Nurdiati 

et al.
Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Terrestrial, Atmospheric
https://doi.org/10.1007/s44195-022-00016-0
and Oceanic Sciences

ORIGINAL ARTICLE Open Access

The impact of El Niño southern oscillation

and Indian Ocean Dipole on the burned area
in Indonesia
Sri Nurdiati1*   , Fahren Bukhari1, Muhammad Tito Julianto1, Ardhasena Sopaheluwakan2, Mega Aprilia1,
Ibnu Fajar1, Pandu Septiawan1 and Mohamad Khoirun Najib1 

Abstract 
Land and forest fire have been identified as one of the main problems contributing to forest biodiversity and Global
Warming and well known as the phenomenon affected by El Niño Southern Oscillation (ENSO) and Indian Ocean
Dipole (IOD). The total burned area becomes higher when either El Niño or positive IOD occur. This research aims to
analyze and quantify the direct correlation of the Niño 3.4 and difference between west and east pole of IOD sea sur-
face temperature anomaly (SSTA) to the burned area in Indonesia and the impact of ENSO and IOD of each category
on the burned area. The correlation between spatial location with Niño 3.4 and difference IOD SST’s will be analyzed
using a heterogeneous correlation map. Meanwhile, the quantitative impact will be calculated based on the singular
value decomposition analysis result to each year categories. The most significant impact of El Niño has occurred on
Merauke following Kalimantan shows the strongest correlation between burned area and Niño 3.4 SST. However, the
significant increase of burned area only occurred during very strong El Niño. Both areas can have double amount of
burned area during peak fire in very strong El Niño. Moderate El Niño have the most diverse impact with the stronger
one occurs on Kalimantan and Merauke. Weak El Niño can have a significant impact if occurred simultaneously with
positive IOD. Even more, it can surpass the effect of a single Moderate El Niño. Meanwhile, the strongest IOD impact
happened in the southern part of Sumatra.

Key points 

1. Correlation map analysis between ENSO and IOD’s sea surface temperature with burned area.
2. Behaviour analysis of burned area in Indonesia in response to ENSO and IOD.
3. Identify and quantify the impact of ENSO-IOD’s teleconnection in Indonesia.

Keywords:  Forest fire, Burned Area, El Niño Southern Oscillation, Indian Ocean Dipole, Sea surface temperature

1 Introduction
In recent years, prolonged droughts and intensive use
of tropical forests have increased land and forest fire
events (FAO 2007). Land and Forest fires have been
*Correspondence: nurdiati@apps.ipb.ac.id
identified as one of the main environmental issues that
1
Department of Mathematics, FMIPA, IPB University, Dramaga, have a long-term impact on biodiversity and global
Bogor 16680, Indonesia climate (Ganjam and Sudhakar 2015). In Indonesia,
Full list of author information is available at the end of the article

© The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this
licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/.
Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 2 of 17

land and forest fires occur annually on Central Suma- emissions (Burton et  al. 2020). A few moderate El Niño
tra, South Sumatra, South Kalimantan, and Merauke happened in 2002, 2006, and 2009 (Burton et al. 2020).
(Nurdiati et  al. 2019). The number of peatland areas In the past few years, much research shows strong El
in those regions produces high carbon emissions dur- Niño in 1997 and 2015 on Indonesia’s land and forest fire
ing the dry season (Osaki et  al. 2016; Septiawan et  al. (Edwards et al. 2020; Tan et al. 2020; Yulianti and Haya-
2019). The biggest forest fire event in Indonesia for the saka 2013). Many people in Indonesia, Southeast Asia,
last 25  years happened in 1997. Most of the burned and worldwide expressed deep concern at the extent
area in 1997 came from widespread peatland fires caus- and severity of fires in Sumatra and Kalimantan forested
ing unusually high carbon emissions in that year (Page areas in the late 1997 dry season. It is generally believed
et al. 2002). In 1997, 9,755,000 hectares (ha) burned in that the health of over 20 million people was adversely
Indonesia (Luca Tacconi 2003). affected in Southeast Asia in late 1997 because causes
1997 El Niño year fire event causing dry peat smoul- were not well analyzed first (Byron and Shepherd 1998).
ders for long periods and burns down to the water table. Field et  al. (2016) state that longer-term records of air-
When this happens, the tree roots are exposed, and the port visibility in Sumatra and Kalimantan show that 2015
peat and forest vegetation become unstable. There- ranked after 1997 and alongside 1991 and 1994 as among
fore, it generates peat subsidence, a massive tree falls, the worst episodes on paper.
and the consequent loss of large forest areas, leading to Another global phenomenon that affects the 2015
more problems for flora and fauna (Harrison et al. 2009). drought season in Indonesia is Indian Ocean Dipole
Although the event happened in Indonesia, haze from (IOD) (Avia and Sofiati 2018). The IOD phenomenon has
the fires can extend to Malaysia, Singapore, and Thai- two phases: the dry phase called positive IOD and the
land, shrouding them in smoke and affecting economic wet phase known as negative IOD. A positive IOD is indi-
activities for millions of more people, resulting in bil- cated by the SST of the western Indian Ocean (Africa)
lions of more dollars of financial losses (Harrison et  al. being warmer than the eastern Indian Ocean (Indonesia
2009). Even more, the carbon emission from Indone- and its surroundings). In contrast, the opposite condi-
sia’s peatland wildfires makes a massive contribution to tion indicates the negative IOD (Li et al. 2002; McPhaden
the increase of global emission (Hooijer et al. 2006). The and Nagura 2014). In the transitional period of the dry
second biggest one happened in 2015 with 4,604,569  ha season, coincided with the lasting period of the positive
were burned during the fire season (overall accuracy IOD phase, there will be increasingly severe drought and
84%) (Lohberger et  al. 2017). Both of those two events widespread Indonesia (Avia and Sofiati 2018). When pos-
happened during strong El Niño in each particular year. itive IOD and El Niño occurred concurrently, Indonesian
Moreover, El Niño events explain most of the year-on- rainfalls decreased significantly than in the single event
year variation in fire (Edwards et al. 2020). of positive IOD or El Niño (Nur’utami and Hidayat 2016).
ENSO is a global phenomenon that describes the recur- In general, positive (negative) phases of the IOD result in
ring pattern of climate variability in the eastern Pacific rainfall extremes becoming drier (wetter) than the nor-
Ocean, characterized by SSTA and anomaly sea level mal phase. However, this generalization impacts ENSO
change (Neelin et  al. 1998; Hu et  al. 2017). The warm- and IOD for all Indonesian regions (Kurniadi et al. 2021).
ing phase of sea surface temperature (SST) describes the In addition, the relations between precipitation and land
El Niño event, while the cooling phase describes the La and forest fire were not uniform in Indonesia (Fanin and
Niña event. During El Niño events, SST in the Eastern Werf 2016). Therefore, the impact of ENSO and IOD are
Equatorial Pacific become hotter than normal conditions. different in each area of Indonesia, which frequently suf-
Meanwhile, during La Niña conditions, SST in the East fers from land and forest fire.
Equatorial Pacific are lower than normal conditions. The Other than global climate phenomena, land and forest
occurrence of El Niño causes the rainy season to come fires in Indonesia are also affected by human behaviour.
later and prolongs the dry season duration in Indone- Examples of the land and forest fire caused by human
sia. Normally dry season in Indonesia occurred around behaviour are land clearing, shifting cultivation (agricul-
may until October, while the rainy season occurred from tural waste), transmigration, reforestation, oil palm/rub-
November until April. Otherwise, La Niña impact on ber and other big-scale plantations (Luca Tacconi 2003).
Indonesia causes the rainy season to come earlier and However, Kemen et  al. (2019) showed that the cause of
a shorter dry season. In the last 25  years, two strong El the fires across Indonesia from 2001 to 2016 has no
Niño happened in 1997 and 2015. Even though 2015 apparent pattern. The leading causes of land and forest
El Niño have highest Niño 3.4 index in those peri- fire are vary depending on the location, such as oil palm
ods (L’Heureux 2016), 2015 El Niño event was not the plantation on Kalimantan and logging roads in Papua.
most extensive fire event in the total burned area or fire Meanwhile, Sumatra has three leading causes: oil palm
 Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 3 of 17

plantations, grassland/shrublands, and small-scale agri- using information of MODIS data from 0.25° × 0.25° to
culture (Kemen et  al. 2019). Furthermore, variation of 0.1° × 0.1°. The value of those data represents the propor-
leading causes is also found on an annual scale, such as tion of observed grid that burned. Therefore, the value
oil plantation in 2013, small agriculture and plantation in will be converted into k­ m2 units (1 grid =  111,321  ­km2)
2015, grassland/shrubland in 2014 and 2016. Even though multiplied by 100 to convert the value into ha. The data
land and forest fire happened annually, the scale of the described in Giglio et  al. (2013) and van der Werf et  al.
event is much smaller than during ENSO years (Luca (2017), as well as can be accessed and downloaded for
Tacconi 2003). Therefore, this research only focused on free from the Global Fire Data website.
causes of land and forest fire related to the climate. Meanwhile, the precipitation variable uses a monthly
Nurdiati et al. (2021a, b) examine the spatial and tem- precipitation average from ECMWF Reanalysis 5th Gen-
poral analysis of the El Niño effect on forest and land fires eration (ERA5) data. The spatial and temporal informa-
on Kalimantan and Sumatra using the hotspot as a fire tion for the data is same as burned area data. For the
indicator. There is a direct spatial and temporal correla- ENSO and IOD variables, this research uses SSTA in
tion between hotspot-ENSO SST and hotspot-IOD SST ERA5 data from 1991 to 2020 (Hersbach et al. 2019). The
(Dafri et al. 2021). However, the result of those research spatial information for both phenomena corresponds
cannot be used to directly describe the impact of ENSO to the Niño3.4 index and DMI index. Nino 3.4 index
and IOD on the burned area in Indonesia. Proportion of is an anomaly of 190°  E–240°  E, 5°  N–5°  S SST (Tren-
burned area recorded as hotspot is less than 0.39 (Tansey berth 1997). The anomaly is based on the average of last
et  al. 2008). Even though there are some improvements 30  years data (1991–2020). Meanwhile, DMI index is
in the recent year, correlation between hotspot and anomalous SST gradient between west IOD (50° E–70° E,
burned area is different on each area caused by different 10°  N–10°  S) and east IOD (90°  E–110°  E, 0°  N–10°  S)
characteristics of the region (Tansey et  al. 2008; Vetrita (Saji et al. 1997). Because of different size of spatial infor-
and Cochrane 2020; Ardiansyah et al. 2017). Meanwhile, mation, the east IOD SST is interpolated so that it has
analyzing burned areas in Indonesia and assessing corre- same spatial dimension with the west one. Both of data
sponding impacts, e.g., ­CO2 emissions, can be potentially are monthly data with 0.25° × 0.25° spatial resolution,
used for climate/economic damage assessments (Kras- produced by the European Centre for Medium-Range
ovskiy et  al. 2018). Therefore, this research analyses the Weather Forecast (ECMWF) and can be accessed for free
interrelationship patterns of burned area and precipita- (Hersbach et al. 2019).
tion in Indonesia with the influence of ENSO and IOD.
The first objective of this research is to analyze the cor-
2.2 Singular value decomposition
relation between global phenomena (ENSO and IOD)
Singular value decomposition (SVD) is matrix decompo-
to the burned area in Indonesia for each observed grid.
sition method that factorize matrix A in the form of:
After that, a detailed analysis was provided by looking at
the different joint patterns for each possible ENSO and A = U V T , with σ1 ≥ σ2 ≥ · · · σr ≥ 0. (1)
IOD category for each region with an annual forest fire.
Lastly, this research indicated ENSO-IOD teleconnec-  represent the singular value matrix of A with σi repre-
tion impact on the land and forest fire in Indonesia. This sent the singular value of ith mode. U and V represent
research can explain how ENSO and IOD affect forest left and right singular vector of A respectively. Therefore,
fires in Indonesia related to the burned area and monthly the equation can be written (Nicholson 2001):
average precipitation clearly to contribute positively to  
σ 0
forest fire modelling in the future. An×p = Un×n VT . (2)
0 σ p×p

2 Data and methodology The singular value decomposition in this research uses

2.1 Data the cross-correlation concept to analyze the joint pattern
This research’s land and forest fire indicator is monthly between burned area and precipitation or SSTA. The idea
burned area data from 1997 to 2016 produced by Global is described by Navarra and Simoncini (2010). Suppose
Fire Emission Database (GFED), and MODIS burned there are two matrices, namely matrix P (burned area)
area data from 2017 to 2020. The spatial information and matrix S (precipitation/SSTA), that have the same
for the burned area data is 95° E–141° E and 6° N–11° S. time span. In that case, a Cross-Correlation matrix is
The burned area variable in GFED4.1s data has a grid formed following the equation C = PS T  . The decompo-
resolution of 0.25° × 0.25° for each data element, while sition is applied to the matrix C  . Therefore, P and S can
in MODIS data has a grid resolution of 0.25° × 0.25° have different spatial dimensions (number of rows). As
and 0.1° × 0.1°. Therefore, the GFED4.1s data is rescale long as the temporal dimension (number of columns) is
Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 4 of 17

the same, the decomposition process can be done. The Table 1  Ranges of Pearson correlation value
result from SVD analysis of Cross-Correlation matrix r (value) Strength of correlation
will be called joint pattern to describe pattern of P that
related specifically with pattern of S. Since it calculate 0 No correlation
related pattern of two variables, dominant pattern of 0 to ± 0.25 Very weak correlation
each variable when calculated individually is not always ± 0.25 to ± 0.5 Sufficient correlation
become dominant joint pattern. ± 0.5 to ± 0.75 Strong correlation
± 0.75 to ± 0.99 Very strong correlation
2.3 Empirical orthogonal analysis 1 Perfect correlation
The Empirical Orthogonal Function (EOF) analysis aims
to transform the original correlated p variables into
k orthogonal (uncorrelated) components. Let X be a m × p , respectively. By decomposing the cross-correlation
matrix of size n × p containing a data set with n number matrix (PST) using SVD (Eq. 1), the U,  , and V matrices
of variables and p time. SVD of X with rank(X) = r is the are constructed. The first EOF mode of HCM is the Pear-
factorization as in Eq. (1) so that we get: son correlation between the expansion coefficient (the
r product of the first column of V notated by v1 and the S
matrix) and the burned area matrix P, i.e., r(v1T S, P) . The

X= ui σi viT . (3)
i=1 expansion coefficient is a time vector with a size of 1 × p .
The HCM is obtained by calculating the Pearson correla-
The V matrix of Eq.  (1) is the EOF matrix or vector
tion of the time vector v1T S with every space dimension of
coefficient and U  is the EOF score matrix or principal
matrix P so that the first EOF mode of HCM has a size of
component. Principal component scores can be written:
n × 1 . Remember that n is a spatial dimension consisting
zi = ui σi , i = 1, 2, . . . , r. (4) of longitude and latitude dimensions. Therefore, we can
visualize the correlation map for each grid location in Indo-
The variance of ith principal components i = 1, 2, . . . , r nesia by transforming the n-vectors to a longitude-latitude
is matrix. In this study, the value of m and p are 1160 and 240,
 r respectively. Meanwhile, the value of m is 8241 and 6537
for the SSTA of ENSO and IOD, respectively. The result

µi = σi2 σi2 , (5)
i=1 of HCM analysis will be called correlation pattern since
it analysed the correlation of related pattern between two
where i = 1, 2, …, r is the singular value of the matrix X. variables.
In practice, the k mode of the 1st EOF or the first princi-
pal component describes the largest proportion of vari- 2.5 Euclidian distance and dynamic time wrapping
ance. The 2nd EOF mode has the second-largest variance Euclidian distance and Dynamic Time Wrapping (DTW)
and so on. The kth EOF mode has the kth maximum vari- are a well-known method that can compare two sig-
ance independent of the previous EOF mode (Hannachi nals by finding the distance between the two signals
2004). (Bayarjargal et  al. 2006; Lhermitte et  al. 2011). Euclid-
ian distance between X = (x1 ≥ x2 ≥ · · · ≥ xn ) and
2.4 Heterogeneous correlation maps Y = (y1 ≥ y2 ≥ · · · ≥ yn ) can be calculated following the
The spatial analysis used the Heterogeneous Correla- equation below:
tion Map (HCM) to describe the correlation of hotspots 
in Indonesia with variances values of climate indicators.
 n

HCM hotspots in Indonesia have been obtained from Deuclid =  (xi − yi )2 . (6)
the Pearson correlation between expansion coefficient i=1
of climate indicators variances with burned area matrix
While the DTW distance of X = (x1 ≥ x2 ≥ · · · ≥ xn )
r(V T S, P) (Björnsson and Venegas 1997). Ranges of Pear-
and Y = (y1 ≥ y2 ≥ · · · ≥ yn ) is defined as:
son correlation value can be seen in Table 1.
From the scratch three-dimensional matrix data (longi- Dtw (X, Y ) = d(m, n), (7)
tude × latitude × time), we rearrange that matrix to two-
dimensional matrix data (space × time). Therefore, suppose   
d(i, j) =xi − yi  + min d(i, j − 1),
that P and S are the two-dimensional matrices representing  (8)
the burned area and SSTA with a matrix size of n × p and d(i − 1, j), d(i − 1, j − 1) ,
 Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 5 of 17

d(0, 0) = 0; d(i, 0) = ∞; d(0, j) = ∞; 2021; Nurdiati et  al. 2021b). In this research, SSTA of
i = 1, 2, . . . , m; j = 1, 2, . . . , n. ENSO and IOD is analyzed with burned area data from
1997 to 2016 to get correlation map that indicates the
In this research, the Euclidian distance was used to get relation of ENSO and IOD with burned area in Indonesia
the different characteristics among each category in the respectively. The first analysis uses ENSO’s SSTA, and the
same month. Meanwhile, the DTW distance is used to second one uses IOD’s SSTA.
nullify the shift between the characteristic of the signal The correlation map analysis between burned area
by comparing the value in the same index and comparing and ENSO’s SSTA result in one dominant pattern with
the value with the value of the previous and next index to 98.78% value of variance explains (Fig.  1). This pattern
find the optimum distance (Muller 2007). represents 98.78% of events from 1997 to 2020 in the cor-
relation of burned area and ENSO. The pattern shows
3 Results and discussion that the largest burned area happened in 1997; more than
In this section, we present the analysis that is divided 15 million ha burned in a month (Fig. 1a). The second one
into two groups. The first one analyses correlation map occurred in 2015, with a maximum value of more than 11
between burned area and SSTA (Niño 3.4 and IOD). The million ha burned in a month. However, each grid loca-
second one explores spatial and temporal character of tion’s spatial correlations with ENSO’s SSTA were less
the burned area and precipitation for each year group than 0.5 (sufficient correlation). The higher correlation
divided based on the overall ENSO and IOD strength. occurred from the centre to the eastern part of Indone-
As already mentioned before, ENSO and IOD’s impact sia (Fig. 1b). This result coincident with previous research
cannot be generalized for all regions in Indonesia. The which shows cold SSTA from centre to the eastern part of
second analysis will be carried out for each region with Indonesian sea during El Niño (Hendon 2003) resulting
recurrent land and forest fire events. less precipitation (Nur’utami and Hidayat 2016), more
burned area on the surrounding region (Pan et al. 2018).
3.1 Analysis of correlation map between burned area Maximum correlation between the temporal pattern
and SSTA (ENSO and IOD) of burned area and ENSO is 0.381 (sufficient correlation,
SSTA of ENSO and IOD are variables that have a domi- Table 1). Those values represent the correlation between
nant joint pattern with a hotspot in Indonesia (Dafri et al. monthly total burned area in Indonesia and monthly

Fig. 1  The correlation map between burned area and ENSO: a the HCM and burned area principal temporal pattern b violin plot of correlation
coefficient on Riau (blue box), South Sumatra (red box), Kalimantan (black box), and Merauke (green box). Red dots and stars represent the average
value and outlier of correlation distribution in each area
Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 6 of 17

Niño 3.4 index. However, the temporal pattern only cap- correlation), which occurred on South Sumatra. Due to
tures the fire event in the middle to the end of the year. the spatial location of east IOD’s SSTA (90°  E–110°  E,
These periods represent the dry season that occurred 0° N–10° S) which is cooler than normal during positive
caused by Australian Monsoon wind. Although the peak IOD is in the middle of Sumatra and Kalimantan which
of El Niño occurs in winter, the burned areas do not resulting less rainfall on the surrounding region. There-
appear in winter because Indonesia has already experi- fore, the spatial correlation of surrounding regions higher
enced the rainy season. compared to the ENSO. Those results can be seen in
There are two land and forest fire events in Indone- the violin plot of South Sumatra and Kalimantan, which
sia which happened in the beginning and the middle of shows a higher average (indicated by the red dot) and
the years (Septiawan et al. 2019). The second dry season maximum value than the ENSO one. Otherwise, Mer-
which occurred in the middle of the year becomes longer auke’s violin plot show the opposite behaviour with a
when El Niño develops. Those different periods make lower average and minimum value than ENSO.
land and forest fire patterns at the beginning of the year The violin plot of Riau (Fig. 2) shows higher maximum
missing from the correlation map. However, these pat- value of correlation than the ENSO one (Fig. 1), but the
terns can be identified in the violin plot of Riau by the average is lower. Moreover, the minimum value belongs
negative value of correlation. These negative values rep- to negative weak correlation which indicate different land
resent the occurrence of burned area in Indonesia when and forest fire event happened oppositely with IOD. As
ENSO is not yet developed. mentioned before, Riau is a region with two forest fire
The correlation map analysis between burned area and events in a year, which happened in the beginning (Feb-
IOD also resulted in one dominant pattern with 96.15% ruary until March) and middle (June until September) of
value of variance explained. Similar to the previous the year. Considering the impact of positive IOD’s is start
analysis, the joint pattern only captured burned area in from June (Yulihastin et  al. 2009), we can conclude that
the middle to the end of the year. This period coincides events that have weak positive correlation occurred in
with the impact of IOD on Indonesia’s monsoonal sea- the middle to end of the year. The weak negative correla-
son, so the temporal plot also showed only from June to tion represents land and forest fire events that happened
November for each year. The maximum spatial correla- at the beginning of the year. Due to the high dominance
tion between burned area and IOD is 0.464 (sufficient

Fig. 2  The correlation map between burned area and DMI: a the HCM and burned area principal temporal pattern b violin plot of correlation
coefficient on Riau (blue box), South Sumatra (red box), Kalimantan (black box), and Merauke (green box). Red dots and stars represent the average
value and outlier of correlation distribution in each area
 Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 7 of 17

of events in the middle of the year, the first EOF mode Sumatra, Kalimantan, and Merauke. Therefore, the analy-
cannot capture the beginning of the year events. sis will be carried out only in those specific regions fol-
The low value of spatial and temporal correlation lowing the average order correlation in ENSO HCM from
means that Indonesia’s land and forest fire influenced the highest to lowest one.
by many factors. Two of them are El Niño and positive
IOD that affected the duration of Indonesian dry season 3.2.1 Merauke
in the particular year. However, the temporal pattern in Land and forest fires on Merauke during strong El Niño
Figs.  1 and 2 looks similar to each other. The first rea- in 2015 significantly impacted Timika. Regional Disaster
son for this similarity is the very high dominance pat- Relief Agency (BPBD) of Merauke claimed 93 hotspots
tern in the burned area data. The second reason for this were detected from burning forests and land activities
similarity is the interconnection (relationship) between (Oja et al. 2019). Lohberger et al. (2017) show that more
ENSO and IOD. In 2011, Cai et al. stated that the Niño3.4 than 800 thousand ha burned, with around 500 thousand
index was moderately correlated with the DMI index in ha occurred on peatland. Those data contributed to the
June–August (JJA). Even more, the correlation becomes total burned area in Indonesia by approximately 20–25%
stronger in September–November (SON). The maximum during 2015’s fire event. HCM analysis in Sect.  3.1 also
positive correlation between ENSO and IOD occurred gives sufficient correlation between burned area and
when the IOD led ENSO by 2  months, while the maxi- ENSO/IOD, indicating that Merauke is a region prone to
mum negative correlation occurs when IOD is leading land and forest fire events.
ENSO by 16 months (Stuecker et al. 2017). The temporal pattern between burned area and precip-
itation on Merauke shown by Fig. 3. Each line in burned
area plot (Fig.  3a) represents an average of total burned
3.2 Classification analysis of joint pattern between burned area in the composite year respective to each category,
area and precipitation while in precipitation plot (Fig. 3b) represents the average
This section analyses the joint pattern between burned monthly precipitation of each analysed grid in the com-
area and precipitation by grouping the year 1997–2016 to posite year. Figure 3 shows that forest fire events on Mer-
each possible category that builds from ENSO and IOD auke commonly happened from August until November,
categories (Behera et al. 2013; Anteneh et al. 2019). The except for WEN-PI and SEN-PI categories, which can
year groups that obtained from 1997 until 2016 data are: end in December. The peak of the fire event occurred
either in September or October, but mainly in September.
1. Normal year: 1998, 1999, 2001, 2003, 2010, 2012, Meanwhile, the monthly average precipitation decreases
2013 to less than 5 mm/day starting from June until October.
2. Weak El Niño and normal IOD (WEN-NI): 2004, Therefore, the peak fire event happened due to an accu-
2014, and 2018 mulated dry condition from June to September. The high
3. Weak El Niño and positive IOD (WEN-PI): 2006 and proportion of burned area that occurred in peatland and
2019 the geographic location of Merauke make the amount
4. Moderate El Niño and normal IOD (MEN-NI): 2002 of burned area remain high in October. Therefore, for-
and 2009 est fires on Merauke relatively have longer duration than
5. Very strong El Niño category and positive IOD (SEN- other regions (Figs. 5, 7, and 10).
PI): 1997 and 2015 The impact of IOD on Merauke is not visible in the
6. La Niña and normal IOD categories for 2000, 2005, burned area temporal pattern (Fig.  3). The WEN-PI
2007, 2008, 2011, 2017, and 2020 burned area graph is lower than the WEN-NI graphs.
7. La Niña and negative IOD categories in 2016 In precipitation, positive IOD decreases the precipita-
tion value until reaching around 1  mm/day in October,
The impact of La Niña in the 6th group increases the the end of the dry season. In other months, the impact is
monthly precipitation in Indonesia and make the dry sea- unclear when comparing WEN-NI and WEN-PI graphs.
son shorter. Therefore, the correlation between burned Considering the result of the HCM analysis in Sect. 3.1,
area and precipitation decreases, making the analysis less which shows a weak to sufficient correlation on Merauke,
relevant. Even more, the 7th group with the contribution we can state that positive IOD affects the joint pattern of
of negative IOD makes the precipitation on Indonesia burned area and precipitation of Merauke only in the end
even higher and makes the correlation even more irrel- period of the dry season. When positive IOD occurs, the
evant. For that reason, the analysis will be carried out in condition is relatively similar with other categories dur-
groups 1–5. The previous research states that the recur- ing the start of dry season (June–August), but it causes
rent land and forest fire occurred mainly on Riau, South
Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 8 of 17

Fig. 3  Temporal of joint pattern for each category on Merauke. Burned area graph (a) represents the average total burned area of the composite
year. In contrast, the precipitation graph (b) represents the average monthly average precipitation from each grid of the composite year for each
category

high burned area and low precipitation during the end of of El Niño, especially strong El Niño. Weak and moder-
dry season (October–December). ate El Niño can cause two to three times area burnt com-
By looking at burned area graph in Fig.  3, we can see pared to a normal year. Even more, a strong El Niño can
that burned area on Merauke is affected by the strength lead to five times amount of burned area than normal
of El Niño. On average, around 35 thousand ha was year. Meanwhile, El Niño impact on Merauke’s precipi-
burned during peak fire events in the Normal year. The tation has prolonged the duration of the dry season not
peak event of weak El Niño can result in around 0.13 only by holding the arrival of the upcoming rainy season
million ha burned. The number of burned areas during but also making the dry season come earlier from May
moderate El Niño is also 0.11 million ha, and the strong or June. This effect resulting in 2 months longer dry sea-
El Niño will cause more than 0.28 million ha burned only son compared to normal conditions. Therefore, the accu-
during the peak event (Fig. 3). This result shows the sen- mulated dry condition from early dry season results in a
sitivity of land and forest fire on Merauke to the strength much more burned area than the normal condition.

Fig. 4  Spatial of joint Pattern from each category on Merauke. In a normal graph, the colour represents an average of total burned area for each
grid in the composite year only during fire season (July–December). Other categories represent the difference between each category with value in
a normal year
 Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 9 of 17

Figure  4 shows the spatial of the first joint pattern was burned on average during the peak fire event. Fol-
between monthly average burned area (ba) and monthly lowing the strength of El Niño, the number of burned
average precipitation (prec) on Mearuke. Other than areas increases to be around 0.3 million ha during
the normal category, the value subtracts with the nor- WEN-NI and around 0.75 million ha during MEN-NI.
mal category’s value to represent the difference between When positive IOD occurred during weak El Niño,
each category with the normal category. The difference around 0.7 million ha was burned during peak fire.
in the spatial pattern of precipitation is inconsistent in However, the high number of burned areas in WEN-
WEN-NI, WEN-PI, and MEN-NI. However, this result PI lasts until October, resulting in a longer duration of
coincides with the result in Fig.  3, which shows a rela- fire event than MEN-NI. Furthermore, SEN-PI doubled
tively similar number of burned areas during those three the impact of WEN-NI and WEN-PI, resulting in more
categories. Even though the precipitation graph of MEN- than 1.5 million ha burned on average during the peak
NI, WEN-NI, and WEN-PI in Fig. 3 is lower than normal fire event and lasted until November.
condition, the difference of average precipitation is only The impact of ENSO and IOD also appeared in the
in the range of 0–1.5 mm/day (Fig. 4) per grid data. Even precipitation variable (Fig.  5b). During peak dry season
more, there are some increases of average precipitation (July–September), the monthly average of precipitation is
during MEN-NI on same area. Therefore, we can con- less than 3 mm/day for all categories, except the normal
sider that WEN-NI, WEN-PI, and MEN- have a similar year. Furthermore, the precipitation is below the peak
impact on the burned area and precipitation of Merauke. of dry season in normal years from July until Septem-
Different from SEN-PI where the difference can be seen ber, representing much more severe than normal year.
clearly in both Figs. 3 and 4. El Niño and positive IOD prolonged the duration of the
dry season by holding the arrival of the next rainy season.
3.2.2 Kalimantan Besides the normal year, the large increases in precipita-
Kalimantan, also known as Borneo, is the region with tion represent the rainy season from October to Novem-
the most significant contribution to the total burned ber, which is 1 month later than the normal year. Figure 6
area in Indonesia for the last 25  years. In 2015 Kalim- shows the spatial of the first joint pattern between
antan contributed 45–50% of the total burned area in burned area and precipitation for each category on Kali-
Indonesia, with one-third of burned area on Kaliman- mantan. The normal category has the least value of vari-
tan occurred on peat (Lohberger et  al. 2017). Figure  5 ance explains caused of the less fire that happened in the
shows that the number of burned areas from the peak category, so that correlation between the burned area
fire event that commonly happened in September and precipitation is reduce. Figure 6 (normal) shows that
increases following the strength of El Niño and posi- during a normal year, up to 700  ha was burned on one
tive IOD. During a normal year, around 0.15 million ha grid data during one fire season in the region in a range

Fig. 5  Temporal of joint pattern for each category on Kalimantan. Burned area graph (a) represents the average total burned area of the composite
year. In contrast, the precipitation graph (b) represents the average monthly average precipitation from each grid of the composite year for each
category
Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 10 of 17

Fig. 6  Spatial of joint Pattern from each category on Kalimantan. In a normal graph, the colour represents an average of total burned area for each
grid in the composite year only during fire season (July–November). Other categories represent the difference between each category with value in
a normal year

of 114°–116°  E and 2°–4°  S where average precipitation during positive IOD compared to normal IOD. From this
value during the dry season is around 4–7 mm/day. The result, we can conclude that the western burned area
number of burned areas in those regions is increased in region on Kalimantan (110°–114°  E, 1.7°–4°  S) is more
WEN-NI and MEN-NI following the decreased precipi- influenced by positive IOD, while the eastern one (114°–
tation. The decreased precipitation is less than 2  mm/ 116° E, 2°–4° S) is affected by both El Niño and positive
day for each grid, the impact can increase burned area by IOD. Not only increase the burned area of the western
around 2500 ha in WEN-PI and up to 5000 thousand ha part, but WEN-PI also increases the burned area on the
per grid during SEN-PI. eastern part to the same level as the MEN-NI category.
WEN-PI and SEN-PI is causing more decreased precip- The impact is even more in SEN-PI with more decrease’s
itation in the region in a range of 110°–114° E and 1.7°– precipitation.
4°  S, causing more burned area in this region (Fig.  6).
This result is related to the SST that contributes to posi- 3.2.3 South Sumatra
tive IOD located on the west of Kalimantan. Therefore, For the last 25 years, South Sumatra has been commonly
the west part of South Kalimantan has less precipitation known as a region that suffers from land and forest fire

Fig. 7  Temporal of joint pattern for each category on South Sumatra. Burned area graph (a) represents the average total burned area of the
composite year. In contrast, the precipitation graph (b) represents the average monthly average precipitation from each grid of composite year for
each category
 Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 11 of 17

Fig. 8  Spatial of joint Pattern from each category on South Sumatra. In a normal graph, the colour represents an average of total burned area for
each grid in the composite year only during fire season (July–November). Other categories represent the difference between each category with
value in a normal year

for almost every year. Centre of the event is usually in occurred. This statement is reinforced with analysis in
Palembang. The high amount of peatland makes the fires Sect. 3.1, which shows that IOD strongly correlates with
difficult to extinguish, especially when long dry season the burned area over South Sumatra while ENSO only
occurred in 1997 and 2015 during the strong El Niño has sufficient correlation. However, the impact of WEN-
(Kirana et  al. 2016). The joint pattern analysis between NI and MEN-NI in the average precipitation variable is
burned area and precipitation shown in Figs. 7 and 8. Fig- less with only − 0.2 to 0.5  mm/day difference compared
ure  7 shows that the fire event on South Sumatra com- to normal year (Fig. 8). This result is the impact of both
monly happened from July to November, which is the phenomena is prolonged the duration of South Suma-
entire monsoonal dry season on South Sumatra. Figure 7 tra’s dry season. At the same time, the precipitation in the
shows that the peak of the fire event always happened in peak condition remains similar value, as shown in Fig. 7.
September for each category. The minimum precipita- From the normal years temporal pattern in Fig. 7b (the
tion value is less than 5  mm/day for each category. On right panel), the dry season on South Sumatra happened
average, the amount of burned area during one fire sea- when the average monthly precipitation value under
son is spread from 150 to 750 thousand ha. However, if 6  mm/day which is during June to September. When
we separate WEN-PI and SEN-PI from the categories, either El Niño or IOD occurs, the precipitation value on
the amount of burned area is in the range of 100–200 June is around 5 mm/day (beginning of dry season) and
thousand ha for each category. Meanwhile, the peak fire decreases in to around 2–3 mm/day during July to Octo-
event in WEN-PI burnt more than 350 thousand ha, and ber which is significantly drier than normal year. Even
the peak fire in SEN-PI has more than 700 thousand ha more, the precipitation value become around 1 mm/day
burned area during one fire season on average. when both phenomena occurred simultaneously (Fig.  7,
From the burned area graph, we can see that El Niño WEN-PI and SEN-PI). In addition, the dry season dura-
and IOD that occurred coincidentally impacted the num- tion in WEN-PI and SEN-PI are 1  month longer than
ber of the burned area shown by the WEN-PI and SEN- other categories with less than 6  mm/day precipitation
PI graphs. Even weak El Niño can have double impact until November, leading to much more burned areas dur-
than that of moderate El Niño, if accompanied by posi- ing the dry season.
tive IOD. The impact of El Niño and positive IOD is rein-
forced by joint spatial pattern for each category, which 3.2.4 Riau
shows how the burned area spread over the region. Riau is one of the regions in Indonesia that has a low
Figure  8 shows that the most widespread increases rainy season. Therefore, Riau has two rainy seasons and
in burned area and decreased precipitation occurs in two dry seasons in 1 year. The consequence of having two
WEN-PI and SEN-PI. This widespread will be causing pairs of seasons is that the season’s duration is shorter
more burned area in total, as shown by Fig. 7. WEN-NI than regions that only have one rainy season and dry sea-
and MEN-NI graph confirm the previous result that the son in 1 year. The duration of each season is around three
impact of El Niño from weak to moderate is quite com- months, including the transition period of the season.
parable if positive IOD did not occur (Fig. 7). The impact This short duration of the dry season makes the number
of El Niño is significantly noticeable when strong El Niño of burned areas on Riau much less than other regions
Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 12 of 17

with annual land and forest fire. However, the two dry category, while the other categories have a comparable
seasons on Riau also make two land and forest fire pat- impact. However, all the categories have a similar impact
terns occurred on Riau. Septiawan et al. (2019) describe on the precipitation of the second dry season. Mean-
patterns at the beginning of the year from January until while, the impact in the first dry season is inconsistent.
March and in the middle of the year from June until Figure  9 shows that the 1st fire season on Riau was
August. Therefore, the singular value decomposition and not affected by El Niño’s strength or IOD’s strength. A
empirical orthogonal analysis will be done separately to normal year and MEN-NI categories show that almost
get a better result. The first one is done from January to 40 thousand ha were burned during peak fire events on
April and May to December for the second one. In order average. The WEN-PI category has more than 20,000 ha
to make an easier comparison between both dry seasons, burned during peak fire events, while SEN-PI has less
the temporal plot for both analyses is joined. Meanwhile, than 20 thousand ha burned. This result corresponded
the spatial pattern is shown separately. The joint pattern with the dry season period before the El Niño and IOD
analysis of Riau is shown in Figs. 9 and 10. effect occurred. The highest burned area occurred in
Figure  9 shows the combined joint pattern from both WEN-PI category, which is influence by the fire’s event in
forest fire seasons on Riau. The worst condition for both 2014 caused by the cold phenomenon and the Intertropi-
dry seasons (February for the first dry season and June/ cal Convergence Zone (ITCZ) contraction to the south
July for the second dry season) is comparable with less and Madden Julian Oscillation (McBride et  al. 2015).
than 4  mm/day on average for each grid. However, the Those two events are resulting in more than 50 thou-
range value for the second dry season is tighter (3.5– sand on average during WEN-NI in February and March.
5.5  mm/day) than the first one (2–5.5  mm/day). The The second fire season on Riau occurred from May until
average monthly precipitation on Riau is 6.29  mm/day. November. Figure  9 (left) shows that the peak of the
Using 6.29  mm/day as an upper bound of dry season, events is different for each category from June to Sep-
duration of the second dry season is longer than the first tember. However, the burned area graph is likely affected
one. The second dry season has 3–4 months of less than by the El Niño and positive IOD. On average, the peak
6.29 mm/day precipitation, while the first dry season only event from the WEN-NI is around 50,000 ha, 65,000 ha
has 1–2 months. Therefore, the fire season of the second for WEN-PI, 55,000  ha for MEN-NI, and more than
one is also longer (Fig. 9a). This feature is not an effect of 80,000 ha for SEN-PI. Meanwhile, the peak fire event of
the separated analysis. The result is nearly identical when the normal year has around 50 thousand ha burned in
the singular value decomposition is applied to the whole June. Previous research states that in June 2013, the total
data. Duration of the second fire season is even longer precipitation was far below normal (of 25 years average),
when strong Niño and positive IOD occurred, resulting 145.06  mm. The precipitation deficit starts from April,
in up to 6  months of high burned area. The impact of allowing the accumulation of dry-spell (number of days
ENSO and IOD to burned area are significant in SEN-PI without drain) and causing many hotspots that lead to

Fig. 9  Temporal of joint pattern for each category on Riau with (a) represent burned area and (b) represent precipitation. The January–April graph
is obtained from the 1st dry season analysis, while the May until December graph is obtained from the 2nd dry season, which is done separately
 Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 13 of 17

Fig. 10  Spatial joint Pattern on Riau (1st ba, 1st prec, 2nd ba, 2nd prec). In a normal graph, the colour represents an average of total burned area
for each grid in the composite year only during fire season (January–March for the 1st dry season, May–November for the 2nd dry season). Other
categories represent the difference between each category with value in a normal year

the burned area in June (Kusumaningtyas and Aldrian Figure  10 (3rd row and 4th row) shows that both
2016). If we exclude the 2013 event from the normal cat- burned area and precipitation in the second dry sea-
egory, the impact of ENSO and IOD are more consistent son affected by El Niño and IOD. It can be seen that
following the strength of ENSO and IOD. the burned area increases and spread wider following
Figure  10 shows the spatial of the first joint pattern the strength of El Niño and IOD, especially when both
between monthly average burned area (ba) and precipita- phenomena occur (SEN-PI and WEN-PI). Even though
tion (prec) on Riau during the first and second dry sea- the difference is not as clear as burned area variable,
son. The odd row represents burned area variable, and the precipitation plot also shows wider area of decrease
the even row represents the precipitation variable. The precipitation in the stronger El Niño and positive IOD
spatial pattern of the first dry season (1st and 2nd row) categories. During WEN-NI and MEN-NI, some areas
shows no consistent increases value of burned area or even have less burned area and higher precipitation
decreased value of precipitation following the increase than normal year. This result shows that the impact of El
of El Niño and IOD strength. Both temporal and spatial Niño is weaker than positive IOD on Riau, which can be
analysis of the first dry season explains the negative cor- seen by comparing the WEN-PI with MEN-NI spatial
relation in Figs.  1 and 2 on both spatial and violin plots pattern. The WEN-PI gave a wider area with increases
of Riau. This result confirms the previous result that the burned area and decreases precipitation compared to
first dry season on Riau was not affected by the strength MEN-NI. This result is related to the location of Riau,
of El Niño and IOD. The higher number of the burned which is closer to a location that contributed to IOD
area occurred anomalously and more correlated with the than El Niño.
dry season condition during the respective years.
Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 14 of 17

3.3 Identifying the impact of teleconnection The yellow row in Table  2 is where El Niño and posi-
between ENSO and IOD tive IOD occurred in the same year. Table 2 shows that,
The previous section analysis shows that the peak from in general, the impact of El Niño and positive IOD that
most of Indonesia’s land and forest fire events hap- occurred simultaneously is bigger for all regions even
pened in September, except events that occurred on Riau compared with moderate El Niño. From the previous sec-
that did not affect ENSO and IOD. For the other three tion analysis, both positive IOD and El Niño affect the
regions, both land and forest fire and the dry season dry season duration. Therefore, the dry season poten-
ended in November. ENSO and the IOD are positively tially becomes even longer when both phenomena hap-
correlated (Cai et  al. 2011). Therefore, identifying the pened compared to a single phenomenon. In general, the
independent impact of ENSO and IOD that occurred in WEN-PI has higher impact than MEN-NI on both meth-
the same dry season is difficult. The correlation between ods except on Merauke. However, the value of WEN-NI,
ENSO and IOD is called ENSO-IOD teleconnection. WEN-PI, and MEN-NI of Merauke is not much differ-
In the past century, the ENSO-IOD teleconnection has ence among each other represent similar impact to the
been found to get stronger (Yuan et  al. 2018). This sec- burned area. This result reinforces previous statement
tion identifies the impact of ENSO-IOD teleconnection that IOD has less impact on Merauke which have over-
on the land and forest fire pattern in Indonesia. lapping temporal graph of burned area and precipitation
The analysis will be provided using Dynamics Time during WEN-NI, WEN-PI, and MEN-NI (Fig. 3).
Wrapping (DTW) and Euclidian distance methods. The least difference also occurred on Riau reinforced
Both methods calculate the distance between signals in the result of previous segment. Even so, the impact of
a normal year with a signal in each category to compare WEN-PI is consistently higher than MEN-NI on Riau.
the whole signal easily among each category. The bigger The big difference is occurred on Kalimantan and South
value represents the bigger difference between both sig- Sumatra. Even though Kalimantan has higher peak fire
nals. The analysis is carried out to the mean and maxi- event during MEN-NI than WEN-PI (Fig.  5), MEN-NI
mum signals to impact the average and worst conditions. impact through the whole signal is less than WEN-PI.
The mean signal contains the average value of all ana- This result related with longer fire season of Kalimantan
lysed grid for each month, while the maximum signal during WEN-PI which is ended in December (Fig. 5). In
contains the maximum value of analysed grid. Strong El line with previous result, South Sumatra is consistently
Niño and positive IOD that occurred simultaneously had affected by the occurrence of El Niño and positive IOD.
the strongest impact both on the peak and whole land Compared to normal year, WEN-PI have significantly
and forest fire signal in all calculations. Therefore, the higher value than WEN-NI or even MEN-NI. The occur-
comparison will be focused on the WEN-NI, WEN-PI, rence of positive IOD during weak El Nino can lead to
and MEN-NI categories. much severe dry season and fire event on South Sumatra.

Table 2  DTW and Euclidian distance results for each category

Mean signal (average condition) Max signal (worst condition)
×105 Normal Riau South Sumatra Kalimantan Merauke Riau South Kalimantan Merauke
Sumatra

WEN-NI 0.54 1.02 3.35 1.88 1.76 1.35 3.72 2.03

WEN-PI 0.96 7.23 10.40 1.13 2.19 7.09 12.26 1.28

DTW
MEN-NI 0.84 2.05 9.65 1.25 1.54 1.99 7.99 1.73

SEN-PI 1.73 13.70 26.71 5.50 3.74 18.76 38.02 5.94

WEN-NI 0.26 0.98 2.59 1.24 0.97 1.15 4.2 1.37

WEN-PI 0.46 4.38 6.91 0.64 1.03 4.72 10.26 0.74

Euclidian
MEN-NI 0.41 1.59 7.12 0.87 0.75 1.02 6.25 1.18

SEN-PI 1.06 8.53 15.89 3.49 2.05 11.95 23.32 3.52

 Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 15 of 17

This result indicated that ENSO-IOD teleconnec- usual, resulting in more total burned areas (Figs. 5, 7, 9).
tion has a strong impact on Indonesia’s land and forest Because of the same period of occurrence, the impact
fire signal. The effect is not always resulting in the big- of weak El Niño combined with positive IOD could be
gest peak fire event, but a bigger fire event in the whole higher than singular moderate El Niño for all regions.
dry season caused by the longer duration. This result is This result indicated the important effect of teleconnec-
supported with previous research about combined roles tion between ENSO and IOD on Indonesia’s precipita-
of the El Niño type and the IOD phase in modulating tion and burned area. The high percentage of peatland
fire activities in Indonesia, especially on South Suma- during the fire’s event makes the impact of the long dry
tra and Kalimantan (Pan et  al. 2018). They found that season caused by weak El Niño and positive can cause
carbon emissions are higher in 2006 (WEN-PI) than more problem than shorter but more severe dry season
in 2009 (MEN-NI) in both southern Sumatra (83.8 ver- caused by moderate El Niño.
sus 29.0  g·C·m−2·month−1) and southern Kalimantan
Acknowledgements
(104.8 versus 63.4  g·C·m−2·month−1), even though 2006 The authors would like to thank The Department of Mathematics, IPB Univer-
is a weaker El Niño year by 0.17 °C in terms of Niño 3.4 sity, and The Meteorological, Climatological, and Geophysical Agency for their
index. Positive IOD during weak El Niño year could fur- strong support and invaluable assistance throughout this research.

ther contribute to the drier conditions and thus more Author contributions
intensive fire activities, which are usually unexpected SN carried out project administration, funding acquisition, conceptualization,
from a weak El Niño year indicated by the relatively low methodology, validation, supervision, and writing (review & editing). FB & MTJ
carried out supervision, methodology, and validation. AS carried out supervi-
Niño 3.4 index. (Pan et al. 2018). sion, resources, data curation, investigation, methodology and validation. MA
This research is limited with the availability of burned & IF carried out conceptualization, methodology, software development, for-
area data in Indonesia which is fundamental to the cli- mal analysis, and writing (original draft). PS & MKN carried out conceptualiza-
tion, methodology, formal analysis, validation, and writing (review & editing).
mate analysis. Most of the classification only contains All authors read and approved the final manuscript.
2  years event in the last 24  years. Thus, it cannot be
guaranteed that future event will have same pattern and Declarations
characteristic explained in this research due to the lack
of sample. However, the result is still can be used to give Competing interests
The authors declare that they have no competing interests.
rough picture of dry season and land and forest fire con-
dition during specific classification of ENSO and IOD Author details
1
based on last 24  years data. Considering future rainfall  Department of Mathematics, FMIPA, IPB University, Dramaga, Bogor 16680,
Indonesia. 2 Applied Climate Information Service Center, Meteorological, Cli-
conditions in Indonesia will become drier than the his- matological, and Geophysical Agency, Kemayoran, Central Jakarta, DKI Jakarta
torical condition during El-Nino and positive IOD (Qalbi 10720, Indonesia.
et al. 2017) which can lead to more severe land and forest
Received: 15 March 2022 Accepted: 20 April 2022
fire, various research needs to be done to help anticipate
the event even if there are fundamental limitation.

4 Conclusion References
ENSO and IOD are phenomena that positively correlate Anteneh ZA, Assefa M, Wondwosen MS, Wossenu A (2019) Drought and
climate teleconnection and drought monitoring. Extreme hydrology and
with land and forest fire and precipitation in Indone- climate variability. Elsevier, San Diego. https://​doi.​org/​10.​1016/​B978-0-​12-​
sia. In terms of burned areas, the strongest impact of El 815998-​9.​00022-1
Niño is occurred on Merauke following by Kalimantan. Ardiansyah M, Boer R, Situmorang A (2017) Typology of land and forest fire on
South Sumatra, Indonesia based on assessment of MODIS data. IOP Conf
The impact is diminished when affected in the western Ser Earth Environ Sci 54:012058. https://​doi.​org/​10.​1088/​1755-a
region, with the least impact appeared in land and for- Avia L, Sofiati I (2018) Analysis of El Niño and IOD phenomenon 2015/2016
est fire on Riau (Fig. 1). Meanwhile, the strongest impact and their impact on rainfall variability in Indonesia. IOP Conf Ser Earth
Environ Sci 166:012034. https://​doi.​org/​10.​1088/​1755-​1315/​166/1/​
of IOD occurs in land and forest fire on South Sumatra 012034
and Kalimantan (Fig.  2). However, the severity of the Bayarjargal Y, Karnieli A, Bayasgalan M, Khudulmur S, Gandush C, Tucker CJ
peak event only increased dramatically when strong El (2006) A comparative study of NOAA-AVHRR derived drought indices
using change vector analysis. Int J Remote Sens 105(1):9–22
Niño and positive IOD occurred simultaneously. Weak Behera S, Brandt P, Reverdin G (2013) Chapter 15—the tropical ocean circula-
to moderate El Niño impact is depended on the charac- tion and dynamics. International geophysics. Academic Press, London.
teristic of each region. The difference between impact https://​doi.​org/​10.​1016/​B978-0-​12-​391851-​2.​00015-5
Björnsson H, Venegas S (1997) A manual for EOF and SVD analysis of climate
between weak and moderate El Niño is insignificant on data. Department of Atmospheric and Oceanic Sciences and Centre for
South Sumatra and Merauke, while it is significant on Climate and Global Change Research, McGill University, Technical Report
Kalimantan. Both El Niño and positive IOD phenomena
make the duration of the dry season even longer than
Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 16 of 17

Burton C, Betts RA, Jones CD, Feldpausch TR, Cardoso M, Anderson LO (2020) Lohberger S, Stängel M, Atwood EC, Siegert F (2017) Spatial evaluation of
El Niño driven changes in global fire 2015/16. Front Earth Sci 8:199. Indonesia’s 2015 fire-affected area and estimated carbon emissions using
https://​doi.​org/​10.​3389/​feart.​2020.​00199 Sentinel-1. Glob Change Biol 24:644–654. https://​doi.​org/​10.​1111/​gcb.​
Byron N, Shepherd G (1998) Indonesia and the 1997–98 El Niño: fire problems 13841
and long-term solutions. Nat Resour Perspect. 28 Luca Tacconi (2003) Fires on Indonesia: causes, cost, and policy implications.
Cai W, van Rensch P, Cowan T, Hendon HH (2011) Teleconnection pathways of CIFOR Occasional Paper no. 38
ENSO and the IOD and the mechanisms for impacts on Australian rainfall. McBride JL, Sahany S, Hassim ME, Nguyen CM, Lim S-Y, Rahmat R, Cheong W-K
J Clim 24(15):3910–3923 (2015) The 2014 record dry spell at Singapore: an intertropical conver-
Dafri M, Nurdiati S, Sopaheluwakan A (2021) Quantifying ENSO and IOD gence zone (ITCZ) drought. Bull Am Meteorol Soc 96:S126–S130
impact to hotspot in Indonesia based on heterogeneous correlation map McPhaden MJ, Nagura M (2014) Indian Ocean Dipole interpreted in terms of
(HCM). J Phys Conf Ser 1869:012150 recharge oscillator theory. Clim Dyn 42:1569–1586. https://​doi.​org/​10.​
Edwards RB, Naylor RL, Higgins MM, Falcon WP (2020) Causes of Indonesia’s 1007/​s00382-​013-​1765-1
forest fires. World Dev 127:104717. https://​doi.​org/​10.​1016/j.​world​dev.​ Muller M (2007) Dynamic time warping. Information retrieval for music and
2019.​104717 motion. Springer, Berlin. https://​doi.​org/​10.​1007/​978-3-​540-​74048-3_4
Fanin T, Werf G (2016) Precipitation-fire linkages on Indonesia (1997–2015). Navarra A, Simoncini V (2010) A guide to empirical orthogonal function for
Biogeosci Discuss. https://​doi.​org/​10.​5194/​bg-​2016-​443 climate data analysis. Springer, Dordrecht
Field RD, Werf GRVD, Fanin T, Fetzer EJ, Fuller R, Jethva R, Levy R, Livesey NJ, Luo Neelin J, Battisti D, Hirst A, Jin F-F, Wakata Y, Yamagata T, Zebiak S (1998) ENSO
M, Torres O, Worden HM (2016) Indonesia 2015 fire and haze. Proc Natl theory. J Geophys Res 103:14261–14290. https://​doi.​org/​10.​1029/​97JC0​
Acad Sci 113(33):9204–9209. https://​doi.​org/​10.​1073/​pnas.​15248​88113 3424
Food and Agriculture Organization (FAO) (2007) Fire management global Nicholson WK (2001) Elementary linear algebra. McGraw-Hill, Singapore
assessment 2006 Nur’utami M, Hidayat R (2016) Influences of IOD and ENSO to Indonesian rain-
Ganjam M, Sudhakar RC (2015) Geospatial monitoring and prioritization of for- fall variability: role of atmosphere-ocean interaction in the indo-pacific
est fire incidences in Andhra Pradesh, India. Environ Monit Assess. https://​ sector. Procedia Environ Sci 33:196–203. https://​doi.​org/​10.​1016/j.​proenv.​
doi.​org/​10.​1007/​s10661-​015-​4821-y 2016.​03.​070
Giglio L, Randerson JT, Van Der Werf GR (2013) Analysis of daily, monthly, and Nurdiati S, Sopaheluwakan A, Agustina A, Septiawan P (2019) Multivariate
annual burned area using the fourth-generation global fire emissions analysis on Indonesian forest fire using combined empirical orthogo-
database (GFED4). J Geophys Res Biogeosci 118(1):317–328. https://​doi.​ nal function and covariance matrices. IOP Conf Ser Earth Environ Sci
org/​10.​1002/​jgrg.​20042 299(1):012048
Hannachi A (2004) A primer for EOF analysis of climate data. Department of Nurdiati S, Sopaheluwakan A, Septiawan P (2021a) Spatial and temporal analy-
Meteorology, University of Reading, Reading sis of El Niño impact on land and forest fire on Kalimantan and Sumatra.
Harrison M, Page S, Limin S (2009) The global impact of Indonesian forest fires. Agromet 35(1):1–10. https://​doi.​org/​10.​29244/j.​agrom​et.​35.1.​1-​10
Biologist 56:156–163 Nurdiati S, Bukhari F, Julianto MT, Najib MK, Nazria N (2021b) Heterogene-
Hendon HH (2003) Indonesian rainfall variability: impacts of ENSO and local ous correlation map between estimated ENSO And IOD from ERA5 and
air–sea interaction. J Clim 16:1775–1790 hotspot in Indonesia. Jambura Geosci Rev 3(2):65–72. https://​doi.​org/​10.​
Hersbach H, Bell B, Berrisford P, Biavati G, Horányi A, Muñoz Sabater J, Nicolas 34312/​jgeos​rev.​v3i2.​10443
J, Peubey C, Radu R, Rozum I, Schepers D, Simmons A, Soci C, Dee D, Thé- Oja H, Fitriani, Samderubun G, Laode I, Maturan A, Betaubun A (2019) The role
paut JN (2019) ERA5 monthly averaged data on single levels from1979 to of indigenous peoples (LMA) in the control of forest and land fires on
present. Copernic Clim Change (C3S) Serv Clim Data Store (CDS). https://​ Merauke. IOP Conf Ser Earth Environ Sci 235:012061. https://​doi.​org/​10.​
doi.​org/​10.​24381/​cds.​f1705​0d7 1088/​1755-​1315/​235/1/​012061
Hooijer A, Silvius M, Wösten H, Page S (2006) PEAT-CO2: assessment of C ­ O2 Osaki M, Nursyamsi D, Noor M, Wahyunto, Segah H (2016) Peatland in Indo-
emissions from drained peatlands in SE Asia. Delft Hydraulics report nesia. In: Osaki M, Tsuji N (eds) Tropical peatland ecosystems. Springer,
Q3943 Tokyo. https://​doi.​org/​10.​1007/​978-4-​431-​55681-7_3
Hu X, Cai M, Yang S, Wu S (2017) Delineation of thermodynamic and dynamic Page S, Siegert F, Rieley J, Boehm HD, Jaya A, Limin S (2002) The amount of car-
responses to sea surface temperature forcing associated with El Niño. bon released from peat and forest fires in Indonesia during 1997. Nature
Clim Dyn. https://​doi.​org/​10.​1007/​s00382-​017-​3711-0 420:61–65. https://​doi.​org/​10.​1038/​natur​e01131
Kemen GA, Schwantes A, Gu Y, Prasad SK (2019) What causes deforestation Pan X, Chin M, Ichoku CM, Field RD (2018) Connecting Indonesian fires and
in Indonesia? Environ Res Lett 14:024007. https://​doi.​org/​10.​1088/​1748-​ drought with the type of El Niño and phase of the Indian Ocean Dipole
9326/​aaf6db during 1979–2016. J Geophys Res Atmos 123(15):7974–7988. https://​doi.​
Kirana AP, Sitanggang IS, Syaufina L (2016) Hotspot pattern distribution in peat org/​10.​1029/​2018J​D0284​02
land area in Sumatera based on spatio-temporal clustering. Procedia Qalbi H, Faqih A, Hidayat R (2017) Future rainfall variability in Indonesia under
Environ Sci 33:635–645 different ENSO and IOD composites based on decadal predictions of
Krasovskiy A, Khabarov N, Pirker J, Kraxner F, Yowargana P, Schepaschenko D, CMIP5 datasets. IOP Conf Ser Earth Environ Sci 54:012043. https://​doi.​org/​
Obersteiner M (2018) Modeling burned areas on Indonesia: the FLAM 10.​1088/​1755-​1315/​54/1/​012043
approach. Forests 9:437. https://​doi.​org/​10.​3390/​f9070​437 Saji N, Goswami B, Vinayachandran P et al (1997) A dipole mode in the tropical
Kurniadi A, Weller E, Min SK, Seong MG (2021) Independent ENSO and IOD Indian Ocean. Nature 401:360–363. https://​doi.​org/​10.​1038/​43854
impacts on rainfall extremes over Indonesia. Int J Climatol. https://​doi.​ Septiawan P, Nurdiati S, Sopaheluwakan A (2019) Numerical analysis using
org/​10.​1002/​joc.​7040 empirical orthogonal function based on multivariate singular value
Kusumaningtyas SDA, Aldrian E (2016) Impact of the June 2013 Riau province decomposition on Indonesian forest fire signal. IOP Conf Ser Earth Envi-
Sumatera smoke haze event on regional air pollution. Environ Res Lett ron Sci. https://​doi.​org/​10.​1088/​1755-​1315/​303/1/​012053
11:075007. https://​doi.​org/​10.​1088/​1748-​9326/​11/7/​075007 Stuecker MF, Timmermann A, Jin F-F, Chikamoto Y, Zhang W, Wittenberg AT,
L’Heureux M (2016) The 2015–16 El Niño’. Science and technology infusion Widiasih E, Zhao S (2017) Revisiting ENSO/Indian Ocean Dipole phase
climate bulletin NOAA’s national weather service. In: Proceedings of the relationships. Geophys Res Lett 44:2481–2492. https://​doi.​org/​10.​1002/​
41st NOAA annual climate. Diagnostics and prediction workshop, 3–6 2016G​L0723​08
October 2016, Orono, ME Tan ZD, Carrasco LR, Taylor D (2020) Spatial correlates of forest and land fires in
Lhermitte S, Verbesselt J, Verstraeten WW, Coppin P (2011) A comparison of Indonesia. Int J Wildland Fire 29:1088–1099
time series similarity measures for classification and change detection of Tansey K, Beston J, Hoscilo A, Page SE, Hernández PCU (2008) Relationship
ecosystem dynamics. Remote Sens Environ 115(12):3129–3152 between MODIS fire hot spot count and burned area in a degraded tropi-
Li T, Wang B, Chang CP, Zhang Y (2002) A theory for the Indian Ocean dipole– cal peat swamp forest in Central Kalimantan, Indonesia. J Geophys Res
zonal mode. J Atmos Sci. https://​doi.​org/​10.​1175/​1520-​0469(2003)​060%​ 113:D23112. https://​doi.​org/​10.​1029/​2008J​D0107​17
3c2119:​ATFTIO%​3e2.0.​CO;2 Trenberth KE (1997) The definition of El Niño. Bull Am Meteorol Soc
78(12):2771–2778
 Nurdiati et al. Terrestrial, Atmospheric and Oceanic Sciences (2022) 33:16 Page 17 of 17

Vander Werf GR, Randerson JT, Giglio L, van Leeuwen TT, Chen Y, Rogers BM,
Mu M, van Marle MJE, Morton DC, Collatz GJ, Yokelson RJ, Kasibhatla PS
(2017) Global fire emissions estimates during1997–2016. Earth Syst Sci
Data 9:697–720. https://​doi.​org/​10.​5194/​essd-9-​697-​2017
Vetrita Y, Cochrane MA (2020) Fire frequency and related land-use and land-
cover changes in Indonesia’s peatlands. Remote Sens. https://​doi.​org/​10.​
3390/​rs120​10005
Yuan D, Hu X, Xu P, Zhao X, Masumoto Y, Han W (2018) The IOD-ENSO precur-
sory teleconnection over the tropical Indo-Pacific Ocean: dynamics and
long-term trends under global warming. J Oceanol Limnol 36(1):4–19.
https://​doi.​org/​10.​1007/​s00343-​018-​6252-4
Yulianti N, Hayasaka H (2013) Recent active fires under El Niño conditions on
Kalimantan, Indonesia. Am J Plant Sci 4:685–696. https://​doi.​org/​10.​4236/​
ajps.​2013.​43A087
Yulihastin E, Febrianti N, Trismidianto (2009) Impacts of El Niño and IOD on the
Indonesian climate. National Institute of Aeronautics and Space (LAPAN),
Indonesia

Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in pub-
lished maps and institutional affiliations.