar-Apr-May (MAM) show a strong correlation with rainfall in JJA. These results indicate that there is a possibility of using ENSO and local SST indicators for rainfall predictions in the dry seasons in JJA and SON in this region.

Finally, the baseflow forecasting was performed on Lombok Island. The study proposes hydrological drought forecasting methods based on two streamflow recession analyses. The first one is based on a recursive digital filter for baseflow separation and recession characterization for the baseflow forecasting. The second one is based on the theory of “simple dynamical systems of catchments.” The applications of the two methods were demonstrated in Lombok Island in Indonesia and showed that the latter method, which reflects more flexible recession characteristics showed better accuracy in the estimations of the low flows. Nevertheless, both of the presented applications showed underestimations in low flow predictions compared to the observed ones. The underestimations were mainly associated with the ignorance of the rainfall, especially for long lead time cases.

Overall this study revealed that Lombok Island is prone to drought disaster especially during El Nino. The meteorological characteristic of the island is drier in the eastern part and wetter in the northwest part. The Batanghari River Basin shows declining trend of rainfall based on the data from 1985-2012. There is seasonal correlation between seasonal rainfall in the Batanghari River Basin with ENSO and Local SST. Based on the result of the lag correlation between hydrological drought and meteorological drought analysis in Lombok Island, there is a possibility to use the ENSO index for drought prediction in the future, especially for SON season.

v vi

Content List

Acknowledgements ...... i

Abstract ...... iii

Content List ...... vii

List of Figures ...... xi

List of Tables ...... xiii

1 Introduction ...... 1

1.1 Droughts in Indonesia ...... 3

1.2 Indonesian Climate Sensitivity to ENSO ...... 7

1.3 Research Problem and Justification ...... 8

1.4 Objectives of the Study ...... 9

1.5 References ...... 9

2 Study Area and Data ...... 13

2.1 Introduction ...... 13

2.2 The Lombok River Basin ...... 13

2.2.1 Social Economic Conditions ...... 15

2.2.2 Climate ...... 15

2.2.3 Water Resources ...... 17

2.3 The Batanghari River Basin ...... 18

2.3.1 Social Economic Conditions ...... 20

2.3.2 Climate ...... 21

2.3.3 Water Resources ...... 23

2.4 Data ...... 23

2.4.1 The Lombok River Basin Hydro-meteorological Data ...... 23

2.4.2 The Batanghari River Basin Hydro-meteorological Data ...... 27

vii

2.4.3 Sea Surface Temperature Data...... 27

2.5 References ...... 28

3 Hydrological Drought Correlation with El Nino in Lombok River Basin...... 31

3.1 Introduction ...... 31

3.2 Study Area and Data ...... 32

3.3 Methods ...... 38

3.3.1 Drought index ...... 38

3.3.2 Calculation of Regional Drought Characteristics ...... 38

3.3.3 The Pearson’s Correlation Method ...... 39

3.4 Result ...... 40

3.4.1 Meteorological and Hydrological Characteristics ...... 40

3.4.2 Drought Identification ...... 47

3.4.3 Drought Characteristics ...... 52

3.4.4 Intermediate Correlation ...... 55

3.4.5 Lag Correlation ...... 57

3.5 Summary ...... 58

3.6 References ...... 58

4 Impacts of ENSO and Local Sea Surface Temperature on Rainfall Patterns in the Batanghari River Basin, Sumatra, Indonesia ...... 61

4.1 Introduction ...... 61

4.2 Study Site and Data ...... 61

4.3 Method ...... 64

4.3.1 Trend Analysis ...... 64

4.3.2 The Pearson’s Correlation Method ...... 65

4.4 Result and Discussion ...... 65

4.4.1 Rainfall Trends...... 65

4.4.2 Intra-seasonal Correlation of Rainfall with ENSO and Local SST ...... 67

viii

4.4.3 Lag Correlation ...... 72

4.5 Summary ...... 73

4.6 References ...... 74

5 Low Flow Forecasting with Recession Analysis Approaches ...... 75

5.1 Introduction ...... 75

5.2 Study Site and Data ...... 76

5.2.1 Study site ...... 76

5.2.2 Data ...... 77

5.3 Recursive Digital Filters...... 77

5.4 Catchment as Simple Dynamical Systems ...... 79

5.5 Flow Characteristic ...... 80

5.6 Parameter Calculation ...... 81

5.6.1 Recession constant ...... 81

5.6.2 Model parameter of simple dynamic systems ...... 82

5.7 Forecasting Result ...... 83

5.7.1 Recursive digital filters model forecasting ...... 83

5.7.2 Simple dynamical systems forecasting ...... 84

5.7.3 Forecasting performance ...... 86

5.8 Summary ...... 86

5.9 References ...... 87

6 Concluding Remarks ...... 89

ix x

List of Figures

Figure 1-1 Drought transfer processes and interactions (Liu et al., 2016)...... 2 Figure 1-2 Type of natural hazard by number of occurrence in Indonesia (source: www.bnpb.go.id)...... 4 Figure 1-3 Number of regencies/municipalities with high risk of hazards (Bappenas & BNPB, 2010)...... 4 Figure 1-4 Three climatic regions in Indonesia (Aldrian & Dwi Susanto, 2003)...... 7 Figure 2-1 Administrative map of the Lombok River Basin...... 14 Figure 2-2 GRDP per sector (BPS, 2014)...... 15 Figure 2-3 Spatial distribution of annual rainfall in the Lombok River Basin...... 16 Figure 2-4 Water usage distribution (BWS-NT1, 2010)...... 17 Figure 2-5 The vulnerability map of water sector of Lombok Island (green color represent not vulnerable area, and red color represent extremely vulnerable area)...... 18 Figure 2-6 Administrative area of the Batanghari River Basin, in Sumatra Island (Source: BMKG)...... 19 Figure 2-7 Livelihood sectors in Jambi Province...... 21 Figure 2-8 Distribution of hydrological station in the Batanghari River Basin (Source: River Basin Organization of Sumatra VI)...... 22 Figure 2-9 Annual average rainfall in the Jambi Province (Source: BMKG)...... 23 Figure 2-10 Rain gauge station distribution in the Lombok River Basin...... 24 Figure 2-11 Discharge station distribution in the Lombok River Basin...... 26 Figure 3-1 Number of villages affected by drought in Lombok Island (Source: BPBD of NTB)...... 31 Figure 3-2 Areas as biggest rice producers in Indonesia (Kompas, 2013) ...... 32 Figure 3-3 Study area: the Jangkok Catchment and the Babak Catchment in Lombok Island...... 33 Figure 3-4 Spatial distribution of rainfall stations used in the meteorological drought analysis...... 34 Figure 3-5 Spatial distribution of annual rainfall characteristics...... 34 Figure 3-6 Average monthly rainfall of Lombok Island...... 35 Figure 3-7 Location of Nino3.4 region. (Source: https://www.ncdc.noaa.gov/teleconnections/enso/indicators/sst.php) ...... 36 Figure 3-8 Illustration of El Nino and La Nina event based on ONI...... 36 Figure 3-9 Average monthly rainfall in Lombok Island (red strip line indicate (12-months running mean of rainfall)...... 40 Figure 3-10. Average monthly rainfall of Lombok Island...... 41 Figure 3-11 Average monthly rainfall during a normal condition and ENSO event...... 41 Figure 3-12 Spatial distribution of monthly rainfall in Lombok Island (normal years)...... 42 Figure 3-13 Spatial distribution of monthly rainfall in Lombok Island (El Nino years)...... 43 Figure 3-14 Spatial distribution of monthly rainfall in Lombok Island (La Nina years)...... 43 Figure 3-15. Monthly dry coverage area during normal and ENSO years...... 44 Figure 3-16. Average of the dry coverage area...... 44 Figure 3-17 Spatial distribution of annual rainfall in Lombok Island...... 45 Figure 3-18 Time series of monthly discharge in Bug Bug Station and rainfall in Sesaot Station...... 46 Figure 3-19 Time series of monthly discharge in Lantandaya Station and rainfall in Lingkok Lime Station...... 46

Figure 3-20 Monthly average discharge in both stations...... 47 Figure 3-21 Seasonal average discharge in both seasons...... 47 Figure 3-22 Dry conditions based on Schmidt (1957)...... 48 Figure 3-23 Drought event based on drought index...... 48 Figure 3-24 Dry periods 1997-1998 based on Schmidt (1957)...... 49 Figure 3-25 Drought periods 1997-1998 based on drought index...... 49 Figure 3-26 Spatial distribution of drought propagation in 1997-1998 drought event...... 50 Figure 3-27 Drought area percentage...... 50 Figure 3-28 Hydrological drought index in Lantandaya Station...... 51 Figure 3-29 Hydrological drought index in the Bug Bug Station...... 52 Figure 3-30 Meteorological drought duration-frequency...... 53 Figure 3-31 Highlight of 7 months drought duration (upper: 1993-1994 drought, lower part: 1997-1998 drought)...... 53 Figure 3-32 Meteorological drought severity-frequency analysis...... 54 Figure 3-33 Catchment drought duration-frequency in the Babak Catchment...... 55 Figure 3-34 Catchment drought duration-frequency in the Jangkok Catchment...... 55 Figure 3-35 Correlation of droughts and ENSO (upper part in the Babak Catchment, lower part in the Jangkok Catchment)...... 56 Figure 3-36 Regional meteorological drought correlation with ENSO (R is performed for 3- month Pi and ONI) (red color shows significant value at P = 95% for a two-tail test)...... 57 Figure 4-1 Study area of the Batanghari River Basin in Jambi Province, Sumatra Island, Indonesia...... 62 Figure 4-2 Comparison of the Batanghari River Basin rainfall from Sultan Thaha Station and GPCC data: a) time series and b) correlation...... 64 Figure 4-3 The Batanghari River Basin monthly and seasonal rainfall characteristic...... 66 Figure 4-4 Inter-annual rainfall and SST correlation pattern...... 69 Figure 4-5 Coefficient correlation (r) between rainfall with Nino3.4 and local SST...... 70 Figure 4-6 Correlation between rainfall and sea surface temperature (DJF, MAM, JJA, and SON seasons respectively from above to below). A correlation of 0.374 is significantly different from zero at the 95% confidence level, assuming the 26 degrees of freedom...... 72 Figure 4-7 Lag correlation of rainfall with local SST & Nino3.4...... 73 Figure 5-1 The Babak River Basin...... 76 Figure 5-2 River flow in Babak River (upper: Gebong Station, below: Lantandaya Station). 81 Figure 5-3 Recession constant...... 82 Figure 5-4 Recession plot of Lantandaya and Gebong Station (log-log scale)...... 83 Figure 5-5 Recession plot of Lantandaya Station and Gebong Station (ln-ln scale). The yellow dots show the plot of recession flow (dQ/dt) with the flow (Q), while the red dots represent the bin average of the –dQ/dt. The red line shows the quadratic regression of the bin average of the –dQ/dt...... 83 Figure 5-6 Forecasted flow based on two methods in Lantandaya Station and Gebong Station...... 85 Figure 5-7 Error value of the low flow forecasting compared with observed baseflow...... 86

xii

List of Tables

Table 1-1 Top 10 Natural Disasters in Indonesia for the period 1900 to 2014 sorted by numbers of total affected people ...... 4 Table 1-2 Area of paddy cultivation affected by flood and drought from 1988-97 (hectare) ... 5 Table 2-1 Administrative area of the Lombok Island ...... 13 Table 2-2 Total population of Lombok Island per Regency ...... 14 Table 2-3 Hydro-meteorological characteristics in Lombok Island ...... 16 Table 2-4 Administrative Area within the Batanghari River Basin boundary ...... 19 Table 2-5 Population density in the Batanghari River Basin ...... 20 Table 2-6 Climate condition in the Batanghari River Basin (upstream, central and downstream) area ...... 21 Table 2-7 Rainfall data availability in the Lombok River Basin ...... 25 Table 2-8 Discharge data availability in the Lombok River Basin ...... 26 Table 3-1 The Oscillation Nino Index (ONI) used to define the El Nino and La Nina events (Source: NOAA) ...... 37 Table 3-2 Drought classification based on the drought index...... 38 Table 3-3 Lag Correlation between ENSO and discharge in Bug Bug Station ...... 57 Table 4-1 Mann Kendall Trend Analysis...... 67

xiii

1 Introduction

Drought is a creeping disaster that is slowly developed in a given area after such period without rainfall. The physical evidence of the drought propagation is somewhat unclear. The dry condition may appear because of no rainfall for few days, but the state could not yet be identified as a drought. After a month without rain, the water scarcity may then appear and give a signal of drought development. However, once the condition has already been felt, the drought event has already developed.

Scientists have introduced several drought definitions. Wilhite and Glantz (1985) in early drought study presented the various ways for defining drought based on two categories; conceptual and operational definitions. Conceptual definitions are formulated in general terms to identify the boundaries of the concept of drought. Operational definitions refer to the identification of the onset, severity, and termination of drought episodes. The operational definitions can also be used to assess the drought frequency, severity, and duration of historical drought events. This kind of definitions requires data on hourly, daily, monthly or seasonal to identify when the drought occurred. Such definitions may also be used to calculate the probabilities of droughts of varying intensity, duration, and spatial characteristics.

Furthermore, Wilhite and Glantz (1985) classified drought into four types; meteorological, agricultural, hydrologic, and socioeconomic drought. Meteorological drought is defined based on the degree of dryness and the duration of the dry period. The most common definition of meteorological drought is a lack of precipitation over a region for a period of time (Mishra & Singh, 2010). The assessment of meteorological drought is commonly done by using precipitation data, e.g., McKee, Doesken, and Kleist (1993), Patel, Chopra, and Dadhwal (2007) and Smakhtin and Hughes (2007). Olukayode Oladipo (1985) suggested that precipitation is the most important climatic element as an input into meteorological drought analysis. While meteorological drought focuses on the precipitation amount, the hydrological drought concerns surface or subsurface water resources for established water use of a given water resources management system (Mishra & Singh, 2010; Wilhite & Glantz, 1985). Agricultural drought definitions refer to a period with declining soil moisture and consequent crop failure without any reference to surface water resources (Mishra & Singh, 2010). Socio- economic drought is related to the failure of water resources systems to meet water demands. It occurs when the demand for water for an economic good exceeds supply as a result of a weather-related shortfall in water supply.

Liu et al. (2016) illustrate the correlation between four types of drought in Figure 1-1. When the amount of precipitation is decreasing, meteorological drought occurs first as a direct consequence. The agricultural and hydrological drought will gradually follow the situation because of continuous water evaporation. The continuous evapotranspiration and decreasing rainfall affect the soil water condition. As agricultural drought concerns the water deficit in crops, loss soil moisture caused by the decreasing precipitation is the earliest step of the drought propagation. If the condition continues for months or seasons, it will cause crop failure. As the precipitation is decreasing continuously and causes the reduction of water resources in the rivers and reservoirs, the hydrological drought is developing. A severe hydrological drought for an extended period may also impact the deficit of groundwater level. As the crop failure happens and water demand increases, the fluctuation in the market is occurring; the socioeconomic drought is developing.

Figure 1-1 Drought transfer processes and interactions (Liu et al., 2016).

Based on the drought transfer processes and interactions, the meteorological drought could be one first step to identify the drought propagation. A better understanding of hydro- meteorological relationships and the development of enhanced forecasts have the potential for forecasting more proactive planning and hazard response activities that will help reduce the loss due to the disaster (Ryu, Svoboda, Lenters, Tadesse, & Knutson, 2010).

1.1 Droughts in Indonesia

In Indonesian Dictionary, drought refers to a condition without water or a natural phenomenon of climate anomaly that may happen anytime when rainfall is below average (The Language Center of National Education Center, 2008). Several drought terms have been used for drought management activities in Indonesia. Indonesian Agency for Meteorology, Climatology, and Geophysics (BMKG) uses the term of meteorological drought which has been routinely monitored and predicted by using the Standardized Precipitation Index (SPI) (Hatmoko, Raharja, Tollenaar, & Vernimmen, 2015). Hydrological drought is used by the Ministry of Public Works to evaluate the water allocation for irrigation, public water supply, and other water usages. Meanwhile, for practical work, the National Disaster Management Authority (BNPB) uses the socio-economic drought term, the condition when the water availability tends to be less than the water demand. The BNPB uses this term as the basis for water distribution to local people during the drought period. The assessment of agricultural drought has been done by assessing the total area affected by drought, which has been done by the Directorate of Plant Protection (Alimoeso et al., 2002; Alimoeso et al., 2000).

Indonesia is a country that is prone to disaster including drought. Indonesia ranks 12th in the world for high mortality risks from multiple hazards. The most frequent natural hazards in Indonesia until 2015 floods events, followed by fire, strong wind, landslides and drought (Figure 1-2). Regarding a number of regencies/municipalities with a high risk of hazards, drought places the third place after earthquake and soil movement (Figure 1-3). However, concerning the number of total affected people, drought is a disaster that has an impact on the most people in Indonesia (Table 1-1).

4,500 3,875 4,000 3,500 3,000 2,500 1,971 2,000 1,639 1,562 1,413 1,500 Number Number of occurence 1,000 850 500 308 273 0 Floods Fire Strong Wind Landslides Drought Floods And Earthquake Others Landslides

Figure 1-2 Type of natural hazard by number of occurrence in Indonesia (source: www.bnpb.go.id).

160 151 140

120 104 100 91 91 90 92 80 70 60 42 45 41 40

Number Number of regions 40 20 13 0 drought earthquake flood soil movement tsunami volcanic eruption

Extremely high risk High risk

Figure 1-3 Number of regencies/municipalities with high risk of hazards (Bappenas & BNPB, 2010).

Table 1-1 Top 10 Natural Disasters in Indonesia for the period 1900 to 2014 sorted by numbers of total affected people No Disaster Date No Total Affected 1 Drought 1972 3,500,000 2 Earthquake (seismic activity) 27-May-06 3,117,923 3 Wildfire Oct-94 3,000,000 4 Earthquake (seismic activity) 30-Sep-09 2,501,798 5 Drought Sep-97 1,065,000 6 Flood 23-Dec-06 618,486 7 Flood 9-Feb-96 556,000 8 Earthquake (seismic activity) 26-Dec-04 532,898 9 Flood 14-Mar-66 524,100 10 Flood 27-Jan-02 500,750

Indonesia drought is often become worst because of El Nino. The data of drought from 1877 to 1997 shows the 93 percent of the disaster in Indonesia has been linked to El Nino events (Kishore, 2000). The strong El Nino in 1997-98 had significant social and economic implications for Indonesia. Most places in Indonesia suffered from drought because of the El Nino. The drought condition besides threated food security due to crop failure will also impoverish Indonesian people who have livelihoods in the field of agriculture. The data compare the area of paddy cultivation area affected by floods and droughts from 1988 to 1997 shows the higher impact of drought (Table 1-2). The severe drought resulted in a massive shortfall in rice production that forced Indonesia to import of over five million tons of rice to ensure food availability to the economically weaker sections of the society (Kishore, 2000).

Table 1-2 Area of paddy cultivation affected by flood and drought from 1988-97 (hectare) Flood Drought Year Remarks Partial damage Total damage Partial damage Total damage 1988 130,375 28,934 87,373 15,115 La Nina 1989 96,540 13,174 36,143 2,116 1990 66,901 9,642 54,125 9,521 1991 38,006 5,707 867,997 192,347 El Nino 1992 59,360 9,615 42,409 7,267 1993 78,480 26,844 66,992 20,415 1994 132,973 32,881 544,422 161,144 El Nino 1995 218,144 46,957 28,580 4,614 La Nina 1996 107,385 38,167 59,560 12,482 1997 58,974 13,787 504,021 88,467 El Nino Average 98,714 22,571 229,162 51,349 Source: Directorate of Crop Protection, Ministry of Agriculture, Government of Indonesia

Droughts also one of the main drivers of forest fires in Indonesia. The drought-driven fires occur typically during the El Nino events (Fernandes et al., 2017). In 1997-98, the fire-induced by drought-related El Nino devastated large areas of tropical rainforest worldwide (Siegert, Ruecker, Hinrichs, & Hoffmann, 2001). Fanin and van der Werf (2017) mentioned in their study, the most robust El Nino years that involved high incidences of fire in Indonesia were 2015 and 1997. The fires typically start during the July to October (JASO) dry seasons for land clearing agriculture (Murdiyarso & Adiningsih, 2007). The 1997 fires have resulted in months- long dangerous atmospheric pollution levels (Marlier et al., 2013) and economic losses estimated at over 4.5 billion US$ (Schweithelm & Glover, 1999).

Several methods for drought monitoring system has been developed in Indonesia. Boer and Subbiah (2005) reviewed the agricultural drought in Indonesia. In their study, they stated that

the cause of the severe crop damage in Indonesia is because of the significant decrease in rainfall. The BMKG developed an operational seasonal climate prediction scheme for Indonesia in 1993. The rainfall forecast is issued every six months; early March for dry season (Apr-Sep) and in early September for the wet season (Oct-Mar). Despite the availability of the forecasting system, the decision makers are still unable to utilize the forecast because of the coarse resolution.

The other approach to estimate the future crop productivity is by using ENSO indices (D'Arrigo & Wilson, 2008). D'Arrigo and Wilson (2008) built a more accurate model of Sep-Dec Palmer Drought Severity Index (PDSI) over Java, Indonesia. The model is a combination of Indian Ocean SST and Nino3.4 SST. The results demonstrate that a simple empirical forecast model of drought for Java using the DMI in combination with Nino3.4 presents a significant improvement over one based only on the recent index, the variable previously utilized in such model.

Another study by Hatmoko et al. (2015) built a hydrological drought monitoring and prediction system in Pemali-Comal River Basin by utilizing a Drought Early Warning System (DEWS) based on Delft-FEWS software. The system is designed to access hydrological drought indicators such as river discharges and reservoir water level by using file transfer facility, offline manual input; telemetering system; and satellite access of TRMM. The online mode of Delf-FEWS can monitor the current river discharge in the irrigation weirs. It can also monitor several hydrological drought indicators which can give information on the updated information of the ongoing drought events and the predicted future drought.

In the future, drought in Indonesia is likely to get worse (BPBD, 2015). The study of King, Karoly, and van Oldenborgh (2016) shows Indonesia endured severe heat and drought during the dry season of 2015. They investigated the attribution of the extreme conditions to human- induced climate change and the concurrent El Nino. They found out that El Nino conditions strongly increased the probability of a drier-than-normal dry season. Besides, the warming trend due to anthropogenic climate change has also increased the likelihoods of high temperatures. The study also shows a trend toward less rain and more extreme dry events, which is smaller than can be significantly detected in the current observations.

1.2 Indonesian Climate Sensitivity to ENSO

Several studies have been done to see the correlation between Indonesian-rainfall with El Nino- Southern Oscillation (ENSO) indices. Kirono (2000) did a preliminary study on the spatial variation of relationships between Indonesia rainfall and the ENSO phenomenon. They analyzed monthly rainfall dataset covering the period of 1950-1997 for 33 synoptic across the region. The data was then compared with the Troup Southern Oscillation Index (SOI) for the same period. The results confirm there is a close relationship between ENSO and Indonesian rainfall. Later on, Aldrian and Dwi Susanto (2003) investigated the correlation between Global Historical Climatology Network (GHCN) rainfall data from 884 stations in Indonesia from 1961 to 1993 with Nino3. While they classified Indonesia into three climatic regions (Figure 1-4), they found out there is a high correlation of the rainfall in the two regions; A and C. The significant responses of the rainfall pattern to ENSO in Region A and C is seen during Jun- Nov (JJA & SON) dry season. Meanwhile, in the region B, the ENSO-related signal is suppressed.

Figure 1-4 Three climatic regions in Indonesia (Aldrian & Dwi Susanto, 2003).

Another study by Haylock and McBride (2001), which determine the spatial coherence of wet season anomalies in Indonesia by using 63 rainfall stations across Indonesia during 1950-1998, they compare Indonesian rainfall with the Southern Oscillation Index (SOI). In their study, the all-Indonesian rainfall indices in the Jun-Aug (JJA) and Sep-Nov (SON) season is the highest while they are the lowest in Dec-Feb (DJF) and Mar-May (MAM) is the lowest one. Hamada et al. (2002) focused on the spatial and temporal of the rainy season over Indonesia and their

link to ENSO. The onset of rainy season comes later in the El Nino years than the average of most stations (particularly in the south-eastern part of Java). The correlation between rainfall amounts and the southern oscillation index are significantly high in SON (dry) season. Meanwhile, Hendon (2003), by comparing rainfall with another ENSO indices (Nino3.4) in the seasonal time scale, concluded that dry season rainfall anomalies are spatially coherent, strongly correlated with SST, and tightly coupled to ENSO variations. In addition to the study of rainfall variability relation with ENSO, some studies focus on the drought variability analysis correlation with ENSO (D'Arrigo et al., 2008; D'Arrigo & Smerdon, 2008; D'Arrigo & Wilson, 2008; D'Arrigo et al., 2006).

1.3 Research Problem and Justification

Drought in Indonesia has been a problem since the 1200s (Rodysill et al., 2013). The study of severe historical drought in Java, Indonesia recorded the drought occurred in a period of intense El Nino events and Asian monsoon failures in the late 1790s. The drought conditions continued and reached peak intensity in East Java in 1810. Based on the available evidence, they suggested severe multi-decadal drought in the East Java throughout the turn of the 19th century was driven by a combination of heightened El Nino activity and volcanic eruptions.

Despite the long historical record of drought in Indonesia, it is still challenging to understand the disaster completely. It is difficult to detect and monitor drought (Hayes, Svoboda, Wilhite, & Vanyarkho, 1999). Lack of physical evidence of the disaster is one of the reasons. The impact of a drought is less visible and spread over a large geographical extent (Wilhite, 1997), different from other natural hazards. Drought also a creeping phenomenon which makes the onset and termination difficult to determine. The effect of drought accumulate slowly and may continue after sometimes, could be months, seasons even years. Also, the existed of various definition of drought adds confusion about drought, whether it exists and if it is severe or not. For these reasons, it is still difficult task to quantify the drought impacts and to define the disaster relief strategy (Wilhite, 1997).

There have been several studies about the relation of rainfall climatology with ENSO condition (Aldrian, 2003; Aldrian & Dwi Susanto, 2003; Aldrian, Gates, & Widodo, 2007; Aldrian, Gates, & Widodo, 2003; Hamada et al., 2012; Hamada et al., 2002; Haylock & McBride, 2001; Hendon, 2003; Kirono, 2000; Kirono et al., 2015; Nur’utami & Hidayat, 2016). Some studies even focus on the linkage to the drought event (D'Arrigo et al., 2008; D'Arrigo & Smerdon, 2008; D'Arrigo & Wilson, 2008; D'Arrigo et al., 2006; Quinn, Zopf, Short, & Yang Kuo, 1978).

However, most of the studies focus on the meteorological drought. There are only a few studies which focus on the correlation of ENSO to streamflow events (Sahu, Behera, Yamashiki, Takara, & Yamagata, 2012; Sahu et al., 2016; Sahu, Yamashiki, Takara, & Singh, 2011). Mishra and Singh (2010) in their study mentioned the necessity of improving the understanding of the relationships between climatological and hydrological parameters to develop measures to reduce the impacts of droughts. The understanding at the local scale is critical due to the heterogeneity in spatiotemporal hydro-meteorological variability. Thus, this study will investigate the correlation between ENSO and drought variability in the local scale, namely the Lombok River Basin and the Batanghari River Basin. The two study sites were chosen as two of the area has a high risk of drought (Sayaka & Pasandaran, 2017; Suroso et al., 2009).

1.4 Objectives of the Study

This study focuses on the development of drought forecasting method. This study also aims at the better understanding of the historical hydro-meteorological drought in relation to the climate phenomenon such as ENSO. Kirono et al. (2015) have already conducted a study about historical seasonal rainfall correlation with ENSO in the study area (Lombok River Basin). Therefore, this study will focus on the seasonal streamflow correlation with ENSO and its relation with drought forecasting system. The specific objectives of the study are as follows:

• To understand the effect of the climate phenomenon, ENSO specifically, to the hydro- meteorological drought condition in the study area. • To develop the method of drought forecasting based on ENSO indices. • To develop the method of hydrological drought forecasting based on the low flow characteristics.

1.5 References

Aldrian, E. (2003). Simulations of Indonesian rainfall with a hierarchy of climate models. University of Hamburg Aldrian, E., & Dwi Susanto, R. (2003). Identification of three dominant rainfall regions within Indonesia and their relationship to sea surface temperature. International Journal of Climatology, 23(12), 1435-1452. doi:10.1002/joc.950 Aldrian, E., Gates, L. D., & Widodo, F. (2007). Seasonal variability of Indonesian rainfall in ECHAM4 simulations and in the reanalyses: the role of ENSO. Theoretical and Applied Climatology, 87(1-4), 41-59. Aldrian, E., Gates, L. D., & Widodo, F. H. (2003). Variability of Indonesian rainfall and the Influence of ENSO and resolution in ECHAM4 simulations and in the reanalyses: Max- Planck-Institut für Meteorologie.

Alimoeso, S., Boer, R., Subroto, S. W. G., Purwani, E. T., Sugiarto, Y., Rahadiyan R.M.K., & Suciantini. (2002). Penyebaran daerah rawan kering di wilayah pertanaman padi Indonesia. Direktorat Perlindungan Tanaman, Depatemen Pertanian, Jakarta, Indonesia. Alimoeso, S., Jasis, Subroto, W. W. G., Zainita, Mutiawari, D., Fitri, A., . . . Issusilaningtyas. (2000). Impact of extreme climate events and food crop management in Indonesia. Directorate of Crop Protection, Directorate General of Food Crops and Horticulture- Ministry of Agriculture Indonesia, NOAA/OFDA-USA and Asian Disaster Preparedness Center, Bangkok, Thailand. Bappenas & BNPB. (2010). National Action Plan for Disaster Risk Reduction 2010-2012. Boer, R., & Subbiah, A. R. (2005). Agricultural droughts in Indonesia. Monitoring and Predicting Agricultural Drought: A Global Study, 330-344. BPBD. (2015). Rencana nasional penanggulangan bencana 2015-2019: Badan Nasional Penanggulangan Bencana. D'Arrigo, R., Allan, R., Wilson, R., Palmer, J., Sakulich, J., Smerdon, J. E., . . . Ngkoimani, L. O. (2008). Pacific and Indian Ocean climate signals in a tree-ring record of Java monsoon drought. International Journal of Climatology, 28(14), 1889. D'Arrigo, R., & Smerdon, J. E. (2008). Tropical climate influences on drought variability over Java, Indonesia. Geophysical Research Letters, 35(5). D'Arrigo, R., & Wilson, R. (2008). El Nino and Indian Ocean influences on Indonesian drought: implications for forecasting rainfall and crop productivity. International Journal of Climatology, 28(5), 611-616. D'Arrigo, R., Wilson, R., Palmer, J., Krusic, P., Curtis, A., Sakulich, J., . . . Ngkoimani, L. O. (2006). Monsoon drought over Java, Indonesia, during the past two centuries. Geophysical Research Letters, 33(4). Fanin, T., & van der Werf, G. R. (2017). Precipitation–fire linkages in Indonesia (1997–2015). Biogeosciences, 14(18), 3995. Fernandes, K., Verchot, L., Baethgen, W., Gutierrez-Velez, V., Pinedo-Vasquez, M., & Martius, C. (2017). Heightened fire probability in Indonesia in non-drought conditions: the effect of increasing temperatures. Environmental Research Letters, 12(5), 054002. Hamada, J.-I., Mori, S., Kubota, H., Yamanaka, M. D., Haryoko, U., Lestari, S., . . . Syamsudin, F. (2012). Interannual rainfall variability over northwestern Jawa and its relation to the Indian Ocean Dipole and El Niño-Southern Oscillation events. Sola, 8, 69-72. Hamada, J.-I., Yamanaka, M. D., Matsumoto, J., Fukao, S., Winarso, P. A., & Sribimawati, T. (2002). Spatial and temporal variations of the rainy season over Indonesia and their link to ENSO. 気象集誌. 第 2 輯, 80(2), 285-310. Hatmoko, W., Raharja, B., Tollenaar, D., & Vernimmen, R. (2015). Monitoring and Prediction of Hydrological Drought Using a Drought Early Warning System in Pemali-Comal River Basin, Indonesia. Procedia Environmental Sciences, 24, 56-64. Hayes, M. J., Svoboda, M. D., Wilhite, D. A., & Vanyarkho, O. V. (1999). Monitoring the 1996 drought using the standardized precipitation index. Bulletin of the American Meteorological Society, 80(3), 429-438. Haylock, M., & McBride, J. (2001). Spatial coherence and predictability of Indonesian wet season rainfall. Journal of Climate, 14(18), 3882-3887. Hendon, H. H. (2003). Indonesian rainfall variability: Impacts of ENSO and local air-sea interaction. Journal of Climate, 16(11), 1775-1790. King, A. D., Karoly, D. J., & van Oldenborgh, G. J. (2016). Climate Change and El Niño Increase Likelihood of Indonesian Heat and Drought. Bulletin of the American Meteorological Society, 97(12), S113-S117.

Kirono, D. G. (2000). On The Spatial Variation Of The Relationship Between El Nino. Southern Oscillation And Indonesian Seasonal Rainfall. The Indonesian Journal of Geography, 32(2000). Kirono, D. G., Butler, J. R., McGregor, J., Ripaldi, A., Katzfey, J., & Nguyen, K. (2015). Historical and future seasonal rainfall variability in Nusa Tenggara Barat Province, Indonesia: implications for the agriculture and water sectors. Climate Risk Management. Kishore, K. (2000). El Niño 1997-98: Indonesia Country Study: Asian Disaster Preparedness Center. Liu, X., Zhu, X., Pan, Y., Li, S., Liu, Y., & Ma, Y. (2016). Agricultural drought monitoring: Progress, challenges, and prospects. Journal of Geographical Sciences, 26(6), 750-767. Marlier, M. E., DeFries, R. S., Voulgarakis, A., Kinney, P. L., Randerson, J. T., Shindell, D. T., . . . Faluvegi, G. (2013). El Niño and health risks from landscape fire emissions in southeast Asia. Nature Climate Change, 3(2), 131-136. McKee, T. B., Doesken, N. J., & Kleist, J. (1993). The relationship of drought frequency and duration to time scales. Paper presented at the Proceedings of the 8th Conference on Applied Climatology. Mishra, A. K., & Singh, V. P. (2010). A review of drought concepts. Journal of Hydrology, 391(1), 202-216. Murdiyarso, D., & Adiningsih, E. S. (2007). Climate anomalies, Indonesian vegetation fires and terrestrial carbon emissions. Mitigation and Adaptation Strategies for Global Change, 12(1), 101-112. Nur’utami, M. N., & Hidayat, R. (2016). Influences of IOD and ENSO to Indonesian rainfall variability: role of atmosphere-ocean interaction in the Indo-Pacific sector. Procedia Environmental Sciences, 33, 196-203. Olukayode Oladipo, E. (1985). A comparative performance analysis of three meteorological drought indices. International Journal of Climatology, 5(6), 655-664. Patel, N., Chopra, P., & Dadhwal, V. (2007). Analyzing spatial patterns of meteorological drought using standardized precipitation index. Meteorological Applications, 14(4), 329-336. Quinn, W., Zopf, D., Short, K., & Yang Kuo, R. (1978). Historical trends and statistics of the Southern Oscillation, El Nino and Indonesian droughts. Fish. Bull, 76, 663-678. Rodysill, J. R., Russell, J. M., Crausbay, S. D., Bijaksana, S., Vuille, M., Edwards, R. L., & Cheng, H. (2013). A severe drought during the last millennium in East Java, Indonesia. Quaternary Science Reviews, 80, 102-111. Ryu, J. H., Svoboda, M. D., Lenters, J. D., Tadesse, T., & Knutson, C. L. (2010). Potential extents for ENSO-driven hydrologic drought forecasts in the United States. Climatic Change, 101(3), 575-597. Sahu, N., Behera, S. K., Yamashiki, Y., Takara, K., & Yamagata, T. (2012). IOD and ENSO impacts on the extreme stream-flows of Citarum river in Indonesia. Climate Dynamics, 39(7-8), 1673-1680. Sahu, N., Robertson, A. W., Boer, R., Behera, S., DeWitt, D. G., Takara, K., . . . Singh, R. (2016). Probabilistic seasonal streamflow forecasts of the Citarum River, Indonesia, based on general circulation models. Stochastic Environmental Research and Risk Assessment, 1-12. Sahu, N., Yamashiki, Y., Takara, K., & Singh, R. B. (2011). An Observation on the Relationship between Climate Variability Modes and River Discharges of the Citarum Basin, Indonesia. Sayaka, B., & Pasandaran, E. (2017). Stage of Development in River Basin Management in Indonesia. Analisis Kebijakan Pertanian, 4(1), 70-82.

Schweithelm, J., & Glover, D. (1999). Causes and impacts of the fires. Indonesia’s fires and haze: The cost of catastrophe, 1-13. Siegert, F., Ruecker, G., Hinrichs, A., & Hoffmann, A. (2001). Increased damage from fires in logged forests during droughts caused by El Nino. Nature, 414(6862), 437-440. Smakhtin, V., & Hughes, D. (2007). Automated estimation and analyses of meteorological drought characteristics from monthly rainfall data. Environmental Modelling & Software, 22(6), 880-890. doi:10.1016/j.envsoft.2006.05.013 Suroso, D., Hadi, T. W., Sofian, I., Latief, H., Abdurahman, O., & Setiawan, B. (2009). Vulnerability of small islands to climate change in Indonesia: a case study of Lombok Island, Province of Nusa Tenggara Barat. The Language Center of National Education Center. (Ed.) (2008) Indonesian Dictionary (4 ed.). Wilhite, D. A. (1997). Improving drought management in the West: The role of mitigation and preparedness. Wilhite, D. A., & Glantz, M. H. (1985). Understanding: the drought phenomenon: the role of definitions. Water international, 10(3), 111-120.

2 Study Area and Data

2.1 Introduction

This chapter introduces the location and physiography of the study area. There two study areas in this study; namely Lombok River Basin and Batanghari River Basin. Both of the River Basin are affected by drought disasters, especially during the dry season (April – October). The description of the dataset used in the study is also presented.

2.2 The Lombok River Basin

Lombok River Basin is located in West Nusa Tenggara (NTB) Province under the management of River Basin Organization of Nusa Tenggara I (BWS-NT1). Based on the administrative area of the government, the Lombok River Basin is divided into four regencies and one city; West Lombok, Central Lombok, East Lombok, North Lombok and Mataram City. The total area of the River Basin is 4,738.65 km2. The area and administrative border of every regency and city in the island are presented in Table 2-1 and Figure 2-1.

Table 2-1 Administrative area of the Lombok Island Area Percentage No. of No. of Regency (km2) (%) Subdistrict Villages Mataram City 61.3 1.29 6 50 West Lombok 1053.87 22.24 10 122 Central Lombok 1208.4 25.50 12 139 East Lombok 1605.55 33.88 20 254 North Lombok 809.53 17.08 5 33 Total 4738.65 100 53 598 Source: BPS, 2007

Figure 2-1 Administrative map of the Lombok River Basin.

The total population of Lombok Island in 2007 was 3,015,245 people (70.83 % of the total population of NTB Province). The average density was 636 people/km2 and the population growth rate of 1.98% per year, as shown in Table 2-2. In its development, the population in each Regency/City in Lombok Island is increasing every year. Based on the results of the analysis, it is known that the percentage of population growth on the Island reaches 1.98% (BPS, 2007).

Table 2-2 Total population of Lombok Island per Regency Male Female Total Regency population population population Density Percentage (people) (People) (people) (people/km2) (%) Mataram City 222,596 227,630 450,226 7,344.63 13 West Lombok 320,103 334,789 654,892 621.42 19 Central Lombok 431,825 481,054 912,879 755.44 27 East Lombok 542,012 622,106 1,164,118 725.06 34 North Lombok 104,573 107,692 212,265 262.21 6 Total 1,621,109 1,773,271 3,394,380 716.32 100 Source: BPS, 2007

2.2.1 Social Economic Conditions

The economic growth drivers of each regency/city in Lombok Island rely on agriculture, commerce, services, and transportation. Through the development of agribusiness, forestry, and growth of small and medium-sized enterprises, as well as tourism, it is expected the development of sustainable economic can be realized. The economic growth in 2007 was 4.89% (BPS, 2007).

Based on the GRDP statistical data, agriculture sector gives the dominant contribution. In 2014, this sector contributed 25.4%, followed by commercial/retail (15%), mining (12.7), construction (10.1%) and transportation (13%) (BPS, 2014b). The GDP growth of each Regency/City in the period 2003 to 2007 is increasing by 17.8%. The per capita income in Lombok Island also shows increased signal by 8.4% (BPS, 2007).

Figure 2-2 GRDP per sector (BPS, 2014).

2.2.2 Climate

Lombok Island has a wet tropical climate which is influenced by the Asian Monsoon. The dry southeast monsoon causes dry season (generally May to October), and wet northwest monsoon causes rainy season (usually in November or December until March or April) with the rainfall characteristics is below normal (B). Average rainfall in Lombok Island is 1593.36 mm. The spatial distribution of annual rainfall in the Lombok River Basin is shown in Figure 2-3. The eastern part of the Lombok River Basin has the lowest range of annual rainfall with the amount of rainfall less 1000mm. The hydro-meteorological characteristics in Lombok Island are presented in Table 2-3.

Figure 2-3 Spatial distribution of annual rainfall in the Lombok River Basin.

Table 2-3 Hydro-meteorological characteristics in Lombok Island No. Description Unit Wet season Dry season Average (Nov - Apr) (May - Oct) 1 Maximum temperature °C 25 - 34 25 - 34 30.5 2 Minimum temperature °C 17 -28 18 - 26 21.8 3 Maximum humidity % 72 - 100 67 - 100 80 4 Minimum humidity % 65 - 67 67 - 84 70 5 Maximum air pressure Mbar 1,010 - 1,014 1,011 - 1,016 1,013.70 6 Minimum air pressure Mbar 1,004 - 1,009 1,006 - 1,013 1,008.30 7 Sunshine % 1 - 91 12 - 95 54 8 Wind direction ∘ 120 - 360 130 - 310 231 9 Wind speed Knot 2 - 6.1 4 - 6 4.8 10 Rainfall mm/month 7 - 458 1 - 335 133 11 Rainy days days 1 - 25 1 - 15 8 3 12 Surface runoff m /s/km 0.0004 - 0.19 0.0004 - 0.0414 0.0161 Source: Hydrological Division of Department of Infrastructure Services of NTB Provinces, 2007

2.2.3 Water Resources

The annual potential of surface water availability with a reliability level of 80% in the Lombok River Basin is 90.2 m3/s. Irrigation sector dominates the surface water usage for about 77.50% which then is followed by domestic, municipal, and industrial (DMI) and fisheries together with livestock sectors for approximately 15.7% and 6.6% consecutively. Along with the population and economic growth, the water needs for various sectors will also continue to increase. The full usage of the water supply is presented in Figure 2-4.

Figure 2-4 Water usage distribution (BWS-NT1, 2010).

Suroso et al. (2009) have done a study about the vulnerability of Lombok Island to Climate Change. They assessed the vulnerability or water resources sector in the island by using water balance as the primary indicator. The water balance consists of supply (surface water, groundwater) and demand (irrigation, domestic and industry). The result of the vulnerability map of water sector is presented in Figure 2-5. Based on the vulnerability map of the water sector, several cities/regency was susceptible to the decreasing of the water availability, including Mataram City, some part of Central Lombok, East Lombok and West Lombok Regency. The level of vulnerability is in accordance with the degree of hazard. The high population density in Mataram City is the main reason for the high vulnerability to water resources.

Figure 2-5 The vulnerability map of water sector of Lombok Island (green color represent not vulnerable area, and red color represent extremely vulnerable area).

2.3 The Batanghari River Basin

The Batanghari River Basin is located in two Provinces; Jambi and West Sumatera Provinces. There are in total 13 regencies and one city that is included in The River Basin. The administrative area of the Batanghari River Basin is presented in Table 2-4 and Figure 2-6. Based on the Decree No. 7 of 2004 on Water Resources Article 14 letter b, the Government has the authority and responsibility to establish a water resources management for inter- provincial river basin areas, cross-country river basin areas, and national strategic river basin areas. In accordance with this article, since the Batanghari River Basin is located in two provinces, the authority stipulates the Water Resources Management of Batanghari River Basin under the Central Government of Indonesia. The River Basin Organization that manages the Batanghari River Basin is the River Basin Organization Sumatera VI.

Table 2-4 Administrative Area within the Batanghari River Basin boundary River Basin Province Regency/City Upstream Batangahari Jambi Bungo, Tebo, Kerinci (12,779.47 km2) West Sumatera Solok, South Solok, Sawah-lunto/Sijunjung, Dharmasraya Batang Tebo Jambi Bungo, Tebo, Kerinci (5,387.25 km2) Batang Tabir Jambi Merangin, Tebo, (3,813.29 km2) Batanghari, Kerinci, Batanghari, Sarolangun Batang Merangin-Tembesi Jambi Merangin, Kerinci (12,819.07 km2) Dwosntream Batanghari Jambi Jambi City, Tebo, (9,795.59 km2) Tanjung Jabung Timur, Muaro Jambi, Batanghari Source: Study of Batanghari River Basin land-use (2004) and map overlap 2008

Figure 2-6 Administrative area of the Batanghari River Basin, in Sumatra Island (Source: BMKG).

Most of the population in the Batanghari River Basin lives in Jambi City, which has the smaller area. The condition makes Jambi City has the highest density of population in the Batanghari River Basin (Table 2-5).

Table 2-5 Population density in the Batanghari River Basin Total Regency/ Area population Density City (km2) (people) (people/km2) Kerinci 4,200.00 306,494 73 Merangin 7,679.00 277,595 36 Sarolangun 6,174.00 205,090 33 Batanghari 4,983 211,897 43 Muaro Jambi 5,419.68 295,319 54 East Tanjab 5,445.00 207,340 38 West Tanjab 5,503.50 239,016 43 Tebo 7,862 246,044 31 Bungo 7,160 250,934 35 Jambi City 205 443,370 2,163 Solok 3,738.00 347,288 93 South Solok 3,346 135,744 41 Sawahlunto 3130.8 196,667 63 Dharmasraya 2,961.13 173,375 59

2.3.1 Social Economic Conditions

The main livelihood in the Jambi Province is agriculture accounting for about 54.81%, followed by commercial and services for 16.26% and 14.39% consecutively (BPS, 2014a). The livelihood sector in Jambi Province is presented in Figure 2-7.

Figure 2-7 Livelihood sectors in Jambi Province.

2.3.2 Climate

Based on Schmidt and Ferguson climatic classification, the Batanghari River Basin climate is Am (wet) with year-round rainfall. In 2002 the average wet month was ten months and dry month average of one month. The annual average rainfall in the river basin is 2500 mm/year with the number of rainy days on average 12 days/month. The average temperature of the region is 28oC with an average humidity of 83% (BWS-SumatraVI, 2012). The climate characteristics in the Batanghari River Basin is presented in Table 2-6.

Table 2-6 Climate condition in the Batanghari River Basin (upstream, central and downstream) area The Batanghari River Basin No Climate condition Upstream Central Downstream 1 Climate classification Very wet (Af) Wet (Am) Wet (Am) (Schimdt&Ferguson) 2 Wet month average 12 months 10 months 10months 3 Dry month average - 1 month 1month 4 Annual rainfall 3000 mm/h 2340 mm/h 2271 mm/h 5 Rainy days 13days/month 6 Altitude (MAMSL) 500 - 3000 100 - 500 0 - 100 Source: Water Resources Management Pattern by River Basin Organization Sumatra IV

In the Batanghari River Basin, there are 12 rainfall stations; four stations are located in the Upstream Batanghari River Basin, four stations are located in the Central Batanghari River

Basin, and four stations are located in the Downstream Batanghari River Basin. From the data of Sicinsin Climatology station, in the Upstream Batanghari River Basin area there are four units of rainfall station with rainfall observed data available from 1993 to 2002, namely; Lubuk Gadang Station, Muara Labuh Station, Sitiung Station, and Tanjung Gadang Station. The rainfall data for the Central Batanghari River Basin and the Downstream Batanghari Basin were recorded from rain gauge stations in Kerinci, Bungo, Tebo, and Sarko Regency. The rainfall data for the downstream of the Batanghari River Basin is recorded from the rain gauge station in the Tanjung Jabung Regency, Batanghari Regency, Muaro Jambi Regency, and Jambi City. The distribution of rain gauge station, water level station, and the climatological station is presented in Figure 2-8. Annual average rainfall in the Jambi Province is presented in Figure 2-9.

Figure 2-8 Distribution of hydrological station in the Batanghari River Basin (Source: River Basin Organization of Sumatra VI).

Figure 2-9 Annual average rainfall in the Jambi Province (Source: BMKG).

2.3.3 Water Resources

According to the Directorate of Water Resources of the Regional West Department of Infrastructure in 2003, in general, the available water discharge on the Batanghari River Basin based on existing potential has been sufficient. It is indicated by the amount of the Batanghari River Basin discharge of 35,953.18 m3/sec (8.3 billion m3/year), while the water demand is only 4,668.96 m3/sec to meet the needs of various sectors (BWS-SumatraVI, 2012). Nevertheless still need to be aware of the tendency to increase the need for various water sectors in line with the increasing development and the regional economy.

2.4 Data

There are three primary data used in this study; rainfall, discharge, and sea surface temperature data.

2.4.1 The Lombok River Basin Hydro-meteorological Data

The rainfall and discharge data of the Lombok River Basin was provided by River Basin Organization of Nusa Tenggara I (BWS-NT1) and Water Resources Information Center (BISDA) of West Nusa Tenggara Province. The data is listed in Table 2-7. In total there are 44 rainfall stations in the river basin that are managed by BWS-NT1 and BISDA. However, 21 stations have no coordinate information. The map of rainfall stations distribution is presented

in Figure 2-10. The stations that do not have coordinate information were omitted from analysis. Finally, only the stations with data longer than 30 years were used in the analysis.

Figure 2-10 Rain gauge station distribution in the Lombok River Basin.

Table 2-7 Rainfall data availability in the Lombok River Basin Data type and Period without Station Period Source X Y station missing data (years) Rainfall Kabul 1975 - 2015 BISDA 38 409813.7 9029403 Rainfall Mangkung 1975 - 2015 BISDA 38 412726.9 9024403 Rainfall Sepit 1974 - 2015 BISDA 38 440971.4 9033175 Rainfall Pringgabaya 1978 - 2015 BISDA 38 459008.3 9053988 Rainfall Sesaot 1974 - 2015 BISDA 35 415992.2 9056473 Rainfall Keru 1981 - 2015 BISDA 34 418627.1 9053561 Rainfall Jurang Sate 1971 - 2015 BISDA 34 420222.8 9050308 Rainfall Gunungsari 1983 - 2015 BISDA 33 400707.2 9055857 Rainfall Santong 1980 - 2015 BISDA 33 421819.2 9080071 Rainfall Pengadang 1982 - 2015 BISDA 33 426078.7 9040368 Rainfall Ijobalit 1982 - 2015 BISDA 33 451589.4 9045903 Rainfall Kuripan 1982 - 2015 BISDA 32 408688.7 9040825 Rainfall Lingkok Lime 1983 - 2015 BISDA 32 429660.1 9055116 Rainfall Loang Make 1982 - 2015 BISDA 32 434671.7 9036728 Rainfall Sapit 1974 - 2015 BISDA 32 450379.6 9061165 Rainfall Rembitan 1975 - 2015 BISDA 31 421895.8 9022118 Rainfall Perian 1989 - 2015 BISDA 16 432840.1 9054661 Rainfall Batujai 1985 - 2015 BISDA 11 418114 9034733 Rainfall Monjok 2009 - 2015 BISDA 7 402794 9052483 Rainfall Keruak 1974 - 2015 BISDA 39 no information Rainfall Paradowane 1980 - 2015 BISDA 34 no information Rainfall Gapit 1978 - 2015 BISDA 33 no information Rainfall Sumi 1981 - 2015 BISDA 33 no information Rainfall Rea Atas 1980 - 2015 BISDA 32 no information Rainfall Sopak 1972 - 2015 BISDA 32 no information Rainfall Utan Rhee 1978 - 2015 BISDA 32 no information Rainfall Taliwang 1975 - 2005 BISDA 31 no information Rainfall Semongkat 1984 - 2015 BISDA 29 no information Rainfall Tepas 1978 - 2007 BISDA 29 no information Rainfall Pungkit Atas 1980 - 2015 BISDA 24 no information Rainfall Sambelia 1990 -2015 BISDA 23 no information Rainfall Kopang 1991 - 2015 BISDA 22 no information Rainfall Kadindi 1980 - 2015 BISDA 21 no information Rainfall Godo 1989 - 2014 BISDA 20 no information Rainfall Pengga 1992 -2008 BISDA 15 no information Rainfall Dompu 1999 - 2015 BISDA 14 no information Rainfall Plampang 1993 - 2015 BISDA 13 no information Rainfall Tawali 2010 - 2015 BISDA 6 no information Rainfall Kumbe 2011 - 2015 BISDA 5 no information Rainfall Belanting 2008 - 2014 BWS - NT1 7 458172 9081231 Rainfall Bertais 2008 - 2014 BWS - NT1 7 408658 9049821 Rainfall Sekotong 2011 - 2014 BWS - NT1 4 no information Rainfall Jurang Malang 2013 - 2014 BWS - NT1 2 420323 9057647 Rainfall Pelangan 2012 - 2014 BWS - NT1 3 382744.4 9027765

The streamflow data used in this study was also collected from two institutions; BISDA and BWS-NT1. The total number of discharge station managed by BISDA and BWS-NT1 is 32 stations. 11 stations of it have no coordinate information. The data for analysis is the data with length more than 20 years. List of discharge station is presented in Table 2-8. The distribution of the discharge station is shown in Figure 2-11.

Figure 2-11 Discharge station distribution in the Lombok River Basin.

Table 2-8 Discharge data availability in the Lombok River Basin Period without Data type Station Period Source X Y missing data (years) Discharge Tanjung 1986 - 2015 BISDA 25 408097.2 9074147.4 Discharge Perampuan 1986 - 2015 BISDA 25 420868.1 9048497.4 Discharge Belencong 1980 - 2015 BISDA 24 402238.1 9054877.4 Discharge Gebong 1985 - 2015 BISDA 24 413257.7 9048175.3 Discharge Semaya 1986 - 2015 BISDA 24 440590.5 9043186.1 Discharge Loloan 1985 - 2015 BISDA 24 442365.6 9086767.2 Discharge Bug-Bug 1985 - 2015 BISDA 23 407043.6 9052277.6 Discharge Dasan Tengak 1986 - 2015 BISDA 23 410640.5 9071849.3 Discharge Santong 1985 - 2015 BISDA 18 422524.2 9079273.7 Discharge Lantan Daya 1985 - 2015 BISDA 17 425139.4 9053019.9 Discharge Tempasan 1989 - 2015 BISDA 15 441341.8 9052345.2 Discharge Sopak 1985 - 2015 BISDA 14 436126.9 9085284.4 Discharge Suradadi 1986 - 2015 BISDA 13 435669.3 9043701.0 Discharge Aiknyet 1985 - 2015 BISDA 11 416267.4 9056473.8 Discharge Ponggong 1985 - 2015 BISDA 11 429559.5 9042155.5 Discharge Kr. Anyar 2014 - 2015 BISDA 0 401939.1 9038875.0 Discharge Kr. Makam 2014 - 2015 BISDA 0 407833.4 9040639.1 Discharge Empang 1994 - 2015 BISDA 17 no information Discharge Keling 1985 - 2015 BISDA 16 no information Discharge Kumbe 2001 - 2015 BISDA 8 no information Discharge Matua 1986 - 2015 BISDA 21 no information Discharge Presak 1985 - 2015 BISDA 12 no information Discharge Sari 1989 - 2015 BISDA 19 no information Discharge Tawali 1999 - 2015 BISDA 14 no information Discharge Tepas 1985 - 2015 BISDA 16 no information Discharge Utan Rhee 1994 - 2015 BISDA 10 no information Discharge Ancar 2008 - 2014 BWS-NT1 7 401896.5006 9050567.53 Discharge Belanting 2008 - 2014 BWS-NT1 7 458046.4336 9081103.63 Discharge Jurang Malang 2013 - 2014 BWS-NT1 2 419436.9279 9056879.14 Discharge Orong 2013 - 2014 BWS-NT1 2 406547.9715 9056619.34 Discharge Petitik 2014 BWS-NT1 1 no information Discharge Srigangga 2014 BWS-NT1 1 no information

2.4.2 The Batanghari River Basin Hydro-meteorological Data

We used daily rainfall data in Sultan Thaha Station which is located at 1.633°S and 103.650°E. The data is managed by the Indonesian Agency for Meteorology, Climatology, and Geophysics (BMKG). The analyses are conducted to the data from 1985 until 2012 (28 years). This station is chosen since the data availability of this station was the most without a significant missing period in the 28 years. We aggregated the daily data into monthly and seasonal for the analysis. We define the season into the following four seasons, namely Dec-Jan-Feb (DJF) as wet season, Mar-Apr-May (MAM) as the transition wet to dry season, Jun-Jul-Aug (JJA) as dry season, and Sep-Oct-Nov (SON) as the transition from dry to wet season based on previous studies (Aldrian & Dwi Susanto, 2003).

2.4.3 Sea Surface Temperature Data

Sea Surface Temperature (SST) data in this study is used to represent the El Nino Southern Oscillation (ENSO) events. The previous studies investigating the relation between ENSO and rainfall variability in Indonesia used some indices to indicate the ENSO condition. Some studies used Sea Surface Temperature (SST) in the Nino 3 area to represent ENSO (Aldrian & Dwi Susanto, 2003; Aldrian, Gates, & Widodo, 2007; Sahu, Behera, Yamashiki, Takara, & Yamagata, 2012; Sahu, Yamashiki, Behera, Takara, & Yamagata, 2012; Trenberth, 1997), while others used the Southern Oscillation Index (SOI) (D'Arrigo et al., 2008; Hamada et al., 2002; Haylock & McBride, 2001; Kirono, 2000) and SST in the Nino 3.4 region (D'Arrigo et al., 2008; D'Arrigo & Smerdon, 2008; D'Arrigo & Wilson, 2008; Hidayat, Ando, Masumoto, & Luo, 2016; Hou et al., 2016; Trenberth, 1997) to examine the ENSO condition.

In this study, we will use two indices to represent ENSO:

• Nino3.4 index The Nino 3.4 index was calculated from HadISST1. It is the area averaged SST from 5°S- 5°N and 170-120°W (Rayner et al., 2003). The data was downloaded from https://www.esrl.noaa.gov/psd/gcos_wgsp/Timeseries/Nino34/. • Oceanic Nino Index (ONI) ONI is 3-month running mean of ERSST.v4 SST anomalies in the Niño 3.4 region (5°N- 5°S, 120-170°W). Warm and cold periods based on a threshold of +/- 0.5oC for the ONI. The data was downloaded from

http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ensoyears.shtml. The ENSO condition is determined based on this data.

2.5 References

Aldrian, E., & Dwi Susanto, R. (2003). Identification of three dominant rainfall regions within Indonesia and their relationship to sea surface temperature. International Journal of Climatology, 23(12), 1435-1452. doi:10.1002/joc.950 Aldrian, E., Gates, L. D., & Widodo, F. (2007). Seasonal variability of Indonesian rainfall in ECHAM4 simulations and in the reanalyses: the role of ENSO. Theoretical and Applied Climatology, 87(1-4), 41-59. doi:https://doi.org/10.1007/s00704-006-0218-8 BPS. (2007). Nusa Tenggara Barat dalam Angka 2007. BPS. (2014a). Jambi in Figure 2014. Jambi, Indonesia: See http://jambiprov. go. id. BPS. (2014b). Nusa Tenggara Barat Dalam Angka 2014. BWS-NT1. (2010). Pola Pengelolaan Sumber Daya Air Wilayah Sungai Pulau Lombok. BWS-SumatraVI. (2012). Pola Pengelolaan Sumber Daya Air Wilayah Sungai Batanghari. D'Arrigo, R., Allan, R., Wilson, R., Palmer, J., Sakulich, J., Smerdon, J. E., . . . Ngkoimani, L. O. (2008). Pacific and Indian Ocean climate signals in a tree-ring record of Java monsoon drought. International Journal of Climatology, 28(14), 1889. D'Arrigo, R., & Smerdon, J. E. (2008). Tropical climate influences on drought variability over Java, Indonesia. Geophysical Research Letters, 35(5). D'Arrigo, R., & Wilson, R. (2008). El Nino and Indian Ocean influences on Indonesian drought: implications for forecasting rainfall and crop productivity. International Journal of Climatology, 28(5), 611-616. Hamada, J.-I., Yamanaka, M. D., Matsumoto, J., Fukao, S., Winarso, P. A., & Sribimawati, T. (2002). Spatial and temporal variations of the rainy season over Indonesia and their link to ENSO. 気象集誌. 第 2 輯, 80(2), 285-310. Haylock, M., & McBride, J. (2001). Spatial coherence and predictability of Indonesian wet season rainfall. Journal of Climate, 14(18), 3882-3887. Hidayat, R., Ando, K., Masumoto, Y., & Luo, J. (2016). Interannual Variability of Rainfall over Indonesia: Impacts of ENSO and IOD and Their Predictability. Paper presented at the IOP Conference Series: Earth and Environmental Science. Hou, X., Zhu, B., Fei, D., Zhu, X., Kang, H., & Wang, D. (2016). Simulation of tropical tropospheric ozone variation from 1982 to 2010: The meteorological impact of two types of ENSO event. Journal of Geophysical Research: Atmospheres, 121(15), 9220- 9236. Kirono, D. G. (2000). On The Spatial Variation Of The Relationship Between El Nino. Southern Oscillation And Indonesian Seasonal Rainfall. The Indonesian Journal of Geography, 32(2000).

Rayner, N., Parker, D. E., Horton, E., Folland, C., Alexander, L., Rowell, D., . . . Kaplan, A. (2003). Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. Journal of Geophysical Research: Atmospheres, 108(D14). Sahu, N., Behera, S. K., Yamashiki, Y., Takara, K., & Yamagata, T. (2012). IOD and ENSO impacts on the extreme stream-flows of Citarum river in Indonesia. Climate Dynamics, 39(7-8), 1673-1680. Sahu, N., Yamashiki, Y., Behera, S., Takara, K., & Yamagata, T. (2012). Large impacts of Indo-Pacific climate modes on the extreme streamflows of Citarum river in Indonesia. Journal of global environment engineering, 17, 1-8. Suroso, D., Hadi, T. W., Sofian, I., Latief, H., Abdurahman, O., & Setiawan, B. (2009). Vulnerability of small islands to climate change in Indonesia: a case study of Lombok Island, Province of Nusa Tenggara Barat. Trenberth, K. E. (1997). The definition of el nino. Bulletin of the American Meteorological Society, 78(12), 2771-2777.

3 Hydrological Drought Correlation with El Nino in Lombok River Basin 3.1 Introduction

Hydrological extreme events such as floods and droughts, cause significant impacts to economic, social, and environment every year in Indonesia. Droughts events happened in Indonesia during dry seasons and became more severe when the El Nino phenomenon occurs at the same time. The 1997-98 drought related to El Nino caused a severe forest fire for about 130,000 km2 area with the economic loss around 5 million USD (BBC, 2015). Besides that, food commodity skyrocketed and became prohibitively expensive. The study by Kirono, Tapper, and McBride (1999) shows that during the El Nino 1997-98, the rainfall amount in whole Indonesia is below the 10th percentile, with many stations recording their lowest rainfall on record.

West Nusa Tenggara (NTB) Province is one of the areas most affected by El Nino in Indonesia (Fitri & Hermawan, 2015). In NTB during 2015 El Nino for example, some areas have experienced a meteorological drought marked with no rain for 60 days by 21 July 2015 (Dayantolis, 2015), while the dry season itself starts from April (Kirono et al., 2015). The severe drought in 2015 caused agricultural drought for 51.1 km2 with 8.76 km2 of it suffered from crop failure (Ridwan & Hazliansyah, 2015). The data from Regional Board of Disaster Management of NTB Province also shows increasing areas affected by drought from 2014 to 2015 by 61.86%. The comparisons of number villages affected by drought in 2014 and 2015 are presented in Figure 3-1.

300 254 250 200 139 150 122

100 67 55 39 33 50 28 30 23 17 27 0 West Lombok Central Lombok North Lombok East Lombok Total Villages 2014 2015

Figure 3-1 Number of villages affected by drought in Lombok Island (Source: BPBD of NTB).

NTB is one of the national granaries in Indonesia (BPS, 2016). Despite the small size of the province, NTB ranks five of the biggest rice producers in Indonesia (Figure 3-2). As NTB is

one of the Provinces as the biggest rice producers in Indonesia, it is necessary to understand the drought mechanism in the area to prevent bigger lost due to the disaster.

1200000

1000000

800000

600000 (tonnes) 400000

200000

Figure 3-2 Areas as biggest rice producers in Indonesia (Kompas, 2013).

Mishra and Singh (2010) stated that the droughts around the globe are related to large-scale climate condition. Understanding the relationship between climate indices and drought will be useful for any drought prediction. It because drought is a complex natural hazard which is best characterized by climatological and hydrological parameters. Improving the understanding of the relationships between parameters is necessary to develop measures to reduce the impacts of droughts. Therefore, the knowledge of the correlation between droughts with climatic, oceanic, and local factors is required to prevent the effects of a drought proactively by addressing vulnerabilities through a risk management approach.

This study aims to investigate the impact of El Nino to the hydro-meteorological drought event in the Lombok River Basin, NTB. The streamflow drought is compared with ENSO indices by using simple correlation (r). The lag correlation is investigated to determine the predictability of severe drought.

3.2 Study Area and Data

Data used in this study consist of rainfall records, discharge records, and Oceanic Nino Index (ONI) as ENSO index. Daily rainfall and discharge records from the Department of Public Works in West Nusa Tenggara were processed into monthly time series data, which later is being used for calculation of drought index. Monthly rainfall is mainly needed for the meteorological drought analysis, while the monthly discharge is needed for the hydrological drought analysis.

Lombok Island is selected to be the study area. There are in total 115 catchments area on the island (BWS-NT1, 2010). Two catchments were selected for the hydrological drought assessment study, namely the Jangkok Catchment and the Babak Catchment. The selection of the two catchments was based on the quality of the discharge data in each catchment. Discharge data quality examination was done for all of the discharge data that has already been mention in Table 2.8. Only two stations with a good and reasonable discharge data for nearly 20 years data were available in the study area. Thus, only two catchments area were selected for the hydrological drought assessment study. The location of the catchment along with the rainfall and discharge stations is presented in Figure 3-3.

Figure 3-3 Study area: the Jangkok Catchment and the Babak Catchment in Lombok Island.

Different from hydrological drought analysis which was done only for two catchments, the meteorological drought analysis was done for the entire study area (Lombok Island). The decision was based on the data availability, which is the availability of rainfall data is more than the discharge data. A total of 15 stations with monthly rainfall data from 1983 to 2015 (23 years) were selected for meteorological drought analysis in the Island. The spatial distribution of rainfall stations used is presented in Figure 3-4. The spatial precipitation pattern in Lombok

varies with annual rainfall maximum (> 2000 mm) in the northwest towards minimum (< 1000 mm) in the southeast part of the island (Figure 3-5). The distribution of the rainfall reflects the topography condition, which the amounts tend to be higher in high elevation. Average monthly rainfall of Lombok Island (Figure 3-6) shows the highest precipitation occurs in the wet season (Dec-Mar) and the lowest precipitation happens in the dry season (Jun-Sep).

Figure 3-4 Spatial distribution of rainfall stations used in the meteorological drought analysis.

Figure 3-5 Spatial distribution of annual rainfall characteristics.

300

250

200

150

Rainfall (mm) 100

Figure 3-6 Average monthly rainfall of Lombok Island.

There are so some ENSO indices that have been already existed and used for ENSO studies for decades. To identify El Nino, Trenberth (1997) suggested to use 5-month running means of SST anomalies in the Nino 3.4 regions ((5°N-5°S, 120-170°W). The El Nino is set to occur if the 5-month running means of SST anomalies in the Nino 3.4 region exceed 0.4°C for six months or more. Another ENSO index that has been widely used by scientist to determine ENSO events is Nino 3 (Aldrian & Dwi Susanto, 2003). In 2007, Aldrian et al. used SSET data from GISST2 (Global Ice and Sea Surface Temperature) dataset and determined the ENSO years from the data. Here are the El Nino and La Nina events based on Aldrian et al. (2007).

El Nino year:

1965/1966, 1969/1970, 1972/1973, 1982/1983, 1987/1988, 1991/1992

La Nina year:

1964/1965, 1970/1971, 1973/1974, 1975/1976, 1988/1989

Aldrian et al. (2007) in their study only defined the historical ENSO events until 1992. To analysis the recent ENSO event occurrences, this study uses Oceanic Nino Index (ONI). ONI is three months running mean of Extended Reconstructed Sea Surface Temperature (ERSST).v4 SST anomalies in the Nino 3.4 region (5oN-5oS, 120o-170oW). El Nino event is defined as the ONI above 0.5, while La Nina event is characterized by the ONI value below 0.5. By using the threshold, we obtained the La Nina and El Nino events. The value of ONI from 1980 to 2016 is presented in Figure 3-2 and Figure 3-8. The red color identifies the El Nino while the blue color shows La Nina events. The climate data used in this study is sea surface temperature (SST) daily data obtained from NOAA Optimum Interpolation (OI) SST

35 v2 downloaded from https://www.esrl.noaa.gov/psd/data/gridded/data.noaa.oisst.v2.html. The spatial coverage of the data is a 1x1 global grid (Reynolds, Rayner, Smith, Stokes, & Wang, 2002). The Nino3.4 data was generated from the OISST v2 data by taking the average of the data in the 5°North-5°South and 170-120°West (Figure 3-7). This Nino 3.4 data then being used to analyze the ENSO condition. The data availability is from 1981/12 to 2017/05.

Figure 3-7 Location of Nino3.4 region. (Source: https://www.ncdc.noaa.gov/teleconnections/enso/indicators/sst.php)

ONI ONI (El Nino) ONI (La Nina) El Nino threshold La Nina threshold 3 2 1 0 -1 -2 -3

Figure 3-8 Illustration of El Nino and La Nina event based on ONI.

Table 3-1 The Oscillation Nino Index (ONI) used to define the El Nino and La Nina events (Source: NOAA)

From 1980 to 2016, it is noted there are 10 El Nino events; 1982-83, 1987-88, 1991-92, 1994- 95, 1997-98, 2002-03, 2004-05, 2006-07, 2009-10, 2014-16. The strongest El Nino event was 1997-98 and 2015-16 events. The effect of the 1997-98 El Nino in Indonesia was quite strong. A severe fire which burns 9.5 million hectares of the plantation and also food shortages was

inevitable at the time. The estimated loss was around US$9billion (Kirono et al., 1999). The 2015 El Nino was also brought a severe forest fire in Indonesia (Whitburn et al., 2016).

3.3 Methods

3.3.1 Drought index

Drought index is commonly used for drought identification. The investigation of meteorological and hydrological drought was done by using discharge index. The discharge index follows the equation:

X m − X (um ,t) X i (um ,t) = (3.1) σ m

where Xi(um,t) is observed time series of rainfall (P) or discharge (Q) at a station m among N (N = 15 stations for rainfall and N = 2 stations for discharge) in a certain domain Ω, u corresponds to a two-dimensional plane u = (x,y) and t is time, Xi is precipitation (P) or discharge (Q) index, is average monthly precipitation (P) or discharge (Q), and σ is the

standard deviation of the𝑋𝑋� monthly discharge.

The equation of drought index is used to show that for the higher value of X (P or Q), which

means the wetter of the river basin condition, the smaller the drought index (Xi) becomes. While for the condition of the dry river basin, the index will become higher. By using this definition of the dry and wet condition, the index then is used to define the drought as a month with drought index value more than 0.5 (Table 3-2).

Table 3-2 Drought classification based on the drought index.

Xi Classification < -0.5 wet/high events -0.5 ~ 0.5 normal > 0.5 dry/low events (drought)

3.3.2 Calculation of Regional Drought Characteristics

The procedure for calculating regional drought characteristics is based on Hisdal and Tallaksen (2003). Three characteristics of drought namely duration, severity and frequency are the focus of this section. To determine the characteristic of the regional drought, drought area mapping is essential. Drought mapping is done by interpolating the drought index from 15 stations by using Kriging interpolation methods. Regional drought occurs when the total area of drought

area is more than 50% of the island. Drought mapping is done for the monthly event. One drought event is defined as a single event of continuous drought months.

Drought severity is defined by the average drought area, following equation:

t ∑ DA D = 1 (3.2) s t

where Ds is drought severity, DA is drought area for every month during drought event, and t is the duration of the drought event. The frequency of the drought event is simply calculated by the number of drought event for every drought severity class or drought duration.

3.3.3 The Pearson’s Correlation Method

The effect of the ENSO and local SST to discharge drought characteristics are analyzed by using a Pearson Correlation Coefficient (Costa, 2017). We assume there is a linear correlation between discharge and SST. Given equation expresses the degree of linear association

σ ρ = x, y x, y σ ,σ x y (3.3)

where σx,y denotes the covariance between x and y, whereas σx and σy correspond to standard

deviations of x and y respectively. The ρx,y is estimated by the sample correlation coefficient, which is given by

N ∑ (xi − x)(yi − y) i=1 s x, y = N −1 (3.4)

where xi and yi is the concurrent of x and y, and and correspond to their sample, and N is the sample size. Then, the Pearson correlation coefficient𝑥𝑥̅ 𝑦𝑦� can be estimated as

s r = x, y s s x y (3.5)

where

N 2 ∑ (xi − x) s = i=1 (3.6) x N −1

and

N 2 ∑ (yi − y) i=1 s y = N −1 . (3.7)

We analyzed monthly and seasonal time series of both discharge data at the Babak Catchment and the Jangkok Catchment.

3.4 Result

3.4.1 Meteorological and Hydrological Characteristics

3.4.1.1 Rainfall characteristics Analysis of rainfall and meteorological drought in this study was done based on the 3-month running mean. Long-term time series of Lombok rainfall (1983-2015) shows the highest rainfall occurred in 2014 and the lowest one occurred in 1997 (Figure 3-9). The maximum rainfall is around 320 mm in February 2014. Meanwhile, the maximum monthly rainfall in 1998 was only 136 mm, occurred in March. The extreme low events in 1998 occurred in the period of EL Nino event.

350 300 250 200 150

Rainfall (mm) Rainfall 100 50 0

Months

Figure 3-9 Average monthly rainfall in Lombok Island (red strip line indicate (12-months running mean of rainfall).

Average monthly rainfall of Lombok Island shows the rainfall start decreasing as the dry start to propagate in May (Figure 3-10). Schmidt (1957) in his study in defining the rainfall types based on wet and dry period ratios in Indonesia suggested rainfall threshold for this purpose. The dry season occurs when the monthly rainfall is less than 60 mm. Following this definition, the dry season in Lombok starts in May. Considering the objective of the study, which aims to

identify the drought event, running mean of 3 months and six months are being considered in the calculation. It is because of drought is a creeping disaster that is slowly developed and also gradually end. The 3-months and 6-months running mean of rainfall is presented in Figure 3-10.

250

200

150

100 Rainfall (mm) Rainfall 50

0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 1-month 3-month 6-month

Figure 3-10. Average monthly rainfall of Lombok Island.

Considering the ENSO phenomenon, the rainfall characteristic during normal condition is being compared to rainfall during ENSO event (Figure 3-11). The monthly precipitation during El Nino tends to be lower than the normal condition for almost every month. During normal years, dry state (rainfall < 60 mm) started from June and lasted for four months. Meanwhile, during El Nino, the dry state starts earlier in May and lasted for six months (May-Oct). It could be evidence of the effect of El Nino to rainfall condition in Lombok Island, especially during the dry season. El Nino intensifies the dry period from 4 months to be 6 months.

300 Normal La Nina El Nino 250

200

150

100 Rainfall (mm) Rainfall

0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Figure 3-11 Average monthly rainfall during a normal condition and ENSO event.

Lombok Island received rainfall varies spatially, as has been already mentioned in Section 3.2. Spatial characteristics of the monthly rainfall distribution during normal and ENSO years are also presented (Figure 3-12, Figure 3-13, and Figure 3-14). From the spatial distribution of monthly rainfall during three conditions (normal, El Nino and La Nina), it could be identified that eastern part of the island tends to be dry. The dry condition even has been already started from April and lasted in November during the normal years. The situation also worst during El Nino. In the eastern part of the island, the dry condition started to propagate in March and lasted in December.

Figure 3-12 Spatial distribution of monthly rainfall in Lombok Island (normal years).

Figure 3-13 Spatial distribution of monthly rainfall in Lombok Island (El Nino years).

Figure 3-14 Spatial distribution of monthly rainfall in Lombok Island (La Nina years).

To compare the conditions, dry area is being calculated. The regional dry season is set to occur when the coverage of the dry area in the island is more than 50% of the total island. It is apparent that during El Nino, the island experience 100% area covered by the dry condition for four months (Jun-Sep), which is longer than the normal condition which only two months (Jul- Aug) (Figure 3-15). Meanwhile, for La Nina condition, 100% area covered by drought only lasted for one month (Aug). These results give evidence that rainfall in Lombok Island is affected by ENSO phenomenon. To summarize, the average area covered by drought during El Nino is 7% higher than during the normal season (Figure 3-16).

Normal El Nino La Nina 100%

80%

60%

40%

drought (%) 20%

0% Average area covered by covered area Average

Figure 3-15. Monthly dry coverage area during normal and ENSO years.

50% 44% 45% 40% 37% 35% 35% 30% 25% 20% 15% 10%

Average areadry percentage Average 5% 0% El Nino Normal La Nina

Figure 3-16. Average of the dry coverage area.

Annual rainfall characteristics of the island also show the effect of El Nino and La Nina (Figure 3-17). Some part of the island less affected by ENSO, as shown that it is always dry event during La Nina event (eastern part of the island) and also always wet during El Nino event (in the mountainous area). However, it could be identified that during El Nino, it is drier than the normal condition.

Figure 3-17 Spatial distribution of annual rainfall in Lombok Island.

3.4.1.2 Discharge characteristics Two discharge data from two stations are presented in this section (Figure 3-18 and Figure 3-19). The two stations are namely Bug Bug Station in the Jangkok Catchment and Lantandaya Station in the Babak Catchment. Bug Bug Station shows characteristic of rainfall with high fluctuation, which is having high maximum rainfall and low baseflow. This characteristic is different from Lantandaya Station. The discharge in Lantandaya Station tends to have stable flow, and high baseflow compare to Bug Bug Station. The conditions may appear from the different characteristic of the two station’s location. Bug Bug Station is located in the more downstream area whereas Lantandaya Station is located in the more upstream area. It is apparent that the discharge in the Lantandaya Station may be affected by groundwater flow.

Monthly Time Series 30.0 0 25.0 100 20.0 200 300 15.0 400 10.0 500 Rainfall (mm) Rainfall Discharge (m3/s) Discharge 5.0 600 0.0 700

Q_BugBug P_Sesaot

Figure 3-18 Time series of monthly discharge in Bug Bug Station and rainfall in Sesaot Station.

Monthly Time Series 5.0 0 4.5 100 4.0 3.5 200 3.0 300 2.5 2.0 400 1.5 500 (mm) Rainfall

Discharge (m3/s) Discharge 1.0 0.5 600 0.0 700

Q_Lantandaya P_LingkokLime

Figure 3-19 Time series of monthly discharge in Lantandaya Station and rainfall in Lingkok Lime Station.

Furthermore, the characteristics of monthly and seasonal flow in both station are presented in Figure 3-23. From the monthly characteristics, discharge at Bug Bug Station is at the lowest state during August, while at Lantandaya Station the lowest state is during September. The seasonal data shows for Bug Bug Station, the driest condition is during JJA season, while for Lantandaya Station, the driest condition is during SON season.

Bug Bug Station Lantandaya Station 9 2.5 8 7 2 6 1.5 5 4 1 3 Dischareg (m3/s) 2 0.5 1 0 0

Figure 3-20 Monthly average discharge in both stations.

Bug Bug Station Lantandaya Station 10 2.5

8 2.0

6 1.5

4 1.0

Discharge (m3/s) 2 0.5

0 0.0 DJF MAM JJA SON DJF MAM JJA SON Figure 3-21 Seasonal average discharge in both seasons.

3.4.2 Drought Identification

3.4.2.1 Meteorological drought Drought characteristic is being identified following the Equation (3.1). The result of drought identification by using drought index and also dry season definition based on Schmidt (1957) is presented in Figure 3-22 and Figure 3-23. Based on Schmidt (1957) the dry season could be identified successfully, however, the drought condition could not be identified easily. Figure 3-22 shows dry season occurrences every year. The lessen rainfall it gets, the drier it is. Meanwhile, Figure 3-23 shows the occurrences of drought event that is identified by drought index. It could be seen that drought did not occur every year. Notable drought occurred on 1997-1998, 1993-1994 and 2015. All of the drought events happened during El Nino years.

Dry season based on Schmidt (1957) 2.5 0 El Nino Rainfall 2 20 1.5 40 ONI 1

60 (mm) Rainfall 0.5

0 80

Figure 3-22 Dry conditions based on Schmidt (1957).

Drought based on Pi 2.5 El Nino Pi 0 2 0.2 0.4 1.5

0.6 Pi ONI 1 0.8 0.5 1 0 1.2

Figure 3-23 Drought event based on drought index.

The following discussion will focus on the drought 1997-1998 (Figure 3-24 and Figure 3-25). Figure 3-24 shows the dry period in 1997 and 1998 based on Schmidt (1957). The dry condition in 1997 started from July and ended in November despite the fact that there is a strong El Nino happened in that year. Meanwhile, based on the drought index, drought started to occur from May 1997 and terminated in April 1998. El Nino in this period began in May 1997 and lasted in May 1998, known as the strongest El Nino in the last two decades. This El Nino resulted in severe drought in Indonesia. Kirono et al. (1999) documented the El Nino event, in this year and it impacts to the rainfall characteristic in Indonesia. They highlighted that the El Nino has prolonged drought in Nusa Tenggara Province until February 1998. The result of the drought identification by using drought index shows that the drought occurrence lasts longer (April 1998). Despite the overestimate result, the drought identification by using drought index shows a better result than based on Schmidt (1957).

Based on Schmidt (1957) 3 0

2.5 50 2 Rainfall 100 1.5 ONI ONI 150 1

0.5 200 Rainfall < 60 mm 0 250

Figure 3-24 Dry periods 1997-1998 based on Schmidt (1957).

Based on index 3 0 2.5 0.2 2 0.4 Pi ONI 1.5 0.6

Indices 1 0.8 0.5 1 0 1.2

Figure 3-25 Drought periods 1997-1998 based on drought index.

Furthermore, the spatial distribution of the drought will be discussed in this section (Figure 3-26). The drought mapping was done by using kriging ordinary interpolation method. As already been mentioned in the section 3.3.2, regional drought is assumed to occur when the total area that is covered by drought is more than 50%. By following the definition, drought starts to happen in October 1997 to April 1998. The area covered by drought is presented in Figure 3-27. Average area covered by drought for the whole event duration is 89%.

Figure 3-26 Spatial distribution of drought propagation in 1997-1998 drought event.

2.5 120% 2 100% 80% 1.5 60% ONI 1 40% 0.5 20% 0 0% % area affected by drought by % area affected

ONI Drought area percentage

Figure 3-27 Drought area percentage.

3.4.2.2 Hydrological drought The hydrological drought was identified only for two catchments, namely the Babak Catchment and the Jangkok Catchment. The Babak Catchment is represented by Lantandaya Station while the Jangkok Catchment is presented by Bug Bug Station (Figure 3-28 and Figure 3-29). In Lantandaya Station, drought in 1997-1998 could be well captured by the drought index. Other intense El Nino event such as 2002, 2005, 2007 also shows drought occurrences at the same time. However, in the earlier year, 1985-1988 the catchment tend to be wetter even during the El Nino event in 1988. This finding may indicate that El Nino gives stronger impact to the hydrological drought in the recent years. It may be so because of the climate change impact to the catchment. Further study about the catchment characteristics that include vegetation, geological and anthropogenic factor may be conducted here to clarify the impact of the climate change and any other factor to the hydrological drought characteristics in the catchment.

Lantandaya Station 3.0 3.0 2.0 2.0 1.0 1.0

Qi 0.0 0.0 ONI -1.0 -1.0 -2.0 -2.0 -3.0 -3.0

Qi ONI

Figure 3-28 Hydrological drought index in Lantandaya Station.

The result of the hydrological drought index calculation in Bug Bug Station (Figure 3-29) shows that most of the drought occurrence from 1987 to 2015 occurs at the same time of El Nino period. However, there is one drought event that happened during La Nina event, vice versa (refers to the event in 1989 and 1991). This finding needs further study on the other climate phenomenon that affects the hydrological and meteorological condition in Lombok Island.

Bug Bug Station 3.0 3.0 2.0 2.0 1.0 1.0 0.0 0.0 Qi -1.0 -1.0 ONI -2.0 -2.0 -3.0 -3.0 -4.0 -4.0

Qi ONI

Figure 3-29 Hydrological drought index in the Bug Bug Station.

3.4.3 Drought Characteristics

3.4.3.1 Drought duration-frequency Drought characteristics calculated here are drought duration, area affected (severity) and frequency. However, for the hydrological drought, the drought area calculation could not be performed since only one data available for one catchment. Spatial interpolation to make drought area mapping for hydrological drought is difficult to do in this condition. Therefore, for hydrological drought, the severity would be presented by the average of the drought index value. Further discussion about catchment drought characteristic may refer to Section 3.4.3.3.

Drought duration-frequency analysis was performed for three conditions, namely monthly (1-month) drought, 3-month running mean drought, and 6-month running mean drought. The result shows that monthly (1-month) droughts are dominated by drought with duration one month with relative frequency 0.7. The most prolonged drought period that happens for 1- month drought calculation is three months drought duration. For 3-month drought, the longest drought lasted for seven months. Meanwhile, for 6-month drought, the longest drought continued for eight months. Based on the result, the 3-month drought is the best to represent the seasonal drought characteristics.

According to drought duration-frequency analysis result, two drought events is lasted for seven months, namely drought 1997-1998 and drought 1993-1994. Both of the drought occurrences are presented in detail in Figure 3-31. Both droughts occurred during El Nino. However, the characteristic of the drought and El Nino is different for both events. For drought 1997-1998, the drought is intense in a long duration, as a result of the strong El Nino that occur at the same

52 time. Meanwhile, for 1993-1994 drought, the drought occurred slowly, with not so high magnitude but lasted in the same duration as 1997-1998 drought. It may be because of the atmospheric condition, an especially sea surface temperature which indicates by ONI that always higher than normal for an extended period.

Pi(1-month) > 0.5 Pi(3-month) > 0.5 Pi(6-month) > 0.5 0.8 0.50 0.50 0.7 0.40 0.6 0.40 0.5 0.30 0.30 0.4 0.3 0.20 0.20 0.2 0.10 0.10 Relative frequency 0.1 0 0.00 0.00 1 2 3 4 5 6 1 2 3 4 5 6 7 1 2 3 4 5 6 7 8 Duration (months) Duration (months) Duration (months)

Figure 3-30 Meteorological drought duration-frequency.

2 ONI Pi 1

Indices 0

-1

3 ONI Pi 2 1 0 Indices -1 -2

Figure 3-31 Highlight of 7 months drought duration (upper: 1993-1994 drought, lower part: 1997- 1998 drought).

3.4.3.2 Drought severity-frequency Drought severity is characterized by the average area covered by drought for a single drought event. Based on the drought area calculation, the most frequent drought is a drought with coverage area 80-90% to the island (Figure 3-32).

0.4 0.35 0.3 0.25 0.2 0.15 0.1

Relative frequency 0.05 0

Average area covered by drought

Figure 3-32 Meteorological drought severity-frequency analysis.

3.4.3.3 Catchment drought characteristic The calculation of drought in the catchment area is based on the rainfall and discharge condition in the catchment. As have been mentioned in Figure 3-3, every catchment has one rainfall station and one discharge station. The calculation of meteorological and hydrological drought then is performed based on the available data.

The result of the drought characteristics calculation in the Babak Catchment shows that the longest drought lasted for 12 months for meteorological drought and 21 months for hydrological drought (Figure 3-33). The same characteristic also for the Jangkok Catchment (Figure 3-34). The result is different from the island drought characteristics, which only lasted for seven months. The different result may happen because of local characteristics of the hydro- meteorological condition is more fluctuating than the regional condition.

Meteorological Drought Hydrological Drought 25% 30%

20% 25% 20% 15% 15% 10% 10% 5% Relative Frequency Relative Frequency 5% 0% 0% 1 2 3 4 5 6 7 8 9 101112 1 3 5 7 9 111315171921 Duration (months) Duration (months)

Figure 3-33 Catchment drought duration-frequency in the Babak Catchment.

Meteorological Drought Hydrological Drought 0.5 0.35 0.3 0.4 0.25 0.3 0.2 0.2 0.15 0.1 0.1 Relative Frequency Relative Frequency 0.05 0 0 1 3 5 7 9 11 13 15 17 1 3 5 7 9 11 13 15 17 Duration (months) Duration (months)

Figure 3-34 Catchment drought duration-frequency in the Jangkok Catchment.

3.4.4 Intermediate Correlation

It is assumed that meteorological drought is correlated with El Nino condition. Then, by assuming that the meteorological drought also relates to hydrological drought, the hydrological drought is then assumed affected by El Nino conditions. Since the hydrological drought condition is needed in this analysis; the calculation will be performed in the catchment scale. The meteorological drought correlation with El Nino will also be presented to be compared with the regional drought condition.

The results of the correlation among meteorological drought, hydrological drought and ENSO are presented in Figure 3-35. The two catchment has different characteristic. The Babak Catchment, the relationship between the three component is happening in Jan-Feb-Mar (JFM), which is the wet season, although it is not significant. Meanwhile, in the Jangkok Catchment, the high correlation between the three component is happening in the dry season between Jun- Jul-Aug (JJA), Jul-Aug-Sep (JAS) and Sep-Oct-Nov (SON). The significant relationship of discharge with ENSO during SON season in Bug Bug Station shows a good agreement with a previous study about the relationship between rainfall with ENSO in Indonesia (Aldrian & Dwi Susanto, 2003; Hamada et al., 2002; Haylock & McBride, 2001; Hendon, 2003; Sahu, Behera, Yamashiki, Takara, & Yamagata, 2012).

1.000 0.800 0.600

R 0.400 0.200 0.000 -0.200 NDJ DJF JFM FMA MAM AMJ MJJ JJA JAS ASO SON OND

Pi - Qi Pi - ONI Qi - ONI 1.000

0.500 R 0.000

-0.500 NDJ DJF JFM FMA MAM AMJ MJJ JJA JAS ASO SON OND

Figure 3-35 Correlation of droughts and ENSO (upper part in the Babak Catchment, lower part in the Jangkok Catchment).

The regional meteorological drought has a significant correlation with ENSO during five months (JJA, JAS, ASO, SON, and OND) (Figure 3-36).

0.70 0.63 0.56 0.60 0.47 0.50 0.44 0.36 0.40 0.35 0.32 0.35 0.29 0.30 0.23 0.25 0.25 0.20 0.10 R (correlation coeeficient) 0.00 NDJ DJF JFM FMA MAM AMJ MJJ JJA JAS ASO SON OND 3 months average

Figure 3-36 Regional meteorological drought correlation with ENSO (R is performed for 3-month Pi and ONI) (red color shows significant value at P = 95% for a two-tail test).

3.4.5 Lag Correlation

The lag correlation of ENSO with drought indices was done only for Bug Bug Station. Lantandaya Station was neglected for the analysis since the discharge in this station does not show any significant correlation with ENSO. From the lag correlation of discharge on SON season is significantly correlated with JJA ENSO index Table 3-3. This result shows the possibility to use ENSO index for hydrological drought prediction in the Bug Bug Station. Short-term drought prediction for SON month by using ONI 3 months before could be performed in Bug Bug Station.

Table 3-3 Lag Correlation between ENSO and discharge in Bug Bug Station ONI - Qi lag correlation (R) Time lag of ONI Rainfall Season (months before) DJF MAM JJA SON 9 -.109 .432 * .012 .561 * 6 -.200 .443 * .281 -.207 3 -.265 .055 -.097 .541 * 0 .524 * .597 * .305 .633 * *significant at 95%

3.5 Summary

In this chapter, meteorological and hydrological drought identification was performed by using drought index. The drought index could be used for meteorological drought identification in Lombok Island and resulted in a good performance. The result of the meteorological drought analysis shows a good agreement with the previous study by (Kirono et al., 1999) regarding drought duration. The 3-month running means drought analysis was selected for meteorological drought analysis considering the physical characteristic of drought and also the seasonal feature of the climate. The most frequent drought happened is the 1-month duration drought, while the most extended duration of drought is seven months.

Hydrological drought analysis was done for two catchment area, namely the Babak Catchment and the Jangkok Catchment. The characteristic of drought in both catchments shows longer drought duration than the regional one. The different result may happen because of local characteristics of the hydro-meteorological condition is more fluctuating than the regional condition.

Intermediate correlation between meteorological drought, hydrological drought, and ENSO shows a different pattern between two catchments. In the Babak Catchment, the non-significant relationship between the three components happens in Jan-Feb-Mar (JFM), which is a wet season. Meanwhile, in the Jangkok Catchment, the high correlation between the three component is happening in the dry season between Jun-Jul-Aug (JJA), Jul-Aug-Sep (JAS) and Sep-Oct-Nov (SON). Lag correlation between hydrological drought and ENSO is there for hydrological drought SON season with ENSO in JJA season.

3.6 References

Fitri, S., & Hermawan, B. (2015, 30 July 2015). Ini Wilayah-Wilayah yang Paling Terdampak El Nino. Republika. Retrieved from https://goo.gl/D6dURx Hamada, J.-I., Yamanaka, M. D., Matsumoto, J., Fukao, S., Winarso, P. A., & Sribimawati, T. (2002). Spatial and temporal variations of the rainy season over Indonesia and their link to ENSO. 気象集誌. 第 2 輯, 80(2), 285-310. Haylock, M., & McBride, J. (2001). Spatial coherence and predictability of Indonesian wet season rainfall. Journal of Climate, 14(18), 3882-3887. Hendon, H. H. (2003). Indonesian rainfall variability: Impacts of ENSO and local air-sea interaction. Journal of Climate, 16(11), 1775-1790. Hisdal, H., & Tallaksen, L. M. (2003). Estimation of regional meteorological and hydrological drought characteristics: a case study for Denmark. Journal of Hydrology, 281(3), 230- 247. Kirono, D. G., Butler, J. R., McGregor, J. L., Ripaldi, A., Katzfey, J., & Nguyen, K. C. (2015). Historical and future seasonal rainfall variability in Nusa Tenggara Barat Province, Indonesia: implications for the agriculture and water sectors. Climate Risk Management. Kirono, D. G., Tapper, N. J., & McBride, J. L. (1999). Documenting Indonesian rainfall in the 1997/1998 El Nino event. Physical Geography, 20(5), 422-435. Mishra, A. K., & Singh, V. P. (2010). A review of drought concepts. Journal of Hydrology, 391(1), 202-216. Reynolds, R. W., Rayner, N. A., Smith, T. M., Stokes, D. C., & Wang, W. (2002). An improved in situ and satellite SST analysis for climate. Journal of Climate, 15(13), 1609-1625. Ridwan, M. F., & Hazliansyah. (2015, 22 August 2015). 876 Ha Lahan Tanaman Pangan di NTB Gagal Panen. Republika. Retrieved from https://goo.gl/bVHt5i Sahu, N., Behera, S. K., Yamashiki, Y., Takara, K., & Yamagata, T. (2012). IOD and ENSO impacts on the extreme stream-flows of Citarum river in Indonesia. Climate Dynamics, 39(7-8), 1673-1680. Schmidt, F. (1957). Rain fall types based on wet and dry period ratios for Indonesia with New Guinea. Verhandelingen, 42. Trenberth, K. E. (1997). The definition of el nino. Bulletin of the American Meteorological Society, 78(12), 2771-2777. Whitburn, S., Van Damme, M., Clarisse, L., Turquety, S., Clerbaux, C., & Coheur, P. F. (2016). Doubling of annual ammonia emissions from the peat fires in Indonesia during the 2015 El Niño. Geophysical Research Letters, 43(20).

4 Impacts of ENSO and Local Sea Surface Temperature on Rainfall Patterns in the Batanghari River Basin, Sumatra, Indonesia

4.1 Introduction

Climate change impacts on the increase of sea surface temperature (IPCC, 2014). The effect of sea surface temperature is also related to El Nino/Southern Oscillation (ENSO) phenomena. The ENSO is an anomaly of the sea surface temperature in the Pacific Ocean which is often related to droughts in some parts of the world and severe floods in other parts (Harrison & Larkin, 1998). Regions that are located in the Pacific Ocean, precipitation patterns could be directly related to the ENSO (Ropelewski & Halpert, 1987).

Being located in near of the Pacific Ocean, Indonesian climate is affected by ENSO. Sumatra Island is one of the islands that the rainfall condition is profoundly influenced by ENSO, especially in the southern part of the island (As‐syakur et al., 2014). In the Batanghari River Basin located on the Sumatra Island, the rainfall condition may be profoundly influenced by the climate phenomena. Besides, the other factor of climate like local Sea Surface Temperature (SST) may have also influenced the regional climate conditions.

As the most priority livelihood is farming, the water resources management in this area is very crucial. One of the crucial aspects to have a proper water resources management is to have a good data and reliable rainfall predictions. Besides, the rainfall prediction is also necessary for disaster management, particularly for water-related disasters including flood and drought. As the climate in the river basin may be affected by the ENSO and local SST, the utilization of these two indicators for rainfall predictions may be possible.

This study investigates general characteristics of rainfall patterns in the region as well as predictability of rainfall by using ENSO and local SST indicators. The study will calculate Pearson’s Correlation Coefficient between rainfall and SST and estimate the trend of historical rainfall by Mann Kendall test. The study also compares the gauged rainfall data in the basin with the global dataset from Global Precipitation Climatology Center (GPCC) to see the possibility of using the global data for long-term analysis.

4.2 Study Site and Data

Our study area is the Batanghari River Basin which is located on Sumatra Island. The Batanghari River Basin is the second largest river basin in Indonesia with the catchment area

around 45,379 km2. 76% of the area is located in Jambi Province, while the rest belongs to West Sumatra Province (19%), South Sumatra Province (4%) and Riau Province (1%). The upstream of the Batanghari River Basin is the Bukit Barisan mountainous area while the downstream area is at Jambi City. The Batanghari River Basin has a wet climate with the annual rainfall 2,500-3,000 mm. However, in the downstream area, e.g., Jambi City, the annual rainfall is less around 2,200 mm. The agricultural sector is the primary livelihood for people living in the Batanghari River Basin (Saputra, 2015).

Figure 4-1 Study area of the Batanghari River Basin in Jambi Province, Sumatra Island, Indonesia.

We used daily rainfall data in Sultan Thaha Station which is located at 1.633°S and 103.650°E Figure 4-1. The data is managed by the Indonesian Agency for Meteorology, Climatology, and Geophysics (BMKG). The analyses are conducted to the data from 1985 until 2012 (28 years). This station is chosen since the data availability of this station was the most without a significant missing period in the 28 years. We aggregated the daily data into monthly and seasonal for the analysis. We define the season into the following four seasons, namely Dec- Jan-Feb (DJF) as wet season, Mar-Apr-May (MAM) as the transition wet to dry season, Jun- Jul-Aug (JJA) as dry season, and Sep-Oct-Nov (SON) as the transition from dry to wet season based on previous studies (Hendon, 2003).

The GPCC Full Data Reanalysis version 7, which is a monthly data with the resolution of 0.7°x0.7°, is also used from January 1901 to December 2013 (Schneider et al., 2015). We calculate the Batanghari River Basin rainfall by averaging the GPCC data located in a range between 1-2°S and 101-104°E (box-shaped shown in Fig.1). The global data was compared with the observed data from January 1985 to December 2012 to see the possibility of using global data for local analysis. We compare the monthly time series of rainfall data from Sultan Thaha station and GPCC by using the Pearson’s Correlation. We found out the two data have similar monthly characteristics with the value of correlation coefficient of r equals to 0.79 (correlation is significant at the 0.01 level (2-tailed)) Figure 4-2.

The other primary dataset we used is daily SST data from NOAA Optimum Interpolation (OI) SST version 2 (Reynolds, Rayner, Smith, Stokes, & Wang, 2002) from 1901 to 2013. We calculate the Nino3.4 index by taking the average of the SST over the region of 5°N-5°S and 170-120°W to represent the ENSO event. In addition to these, we used the South China Sea to represent the local SST that affect rainfall in the Batanghari River Basin (Aldrian & Dwi Susanto, 2003). The local SST was derived from the average of the OISST data located between 8°N-1.5°S and 102-110°E.

600 a) GPCC 500 Sultan Thaha 400

300

200 Rainfall (mm)

100

0 1-Jan-85 1-Jan-90 1-Jan-95 1-Jan-00 1-Jan-05 1-Jan-10

600 b) 500

400 y = 0.9528x - 9.411 R² = 0.6312 300

Sultan Thaha Sultan 200

100

0 0 100 200 300 400 500 GPCC

Figure 4-2 Comparison of the Batanghari River Basin rainfall from Sultan Thaha Station and GPCC data: a) time series and b) correlation.

4.3 Method

4.3.1 Trend Analysis

We used Mann-Kendall test to analyze the rainfall trend in the Batanghari River Basin. The test is applicable in cases when the data values xi of a time series can be assumed to follow the model:

= ( ) + ε xi f ti i (4.1) where f(ti) is a continuous monotonic increasing or decreasing function of time and the residuals εi can be assumed to be from the same distribution with zero mean. It tests the null hypothesis of no trend, Ho, against the alternative hypothesis, H1, where there is an increasing or decreasing monotonic trend.

The Mann-Kendall test statistic S is calculated using the formula,

n−1 n S = ∑ ∑ sg(x j − xk ) k =1 j =k +1 (4.2)

The two-tailed test is used for four different significance levels α: 0.1, 0.05, 0.01 and 0.001. If the absolute value of S equals or exceeds a specified value Sα/2 we reject the Ho which means significantly decreasing or increasing trend in the rainfall (Salmi, 2002).

4.3.2 The Pearson’s Correlation Method

The calculation of the Pearson’s Correlation following the method in the Section 3.3.3.

4.4 Result and Discussion

4.4.1 Rainfall Trends

The monthly rainfall in the Batanghari River Basin has two peaks, in Mar-Apr and Nov-Dec (Figure 4-3). Based on this seasonal characteristic, we classified this region into Region B of Indonesian climate region based on Aldrian and Dwi Susanto (2003). The cause of those two peaks is the southward and northward movement of the inter-tropical convergence zone (ITCZ) (Aldrian & Dwi Susanto, 2003). Meanwhile, the seasonal rainfall shows that the lowest rainfall is in the JJA season.

300

250

200

150

100 Rainfall (mm) Rainfall 50 GPCC Sultan Thaha 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

1000

800

600

400

Rainfall (mm) Rainfall 200 GPCC Sultan Thaha 0 DJF MAM JJA SON Figure 4-3 The Batanghari River Basin monthly and seasonal rainfall characteristic.

The annual rainfall trend based on the Mann Kendall test is shown in Table 4-1. From the result, there is no significant increasing and decreasing trend of annual rainfall in the river basin. The trend analysis of monthly and seasonal rainfall was also done in the both GPCC and Sultan Thaha data. The results show there is no significant change for GPCC data for every month and season while Sultan Thaha data shows decreasing trend only in MAM with 0.05 significant level.

Table 4-1 Mann Kendall Trend Analysis

Test statistic S Time series GPCC Sultan Thaha Jan -1.21 -0.69 Feb 0.85 0.38 Mar -1.64 -1.60 Apr 0.41 -1.32 May -0.49 -0.69 Jun 0.18 1.13 Jul -0.30 1.32 Aug 0.34 0.97 Sep -0.22 -0.41 Oct 0.18 -0.53 Nov 0.61 0.61 Dec -0.06 -0.65 DJF -0.18 -0.71 MAM -0.89 -2.11 * JJA -0.22 1.56 SON 0.69 -0.10 Annual -0.61 -0.18 *Significant at < 0.05

4.4.2 Intra-seasonal Correlation of Rainfall with ENSO and Local SST

We investigated the monthly and seasonal patterns of the rainfall and SST. However, there is no clear pattern of the monthly precipitation, local SST and Nino3.4 relationship (Figure 4-4). The rainfall occurred from the 26.5 to 31°C of local SST, and lower temperature of Nino3.4 between 24-30°C with the monthly rainfall amounts range from 0 to 507 mm. The largest fluctuation of the sea temperature of Local SST is in March for about 2.17°C and in November for about 5.05°C of Nino3.4. The seasonal pattern shows the DJF rainfall occurred in the lowest range of sea surface temperature between 26.5 to 28.6°C of local SST. Meanwhile, the other season's rainfall takes place in almost the same range between 28.5 to 31°C of local SST. The JJA season has the lowest amount of rainfall ranging between 95.9 to 902.6 mm. Meanwhile, The Nino3.4 rainfall relationship pattern is less clear compared with the local SST. The largest rainfall amount happens for two seasons SON, and MAM with the rainfall amounts range from 272 to 1,018 mm and 125.5 to 1,202 mm respectively.

Monthly and seasonal relationships between rainfall in the Batanghari and SST are explored by correlating monthly and seasonal rainfall with SST values. Figure 4-5 shows the correlation coefficient (r) of the monthly and seasonal rainfall with Nino3.4 and local SST. It is apparent that

there is a strong seasonality of rainfall relationship with ENSO and local SST. The strongest relationship is seen in the dry season (JJA) for both ENSO and local SST. The monthly correlation shows the highest correlation between local SST and rainfall happens in September with a positive correlation (r = 0.61) for Sultan Thaha data and r = 0.47 for GPCC data. The other months that have a significant correlation with local SST are July and August with r = 0.37 and 0.45 respectively for Sultan Thaha and r = 0.47 and 0.59 respectively for GPCC data. Except for September, the GPCC rainfall and local SST have a higher correlation than Sultan Thaha data. The others month have no significant correlation with local SST. Thus, the relationship of rainfall and local SST is significant during Jul-Aug-Sep.

Meanwhile, the significant correlation of rainfall and ENSO happens for four months; Jul-Aug- Sep-Oct. The highest correlation is in October with r = -0.603 for Sultan Thaha rainfall and - 0.583 for GPCC data. For those all four months, the GPCC data has higher correlation than Sultan Thaha data. Nevertheless, the pattern of correlation for GPCC and Sultan Thaha data is similar.

Seasonal correlation of rainfall and SST shows a different pattern between local SST and Nino3.4. The high correlation of rainfall and local SST only seen during JJA season (r = 0.55), while for Nino3.4-rainfall, strong correlation observed during two seasons; JJA and SON with r -0.634 and -0.626 respectively.

600 Jan 500 Feb Mar 400 Apr May 300 Jun Jul 200 Rainfall (mm) Rainfall Aug 100 Sep Oct 0 Nov 24 25 26 27 28 29 30 Dec Nino3.4 (°C)

600 Jan 500 Feb Mar 400 Apr May 300 Jun 200 Jul

Rainfall (mm) Rainfall Aug 100 Sep Oct 0 Nov 26 27 28 29 30 31 32 Dec Local SST (°C)

1400 1200 DJF 1000 MAM 800 JJA 600 SON Rainfall (mm) Rainfall 400 200 0 24 25 26 27 28 29 30 Nino3.4 (°C)

1400 1200 DJF 1000 MAM 800 JJA 600 SON Rainfall (mm) Rainfall 400 200 0 26 27 28 29 30 31 Local SST (°C) Figure 4-4 Inter-annual rainfall and SST correlation pattern.

1.00 Monthly correlation with Nino3.4 0.80 GPCC 0.60 0.40 Sultan Thaha 0.20

r 0.00 -0.20 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec -0.40 -0.60 -0.80 -1.00 Monthly correlation with local SST 1.00 0.80 GPCC 0.60 Sultan Thaha 0.40 0.20

r 0.00 -0.20 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec -0.40 -0.60 -0.80 -1.00

1.00 Seasonal correlation with Nino3.4 0.80 GPCC 0.60 Sultan Thaha 0.40 0.20

r 0.00 -0.20DJF MAM JJA SON -0.40 -0.60 -0.80 -1.00

1.00 Seasonal correlation with local SST 0.80 0.60 GPCC 0.40 Sultan Thaha 0.20

r 0.00 -0.20DJF MAM JJA SON -0.40 -0.60 -0.80 -1.00

Figure 4-5 Coefficient correlation (r) between rainfall with Nino3.4 and local SST.

The spatial analysis of the correlation between rainfall and SST was also done and shown in Figure 4-6. During the dry season of JJA, there is a negative correlation of rainfall with SST in the Nino3.4 region and positive correlation with the local SST in the South China Sea and Indonesia Sea (Figure 4-6). Based on this correlation, it is likely when the temperature in the

Pacific Ocean increases and the temperature in the South China Sea decreases, the amount of rainfall in the Batanghari River Basin decreases.

The correlation of ENSO is also seen in the transition season of SON, where the rainfall has a relationship with ENSO and Indonesian SST. However, the rainfall has a weak correlation with local SST in the South China Sea around 0.1 (Figure 4-6). Meanwhile, the JJA season has a high relationship with the sea temperature in the South China Sea around 0.5. It may indicate that the rainfall in JJA seasons is affected by both ENSO and local SST while rainfall in SON season is affected by only ENSO. Contrast with the dry season, the wet season in DJF and MAM shows no correlation either with ENSO, Maritime continent, and local SST.

Figure 4-6 Correlation between rainfall and sea surface temperature (DJF, MAM, JJA, and SON seasons respectively from above to below). A correlation of 0.374 is significantly different from zero at the 95% confidence level, assuming the 26 degrees of freedom.

4.4.3 Lag Correlation

The relationship of each season rainfall with previous season local SST and Nino3.4 are investigated to see the possibility of using both indicators for rainfall prediction. The lag correlation of JJA season rainfall is high with local SST in the MAM season as shown in Figure 72

4-7 (significant three months lag correlation is shown with x-marker). The JJA rainfall shows lag-correlation to local SST at three months before with r = 0.45, which is statistically significant (at < 0.01). Meanwhile, for Nino3.4 lag correlation, the highest one happens for SON rainfall with Nino3.4 JJA for r = -0.642 (at < 0.01). From this result, it shows the possibility of using local SST indicator for JJA rainfall prediction and Nino3.4 (ENSO) indicator for SON rainfall prediction.

1.000 DJF-lag MAM-lag .500

r .000 -.500 6 3 0 6 3 0 -1.000 1.000 JJA-lag SON-lag .500

r .000 -.500 6 3 0 6 3 0 -1.000 time lag (months before rainfall) Local SST Nino3.4 Figure 4-7 Lag correlation of rainfall with local SST & Nino3.4.

4.5 Summary

This study investigated the trend of historical rainfall in the Batanghari River Basin. The results of Mann-Kendall trend analysis, in general, show no significant trend of any changes either for monthly, seasonal nor annual rainfall in Sultan Thaha station and GPCC data. The only trend seen is in MAM season for Sultan Thaha data with the significant decreasing trend (at < 0.05). As MAM season is a transition from wet to dry season the decreasing trend of rainfall in this season may indicate the earlier developing of the dry season.

The impact of local SST and Nino3.4 (represent ENSO) to the rainfall shows a significant positive influence of local SST to rainfall during Jul-Aug-Sep and negative correlation of Nino3.4 with rainfall during Jul-Aug-Sep-Oct. For the seasonal correlation, JJA season shows significant correlation with both local SST and Nino3.4 while SON season correlates only with Nino3.4. The significant negative correlation of Nino3.4 with both JJA and SON seasons may indicate a strong and prolong impacts of ENSO to SON season in the river basin.

Furthermore, we also examined the lag correlation of rainfall with both climate factors to see the predictability of rainfall by using both indicators. A significant correlation is seen in JJA rainfall with MAM local SST. On the other hand, high lag correlation is also seen for SON rainfall with JJA Nino3.4. It can be a good signal for using those two indicators in predicting the rainfall in the dry season (JJA and SON). Further study about drought forecasting by using those indicators may be done for drought and water management purposes.

4.6 References

Aldrian, E., & Dwi Susanto, R. (2003). Identification of three dominant rainfall regions within Indonesia and their relationship to sea surface temperature. International Journal of Climatology, 23(12), 1435-1452. doi:10.1002/joc.950 As‐syakur, A., Adnyana, I., Mahendra, M. S., Arthana, I. W., Merit, I. N., Kasa, I. W., . . . Sunarta, I. N. (2014). Observation of spatial patterns on the rainfall response to ENSO and IOD over Indonesia using TRMM Multisatellite Precipitation Analysis (TMPA). International Journal of Climatology, 34(15), 3825-3839. Harrison, D., & Larkin, N. K. (1998). El Niño‐Southern Oscillation sea surface temperature and wind anomalies, 1946–1993. Reviews of Geophysics, 36(3), 353-399. Hendon, H. H. (2003). Indonesian rainfall variability: Impacts of ENSO and local air-sea interaction. Journal of Climate, 16(11), 1775-1790. IPCC. (2014). Climate change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. IPCC, Geneva, Switzerland, 11, 151 pp. Reynolds, R. W., Rayner, N. A., Smith, T. M., Stokes, D. C., & Wang, W. (2002). An improved in situ and satellite SST analysis for climate. Journal of Climate, 15(13), 1609-1625. Ropelewski, C. F., & Halpert, M. S. (1987). Global and regional scale precipitation patterns associated with the El Niño/Southern Oscillation. Monthly Weather Review, 115(8), 1606-1626. Salmi, T. (2002). Detecting trends of annual values of atmospheric pollutants by the Mann- Kendall test and Sen's slope estimates-the Excel template application MAKESENS: Ilmatieteen laitos. Saputra, F. M. (2015). Daerah Aliran Sungai Batanghari (\). Schneider, U., Becker, A., Finger, P., Meyer-Christoffer, A., Rudolf, B., & Ziese, M. (2015). GPCC Full Data Reanalysis Version 7.0 at 0.5°: Monthly Land-Surface Precipitation from Rain-Gauges built on GTS-based and Historic Data. doi:10.5676/DWD_GPCC/FD_M_V7_050

5 Low Flow Forecasting with Recession Analysis Approaches 5.1 Introduction

Drought is a creeping disaster which is difficult to be identified because of less evidence of physical effect. The effect of the disaster could be seen after some time since the disaster happens. However, once the disaster is apparent, it is already late to take mitigation action. Therefore, the drought forecasting is essential for drought mitigation and preparedness to minimize the impact of the disasters (A. K. Mishra & Singh, 2011).

There are four types of drought, namely meteorological drought, hydrological drought, agricultural drought, and socioeconomic drought. Among those, this study focuses on hydrological drought, and particularly this study concern low streamflow conditions in dry seasons. Many of the drought impacts, such as ecosystems and social impacts are associated with the hydrological drought (Van Loon, 2015).

In terms of the forecasting, there are number of methods proposed for meteorological drought in recent years, e.g. drought forecasting by using drought index (Cancelliere, Di Mauro, Bonaccorso, & Rossi, 2007; Özger, Mishra, & Singh, 2012), stochastic models (Durdu, 2010; A. Mishra & Desai, 2005; A. Mishra, Desai, & Singh, 2007) and remote sensing (Han, Wang, Zhang, & Zhu, 2010). For hydrological drought, there are many approaches proposed, including the probabilistic method (Araghinejad, 2011; Kikkawa & Takeuchi, 1975; Madadgar & Moradkhani, 2013) and hydrological drought index (Hatmoko, Raharja, Tollenaar, & Vernimmen, 2015). Kikkawa and Takeuchi developed one of the hydrological statistic using a probabilistic method called a drought duration curve (DDC) which estimate a sum of inflows over m days in a T-year drought (Kikkawa & Takeuchi, 1975).

One of the common approaches for forecasting low flow with some lead time is to use hydrologic models. They are mainly important because of the predicted discharge information used in water allocation under drought conditions. However, the application of hydrologic models in low flow forecasting is limited when we have no sufficient rainfall records. For example, in our case study in a mountainous catchment in Lombok Island in Indonesia, some streamflow records are available as they are primarily crucial for water allocations, we have no sufficient rainfall records as the catchment covers high mountains. In such a case, it may be more feasible to forecast low flow based on recession characteristics of stream flow. Since the low flow discharge is determined by catchment storage conditions and less affected by rainfall,

75 the recession approach for the distinct dry season is particularly suitable. Hence the objective of this study is to investigate the applications of two different recession analysis methods for low flow forecasting in the Babak River Basin in Indonesia. The two methods are recursive digital filters and simple dynamical systems model. The recursive digital filters model was chosen as it is simple, with a single parameter for baseflow separation, while the simple dynamical systems were selected since it can represent more flexible recession characteristics.

5.2 Study Site and Data

5.2.1 Study site

Lombok Island, Indonesia is one of the islands that are prone to drought with the high variability of annual rainfall from 800 – 1970 mm/year. In 2015, the number of villages affected by the drought accounted for around 31% of the total villages in the Island. The Babak River Basin is selected for the study as this area has very high utility for supporting irrigation. The water allocation system in the study area is currently using rainfall forecast. The discharge forecasting may contribute to improve the accuracy of the water allocation. The area of the Babak River Basin is 258.1 km2. The Babak River Basin area is shown in Figure 5-1.

Figure 5-1 The Babak River Basin.

5.2.2 Data

River discharge data from two automatic water level recorder (AWLR) station is used for the analysis. The two stations are Lantandaya and Gebong Station. The Lantandaya Station is located in the upstream part of the river basin, while Gebong Station is located in the downstream area, as shown in Figure 5-1. 31 years data recorded from 1985 until 2015 is used for the analysis.

5.3 Recursive Digital Filters

Recursive digital filters for baseflow separation was used in the analysis. The basic concept of the method is to separate the streamflow into direct runoff and baseflow following the Eq. (5.1).

yk = fk + bk (5.1) with yk is total streamflow, fk is a direct runoff, bk is baseflow, and k is time step number. Two parameters are used to calculate the baseflow, namely recession constant a and BFImax. The

recession constant a is objectively determined by a recession analysis while BFImax is the maximum value of the baseflow index (long-term ratio of baseflow to total streamflow) (Eckhardt, 2005).

The general formulation of baseflow is

bk = Abk −1 + Byk . (5.2)

The parameter of A and B could be defined by a and BFImax parameters, which are

1− BFI A = max a 1− aBFI max (5.3) and

(1− a)BFI B = max . (5.4) 1− aBFImax

By substituting the parameters of A and B in Eq.(5.3) and Eq.(5.4) to Eq.(5.2), the formula of a recursive one-parameter filter could be obtained as follow (Eckhardt, 2005):

(1− BFI )ab + (1− a)BFI y b = max k −1 max k k 1− aBFI max (5.5)

subject to bk ≤ yk .

The recession constant a is determined by the recession analysis of the streamflow discharge

(Kirchner, 2009). Every streamflow value of yk is taken into account, which makes sure a sufficient recession period at least for five days, represented by the following condition.

yk −3 > yk −2 > yk −1 > yk > yk +1 > yk +2 . (5.6)

The value of a is defined as an envelope of a plot of yk+1 against yk. The accuracy of a may depend on the available data. The short record may result in the underestimation of a as the parameter is obtained from the envelope of the data.

The value of BFImax is determined from the hydro-geological characteristic of a river basin.

Eckhardt (2005) suggested some value of BFImax, e.g., BFImax ≈ 0.80 for perennial streams with

porous aquifers, BFImax ≈ 0.50 for ephemeral streams with porous aquifers, BFImax ≈ 0.25 for

perennial streams with hard rock aquifers. We used BFImax ≈ 0.80 since the study area has a characteristic of perennial streams with the porous aquifer.

By following the assumption that the outflow from the aquifer is linearly proportional to its storage, that is resulting the model of an exponential baseflow recession during periods without groundwater recharge (Eckhardt, 2008), the baseflow forecasting could be done by following the equation

bk +1 = abk (5.7) which is equivalent to

bk +1 = bk exp(− ∆t τ ). (5.8)

The equation (8) can also be expressed as

bk = b0 exp(− k∆t τ ) (5.9) with a = exp(-∆t/τ) and τ is a constant parameter.

The low flow forecasting in this study follows the Eq.(5.7), where the input for the forecasting is recession constant (a) and the discharge of the river in the starting time of the forecasting

(b0).

5.4 Catchment as Simple Dynamical Systems

The theory of “Catchments as simple dynamical systems”(Kirchner, 2009) leads to a quantitative estimation of catchment’s dynamic storage, recession time scales, and sensitivity to antecedent moisture, which is suggested to be useful for catchment characterization. This theory proposes a first-order nonlinear differential equation that can be used to simulate streamflow hydrograph from precipitation and evapotranspiration time series.

The basic concept of this theory is, to begin with, the conversion-of-mass equation,

dS = P − E − Q (5.10) dt where S is the volume of water stored in the catchment (mm), P is precipitation (mm), E is evapotranspiration (mm), Q is discharge (mm/day). The Eq.(5.10) illustrated the changing of the storage volume on the change in the time t while it is a function of P, E, and Q.

From the Eq.(5.10) the precipitation, evapotranspiration and storage volume are measured while the discharge is calculated from the other three components. Since precipitation and evapotranspiration are locally characterized, discharge is strongly related to the storage capacity. The condition can be expressed by the equation,

Q = f (S) . (5.11)

Eq.(5.12) is then used to estimate the catchment sensitivity to changes in storage. The equation formed by substituting Eq.(5.11) to Eq.(5.10).

dQ dQ dS dQ = = (P − E − Q) dt dS dt dS . (5.12)

The changing of the discharge on the storage change is defined as the function of the discharge itself,

dQ = f '(S) = f '(f −1 (Q)) = g(Q) dS . (5.13)

The g(Q) function is called the “sensitivity function” as it expresses the sensitive nature of the discharge to changes in storage. By assuming there is no precipitation and evapotranspiration in the period of the analysis, the g(Q) function then is determined by,

dQ − dQ dt g(Q) = ≈ dS Q P≤Q,E≤Q . (5.14)

The g(Q) function is solved by plotting ln(–dQ/dt) and ln(Q) and making quadratic polynomial regression from the plot. The quadratic regression function of ln(g(Q)) will be Eq.(5.15). After this, the g(Q) function could be determined from Eq.(5.15) as Eq.(5.16).

 − dQ dt  ln(g(Q)) = ln  ≈ c + c ln(Q) + c (ln(Q))2  Q  1 2 3  P≤Q,E ≤Q  . (5.15)

( ) = + + ( )2 g Q exp(c1 c2 lnQ c3 lnQ ). (5.16)

After we obtain the function of g(Q), the function of dQ/dt could be defined as,

dQ 2 = exp(c1 + (c2 −1)ln Q + c3 (ln Q) )(− Q) dt . (5.17)

As the parameters of c1, c2, and c3 were already obtained from the quadratic regression plot, we only need to have Q as the input for the Eq. (5.17). Finally, the flow forecasting is obtained from dQ/dt function.

5.5 Flow Characteristic

The flow characteristics of the Babak River Basin show high seasonal variability; high stream flow in a rainy season, starting from November until April and start to decrease in the dry season, starting from May until September, as shown in Figure 5-2

Time Series: Discharge(Q) 150

100

50 mm/day

0 Jan 01, 07 Apr 01, 07 Jul 01, 07 Oct 01, 07 Jan 01, 08 Date

Time Series: Discharge(Q) 8

mm/day 2

0 Jan 01, 07 Apr 01, 07 Jul 01, 07 Oct 01, 07 Jan 01, 08 Date

Figure 5-2 River flow in Babak River (upper: Gebong Station, below: Lantandaya Station).

5.6 Parameter Calculation

5.6.1 Recession constant

The value of a is estimated from the linear regression plot of Qt and Qt+1 using discharge data from 1985 until 2005 for both discharge station (Figure 5-3). The data plot should follow the recession condition, which is at least having five days decreasing flow continuously. The result of a for Lantandaya Station is 0.95 and 0.94 for Gebong Station. These values indicate that Lantandaya Station has slower recession than Gebong. This condition may be caused in the different condition of physical characteristics of the two rivers where the station is located.

Lantandaya Station: Q vs Q t (t+1) 4 Data 3.5 Fit Y = T 3

2.5

2 (mm/day)

1.5 (t+1) Q 1

0.5

0 0 1 2 3 4

Q (mm/day) vs Q Gebong Station:t Q t (t+1)

Data 40 Fit Y = T

(mm/day) 20 (t+1) Q 10

0 0 10 20 30 40

Q (mm/day) t

Figure 5-3 Recession constant.

5.6.2 Model parameter of simple dynamic systems

The characteristic of the recession flow for two discharge stations is illustrated in the plot of – dQ/dt against Q observed as shown in Figure 5-4. The data plotted in the Figure 5-4 include all of the discharge data from 1985 until 2005 which has a recession characteristic of at least five days declining flow as Eq.(5.6). The plot is presented in the log-log axis to see the pattern of the recession for high flow and low flow conditions. The recession for high flow is faster than the low flow. Moreover, the different patterns of the recession characteristics are confirmed for the two stations. In the plot of –dQ/dt and Q in Lantandaya Station, the data shows two group of higher and lower part which may happen in result of the transformation process of the water level data to discharge, which is sensitive for low discharge data.

Lantandaya Station Gebong Station 1 2 10 10 ) ) 2 2

0 0 10 10

-1 -2 10 10 -dQ/dt (mm/day -dQ/dt (mm/day -2 -4 10 10 -1 0 1 2 2 3 4 5 6 7 8 10 10 10 10 Q(mm/day) Q(mm/day) Figure 5-4 Recession plot of Lantandaya and Gebong Station (log-log scale).

The parameters of c1, c2, and c3 were defined by the quadratic regression of a plot of ln(-dQ/dt) against ln(Q) as shown in Figure 5-5. The values of c1, c2, and c3 for Lantandaya Station are - 3.3995, -0.8166, and 1.0385 while for Gebong Station are -2.7266, 0.9284 and 0.0063. These values were then used for dQ/dt function. Therefore, we obtain two dQ/dt functions for Lantandaya and Gebong Stations, which later will be used for flow forecasting following Eq.(5.17).

Lantandaya Station Gebong Station 1 4 ) ) 0 2 2 2

-1 0 -2 -2 -3

-4 -4 ln(-dQ/dt, mm/day ln(-dQ/dt, mm/day

-5 -6 0 0.5 1 1.5 2 2.5 -2 0 2 4 6 ln(Q, mm/day) ln(Q, mm/day)

Figure 5-5 Recession plot of Lantandaya Station and Gebong Station (ln-ln scale). The yellow dots show the plot of recession flow (dQ/dt) with the flow (Q), while the red dots represent the bin average of the –dQ/dt. The red line shows the quadratic regression of the bin average of the –dQ/dt.

5.7 Forecasting Result

5.7.1 Recursive digital filters model forecasting

The recession constant a is used for forecasting the low flow. The forecasting result shows that for both of the station has the underestimate discharge. The forecasting result (Figure 5-6)

following the exponential graph while the observed discharge in the Lantandaya Station shows almost stable flow while in Gebong Station shows the fluctuation of the flow, which may be influenced by rainfall that is still happening by the time of the forecasting.

5.7.2 Simple dynamical systems forecasting

The result of the flow forecasting by using simple dynamical systems (Figure 5-6) shows that the forecasting result by using the model tends to be lower for both stations compared to the measured river discharge. The forecasting result also shows some discontinuation flow in Lantandaya Station, e.g., the forecasting flow starts from 01 June 2017. The forecasted flow ends up before 60 days of forecasting. The discontinuity of the forecasted result may be because of the poor quality of the calibration data as it is strongly influenced by the values of parameters c1, c2, and c3.

Lantandaya Sta.: starting time 01 June 2006 20

Simple dynamical system 15 Recursive digital model Observe discharge

Q(mm/day) 5

0 10 20 30 40 50 60 70 t(days)

Gebong Sta.: starting time 01 June 2006 20

Simple dynamical system 15 Recursive digital model Observed discharge

Q(mm/day) 5

0 10 20 30 40 50 60 70 t(days)

Lantandaya Sta.: starting time 01 June 2007 20

Simple dynamical system 15 Recursive digital model Observe discharge

Q(mm/day) 5

0 10 20 30 40 50 60 70 t(days)

Gebong Sta.: starting time 01 June 2007 20

Simple dynamical system 15 Recursive digital model Observed discharge

Q(mm/day) 5

0 10 20 30 40 50 60 70 t(days)

Figure 5-6 Forecasted flow based on two methods in Lantandaya Station and Gebong Station.

5.7.3 Forecasting performance

Each of the forecasted results is evaluated by calculating the error value between the forecasting result and the baseflow based on the calculation of observed data to evaluate the quality of both forecasting method. The horizontal axis shows the starting time of the baseflow forecasting while the vertical axis shows the lead time of the baseflow forecasting.

The graph in Figure 5-7 indicates the error value for each lead time forecasting. The shorter the forecasting, the error value becomes less. The smallest error value is around 40% which happens in the 15 days after the baseflow forecasting was started. The error increases as the lead time increases. It indicates that the forecasting method should be still improved. The forecasting result tends to underestimate compared to the baseflow based on the observed values.

Recursive Digital Simple Dynamical Systems Lantandaya Station: Error (%) Lantandaya Station: Error(%) 200 100 200 100

80 80 150 150

60 60 100 100 40 40 50 50 Lead time (day) Lead time (day) 20 20

0 0 0 5 6 7 8 9 5 6 7 8 9 Month Month

Gebong Station: Error(%) 200 100 Gebong Station: Error(%) 200 100

80 150 80 150

60 60 100 100 40 40

50 50 Lead time (day) 20 Lead time (day) 20

0 0 0 0 5 6 7 8 9 5 6 7 8 9 Month Month

Figure 5-7 Error value of the low flow forecasting compared with observed baseflow.

5.8 Summary

The baseflow forecasting based on the recursive digital filters model tends to result in underestimated value compared with the simple dynamic system. The underestimate value of

the recession forecasting may happen because of the assumed exponential decay of streamflow parameterized only with a single decay value of a in this approach.

Although the results of the simple dynamic system also show some underestimations, it provided higher accuracies at least for the shorter lead time ranges (<15 days) than the recursive digital filters model. The reason for both of the approaches showing underestimations may be mainly because of the ignorance of rainfall events during the forecasting periods. In some cases, the rainfall still happens in the middle of the forecasting period even in dry seasons, which leads to the increase in the observed baseflow. The result of the flow forecasting in Lantandaya Station for simple dynamic system shows discontinuation of the flow. It may happen due to the poor quality of the data for calibration.

5.9 References

Araghinejad, S. (2011). An approach for probabilistic hydrological drought forecasting. Water Resources Management, 25(1), 191-200. Cancelliere, A., Di Mauro, G., Bonaccorso, B., & Rossi, G. (2007). Drought forecasting using the standardized precipitation index. Water Resources Management, 21(5), 801-819. Durdu, Ö. F. (2010). Application of linear stochastic models for drought forecasting in the Büyük Menderes river basin, western Turkey. Stochastic Environmental Research and Risk Assessment, 24(8), 1145-1162. Eckhardt, K. (2005). How to construct recursive digital filters for baseflow separation. Hydrological processes, 19(2), 507-515. Eckhardt, K. (2008). A comparison of baseflow indices, which were calculated with seven different baseflow separation methods. Journal of Hydrology, 352(1), 168-173. Han, P., Wang, P. X., Zhang, S. Y., & Zhu, D. H. (2010). Drought forecasting based on the remote sensing data using ARIMA models. Mathematical and Computer Modelling, 51(11), 1398-1403. Hatmoko, W., Raharja, B., Tollenaar, D., & Vernimmen, R. (2015). Monitoring and Prediction of Hydrological Drought Using a Drought Early Warning System in Pemali-Comal River Basin, Indonesia. Procedia Environmental Sciences, 24, 56-64. Kikkawa, H., & Takeuchi, K. (1975). Use of statistical knowledge of droughts for alleviating drought problems. Paper presented at the Water for Human Needs; Proceedings of the World Congress On Water Resources. Kirchner, J. W. (2009). Catchments as simple dynamical systems: Catchment characterization, rainfall‐runoff modeling, and doing hydrology backward. Water Resources Research, 45(2). Madadgar, S., & Moradkhani, H. (2013). A Bayesian framework for probabilistic seasonal drought forecasting. Journal of Hydrometeorology, 14(6), 1685-1705. Mishra, A., & Desai, V. (2005). Drought forecasting using stochastic models. Stochastic Environmental Research and Risk Assessment, 19(5), 326-339. Mishra, A., Desai, V., & Singh, V. (2007). Drought forecasting using a hybrid stochastic and neural network model. Journal of Hydrologic Engineering, 12(6), 626-638. Mishra, A. K., & Singh, V. P. (2011). Drought modeling–A review. Journal of Hydrology, 403(1), 157-175.

Özger, M., Mishra, A. K., & Singh, V. P. (2012). Long lead time drought forecasting using a wavelet and fuzzy logic combination model: a case study in Texas. Journal of Hydrometeorology, 13(1), 284-297. Van Loon, A. F. (2015). Hydrological drought explained. Wiley Interdisciplinary Reviews: Water, 2(4), 359-392.

6 Concluding Remarks The study aims to understand the effect of the climate phenomenon, ENSO specifically, to the hydro-meteorological drought condition in the study area, to develop the method of drought forecasting system based on ENSO indices, and to build hydrological drought forecasting system based on the low flow characteristics in Lombok and Sumatra Islands. Conclusion for each chapter are as follows:

1. In chapter 3, meteorological and hydrological drought identification was performed in Lombok River Basin by using drought index. The drought index could be used for meteorological drought identification in Lombok Island and resulted in a good agreement with the previous study by (Kirono et al., 1999) regarding drought duration. The 3-month running means drought analysis was selected for meteorological drought analysis considering the physical characteristic of drought and also the seasonal feature of the climate. Drought condition of the Lombok Island characterized by a duration, frequency and also drought area coverage. Two catchments were selected for hydrological drought analysis. Intermediate correlation between meteorological drought, hydrological drought, and ENSO shows a different pattern between two catchments. In the Babak Catchment, the non-significant relationship between the three components is happening in Jan-Feb-Mar (JFM), which is a wet season. Meanwhile, in the Jangkok Catchment, the high correlation between the three component is happening in the dry season between Jun-Jul-Aug (JJA), Jul-Aug-Sep (JAS) and Sep-Oct-Nov (SON). Lag correlation between hydrological drought and ENSO is there for hydrological drought SON season with ENSO in JJA season. 2. In chapter 4, the study of the trend of historical rainfall in the Batanghari River Basin was done. The results of Mann-Kendall trend analysis, in general, show no significant trend of any changes either for monthly, seasonal nor annual rainfall in Sultan Thaha station and GPCC data. The only trend seen is in MAM season for Sultan Thaha data with the significant decreasing trend (at < 0.05). As MAM season is a transition from wet to dry season, the decreasing trend of rainfall in this season may indicate the earlier development of the dry season. The impact of local SST and Nino3.4 (represent ENSO) to the rainfall shows a significant positive influence of local SST to rainfall during Jul- Aug-Sep and negative correlation of Nino3.4 with rainfall during Jul-Aug-Sep-Oct. For the seasonal correlation, JJA season shows significant correlation with both local SST and Nino3.4 while SON season correlates only with Nino3.4. The significant negative

correlation of Nino3.4 with both JJA and SON seasons may indicate a strong and prolong impacts of ENSO to SON season in the river basin. 3. In chapter 5, the low flow forecasting based on recession characteristics was performed by using two methods, namely recursive digital filters and simple dynamic systems. The result shows that the baseflow forecasting based on the recursive digital filters model tends to result in underestimated value compared with the simple dynamic system. The underestimate value of the recession forecasting may happen because of the assumed exponential decay of streamflow parameterized only with a single decay value of an in this approach. Although the results of the simple dynamic system also show some underestimations, it provided higher accuracies at least for the shorter lead time ranges (<15 days) than the recursive digital filters model. The reason for both of the approaches showing underestimations may be mainly because of the ignorance of rainfall events during the forecasting periods. In some cases, the rainfall still happens in the middle of the forecasting period even in dry seasons, which leads to the increase in the observed baseflow. The result of the flow forecasting in Lantandaya Station for simple dynamic system shows discontinuation of the flow. It may happen due to the poor quality of the data for calibration.

Overal, this dissertation work result has implied the integrated method of drought analysis in two study area. Drought identification for both meteorological and hydrological drought was done by using drought indices. Although the result is still overestimated, could reflect the drought duration and drought coverage area in Lombok Island. Drought forecasting system by using low flow characteristic is still underestimated and need to be improved. Overall, after assessing the correlation between meteorological and hydrological drought with ENSO, it is apparent that there is seasonal correlation between ENSO and the drought condition in the two study area. Further assessment on how to use the ENSO information for drought forecasting system is necessary.