You are on page 1of 14

INTERNATIONAL JOURNAL OF CLIMATOLOGY

Int. J. Climatol. 28: 435–448 (2008)
Published online 12 June 2007 in Wiley InterScience
(www.interscience.wiley.com) DOI: 10.1002/joc.1543

Spatio-temporal climatic change of rainfall in East Java
Indonesia
Edvin Aldriana,b * and Yudha Setiawan Djamila
aAgency for the Assessment and Application of Technology, UPTHB - BPPT, Jl MH Thamrin 8, Jakarta, Indonesia
b Marine Study Program, Faculty of Mathematics and Natural Science, University of Indonesia, Kampus FMIPA-UI, Depok, Indonesia

ABSTRACT: Spatial and temporal rainfall analysis of the Brantas Catchment Area (DAS Brantas), East Java, from 1955
to 2005 based on 40 rainfall stations with monthly rainfall data derived from daily rainfall data has been performed. To
identify the climatic trend and annual changes in the area over the last five decades, we use the empirical orthogonal
function (EOF) method based on multivariate statistics, followed by the fast Fourier transform (FFT) method for the power
density spectrum analysis, the non-parametric Mann-Kendall trend test and the wavelet transform method. With EOF, we
found the monsoonal rainfall pattern as the most dominant in this area, which explains about 72% of all variances. Without
the annual signal, the leading EOF shows significant ENSO-modulated inter-annual and seasonal variabilities, especially
during the second transitional period. We found a common and significant negative trend of accumulated rainfall and a
negative trend of the monsoonal strength and dominance. This finding leads to changes in the annual pattern, which are
increase in the ratio of rainfall during the wet season and increase of the dry spell period or the imbalance of the annual
pattern. The increased ratio of the rainfall in the wet season has led to an increased threat of drought in the dry season
and extreme weather in the wet season in recent decades. The role of the orographic effect had been detected from the
decadal pattern, in which the high-altitude areas have greater rainfall amount all year round. From the decadal isohyets in
December/January/February (DJF) and June/July/August (JJA), the rainfall amount decreased significantly during the last
five decades as shown by a persistent increase of areas with low rainfall amount. By comparing the time series of rainfall
data in two locations, the mountain and coastal areas, we discovered that the dry periods have increased, mainly in the
low altitude area. Copyright  2007 Royal Meteorological Society

KEY WORDS Indonesia; Java rainfall; climate trend; ENSO; decadal; EOF; wavelet
Received 9 June 2006; Revised 7 January 2007; Accepted 17 March 2007

1. Introduction About the region, Suppiah and Hennessey (1998) and
According to the intergovernmental panel on climate later Haylock and Nicholls (2000) have studied and
change (IPCC) report (IPCC, 2001), the surface tempera- reported an increasing trend of the Australian rainfall.
ture of the earth has risen steadily since the post-industry The closest effort to this study for countries in Southeast
era of the 19th century. It is very likely that the observed Asia was made by Manton et al. (2001), who analysed
global warming has contributed to a change in rainfall trends of daily rainfall and temperature extremes of
patterns on the local or regional scale. So far, no regional countries in Southeast Asia and Australia, including
scale study has been performed over Indonesia to assess Indonesia. For the Indonesian region, they used six
the possible interdependencies between the observed rain- inadequate rainfall stations (Pangkal Pinang, Jakarta,
fall variability and climate trends over the last five Balikpapan, Manado, Ambon and Palu), from which they
decades. In their report, IPCC (2001) suggested that some found no significant trend in any of the extreme rainfall
model studies predicted a small change (below 5%) of indices in Indonesia. Moreover, for the entire region,
rainfall over the Southeast Asian region as resulted from they found that the number of rainy days (with at least
the inter-model consistency or no conclusive findings. 2 mm of rain) has decreased significantly throughout
Has the climate change in reality reduced or increased the Southeast Asia and the western and central South Pacific,
rainfall over this region during past century or decades?. but increased in the north of French Polynesia and
Such a question is very important nowadays, since during Fiji, and at some stations in Australia. Hence, to the
recent decades, pressure on demographic and population authors’ knowledge, there is no such study reported on
tension as well as agriculture and land cover change has climatic trend over local or regional maritime continent.
put some areas of the world in water deficiency. Besides, spatial assessment of the climate impact and
the trend is a relatively new topic for rainfall study in
Indonesia.
* Correspondence to: Edvin Aldrian, BPP Teknologi (UPTHB), Jl MH
Thamrin no 8, Lt 19, Jakarta 10340. Indonesia. While studies on the Indonesian rainfall regime date
E-mail: edvin@webmail.bppt.go.id back to early 20th century (Braak, 1921), more detailed

Copyright  2007 Royal Meteorological Society

Those data are man- reaches only 2. The grid system Tangkil. for our area of study. Among the different homogeneity tests described by Peterson et al. One of the of the study.7 million in 1994. Before 1991. Data and model The amount of annual average rainfall reaches 2300 mm (1991–2005 average) and about 80% of this falls during We used daily rainfall data from 40 rain gauge sta- the wet season. The analyses are conducted with is about 43. Moreover. for example. important for the future projection of the water manage- Konnen et al. The result of this will discuss the data and method of analyses followed study will be useful in understanding the possible local by some of our results. the local authority paper is the analysis of the spatial and temporal pre. Thus. we checked for missing data and data consistency with a homogeneity test. DJAMIL studies on the Java rainfall were published long after behaviour in recent decades temporally or spatially are that.5 ° E to 113. Figure 1). since 1991. (1998). In the following we climatic changes that could be assessed. Also. the interesting local areas for such a study is the Brantas wavelet method. The data from this area and spectral analysis were used. The population of aged and collected by the Brantas catchment authority the Brantas basin was about 13. 1959) gridding method to minimize the risk where the rain gauge stations are located. in the final section. conclude the The Brantas catchment covers about 12 000 km2 or highlights of our study. East Java. The Brantas river catchment area in East Java. (1998).0 ° E. factor for higher frequency from daily up to intra-seasonal this data is a valuable source of information of the local will be disregarded in this study. while the total dam capacity in the area 111. (Cressman. ALDRIAN AND Y. The climatic five decades). 2. the time series of monthly precipi- Aldrian and Susanto (2003). during the last five decades. The regard to rainfall data of 51years from 1955 to 2005. Changes in climatic Before we changed the data format into monthly. there was a consistency problem from data before and after 1991. 1979) by comparing with the nearby stations.6–3 billion m3 yearly. discuss some topics related to impacts of the global climate change. we rearranged the daily data into monthly data. The study is limited to has been collected daily for quite a long time (more than temporal analyses from monthly to decadal. methods such as non-parametric tests. about Copyright  2007 Royal Meteorological Society Int. therefore the record will measure instantaneous data and not once daily like before 1991.1002/joc . During the gridding process. population density over the basin area is about 1. The homogeneity was detected using the Craddock test (Craddock.4 ° S and from 12 billion m3 . area. a study of its period. In order to achieve the objectives decades from the real observation data. which or Perum Jasa Tirta I. The total length of the main river is 320 km with the Brantas River itself being the second largest river in Java.436 E.2 ° S–08. has modernized the data collection using an integrated cipitation variability and trend in the Brantas catchment automatic telemetry system. Haylock and McBride (2001) and ment. after 1991 used for this study.2% of the total population of East Java. previous results and. we performed a gridding process of the rainfall data in the study area using the Cressman objective analysis Figure 1. Malang and Poncokusumo.08° or about 8. at the surface and in the atmosphere. With a good spatial and temporal coverage. westward and northward before it uses a spatial resolution 0. Since the beginning of 1991. Indonesia. J.9 km2 and has turns eastward and ends in two estuaries near Porong and Surabaya 25 grid cells longitudinally and 15 grid cells latitudinally city. With such different measurement methods. Among those are 13 sta- Considering the importance of the Brantas catchment tions with continuous data record over the entire analysis area to the local and regional economy. all rain gauge stations recorded manual data from each rain gauge taken once daily in the morning (around 00 UTC). In this study. Climatol. The aim of this for this study. data have been recorded automatically. and then goes around the Arjuna and Kelud Mountains southward. by Hackert and Hastenrath (1986). The Brantas river starts near of missing data of one or a few stations. Such a Since we are dealing with the long-term climatic study is important for understanding the water potency trend. The data before 1991 on several stations were adjusted according to the level of data after 1991 to preserve homogeneity of the monthly data. 28: 435–448 (2008) DOI: 10. S. this method is the most feasible and easy to implement. Hence. we have in total 40 rain gauge stations long-term climatic trend is essential. Total potency of surface water reaches tions over the catchment area (07.5 times There are 27 rain gauges before and 26 rain gauges the provincial average. about 35% area of the East Java Province (Figure 1). we found a problem of data before and after 1991 in several stations. the empirical orthogonal function (EOF) catchment in East Java province. It is interesting to see the tation data of 40 rain gauge stations between 1955 and past changes of the rainfall pattern during the previous 2005 were analysed.

On the stations for monthly as well as for annual rainfall trends. as mountain areas always have a larger amount of rainfall ested in the annual trend of rainfall. or coastal area up to the mountainous area. Results and analyses transform (FFT) methods. The preserve a high rainfall amount all year round and even method was applied on all continuous station data. After looking at look at the isohyet (rainfall iso-contour) of each data the annual spatial pattern. Mann. we are inter. sponding statistical significance using the Mann–Kendall the increase of rainfall from the dry to the wet season trend test. the mountainous areas receive more rainfall and. then we calculated the linear trend as well as the corre. an taneously. we were interested to see the group and compare it with the whole time series (inter. Thus. The method will calculate or about 150 mm/month. Wilks. (Preisendorfer. The wavelet analysis (Torrence behaviour in the Brantas catchment area with high- and Compo. in fact. computing the FFT at each time using only the as shown in Figure 2 clearly shows the monsoonal data within the window. This test is performed on stations with no loca. decadal groups and as a whole. Data from the stations 1995. otherwise the EOF the dominated spatial pattern. By a definition decompose a time series into time/frequency space simul. first EOF (PC) is investigated with respect to its temporal parametric Mann–Kendall trend test and the spectral variability. J. The high mountainous area seems to from information after using the wavelet analysis. 1975–1984. Isohyets are spatial lines showing the same Then we chose December/January/February (DJF) and amount of rainfall in a certain area. there are 13 wet season. The second maximum value occurring during the peak of the wet advantage of the test is its low sensitivity to abrupt breaks season from December to February. 1938. which is derived than lowland areas. 1955–1964. analyses of the wavelet transform. Such a dry season from June to August. The time series associated with the were used for individual trend analysis using the non. 1998) works furthermore by attempting to intensity rainfall from October to April. high humidity from the South China Sea at the time of tion change during the last five decades. and focus on spatial trend analyses. The EOFs method will not work. we managed to fill all grid analyses on each data group in order to obtain dominant cells homogeneously for the period from January 1955 spatial patterns as well as their long-time behaviour and to December 2005. During the wet season.1002/joc . 1970) test is a non-parametric test. known locally as the West Monsoon. Copyright  2007 Royal Meteorological Society Int. we use several trend analysis methods derived from the time series of the data field (von Storch. We will to dryness than the mountainous area. in dry season the wind brings dry air from The wavelet transform is an extension of the Fourier the Australian continent in a period known locally as the transform method or a running (windowed) FFT method. Annual pattern and inter-decadal variability Kendall. with the and the frequency of the circulation forms. spatial trend changes occur during the past five decades. over mountainous areas is more than those over low- In order to classify the climatic trend for the last land areas. which brings moist air with analysis. Moreover. 1995).1. But the The spatial rainfall pattern of the Brantas catchments time series used in this study have a normal distribution area for the monthly average data clearly indicates the with some exceptions such as in the case of precipitation 1-year periodicity of the monsoonal pattern. The wavelet anal. from the Indonesian Bureau of Meteorology (BMG). The amount of due to an inhomogeneous time series.e. to April is considered as the wet season. The rest of analyses were performed under gridded data for trend analyses using a combination of the EOF and the fast Fourier 3. 1965–1974. Climatol. 3. This pattern follows test will be performed on all trend analyses in this study the Southeast Asian monsoonal wind pattern that also to derive significant level and confidence on the trend passes over the catchment. we ysis is calculated using the Morlet mother wavelet for also see that the rainfall contour follows the orography. Thus. SPATIO-TEMPORAL CLIMATIC CHANGE OF RAINFALL IN EAST JAVA 437 0. 1989) after the gridding process. are defined as the eigenvectors of the covariance matrix In this study. is more susceptible the whole data 1955–2005 (inter-annual data). This test is used rainfall decreases to a minimum during the peak of the to analyse the statistical significance of all trends. After we classified data into their them with their respective decadal monthly averages. and during the peak of the dry season. annual). especially the coastal area. we will June/July/August (JJA) as months representing the wet have the average spatial pattern of each decade and and the dry seasons. respectively.18% of total grid cells were missing and we replaced the total time series. The propagation of the wet area on the on-set five decades. The Mann–Kendall (Kendall. Hence. using a certain window size and sliding it along in The average annual rainfall pattern (1955–2005) time. we separate monthly data into decadal of the wet season and its offset begin from the lowland data of five groups. respectively. i. The latter condition is essential to frequency distributions of the principal components of run the EOF method (explained later). East Monsoon. we notice that November the power spectrum of the data series. and how this period passes with a rainfall amount exceeding 50 mm amplitude varies with time. 28: 435–448 (2008) DOI: 10. One gets information on both the amplitude above-normal or wet month is declared when a 10-day of any ‘periodic’ signals within the series. which does not require the data to be distributed normally. other hand. In this study. the power and global spectra. we performed the EOF Although the data period of each station is not the same. 1945. the 1985–1994 and 1995–2005 and one dataset covering lowland. 1988) and FFT (Press and Rybicki.

8E 112E 112.8E 112E 112.8E 112E 112. J.8E 112E 112.2S 7.6S 8S 8.8E 112E 112.8E July August September October November December 10 50 100 150 200 250 300 350 400 450 500 Figure 2. Climatol. in recent years.8E 112E 112.2S 7. Besides. Using monthly and annual trend tests are given in Tables II a similar analysis as in the monthly basis.8E 112E 112. Those spatial patterns show a greater greater rainfall intensities over the mountains during the expansion of the dry area in the later decades. DJF and JJA.2S 7. indicates the decrease of the yearly amount of rainfall for the last five decades.8E 112E 112.4S 112E 112. Results of areas are still receiving a high amount of rainfall. we will investigate mountainous areas. S.8E JJA 1955/64 JJA 1965/74 JJA 1975/84 JJA 1985/94 JJA 1995/05 10 50 100 150 200 250 300 350 400 450 500 Figure 3. Thus.8E DJF 1955/64 DJF 1965/74 DJF 1975/84 1985/94 DJF 1995/05 7. 28: 435–448 (2008) DOI: 10.8E 112E 112. DJAMIL Rainfall Monthly Average (1955−2005) 7.8E 112E 112.8E 112E 112.438 E.2S 7. The pattern shows that such 3.8E 112E 112. on the basis of the information Figure 3 illustrate a regular trend on certain months like of the last two paragraphs. It is apparent from Table II that declining trend on a decadal scale occurs in December almost all stations have a decreasing trend except Semen.6S 8S 8.8E 112E 112.8E 112E 112.4S 112E 112. The decadal average changes of seasonal rainfall values in DJA and JJA as represented by the wet and dry seasons over the Brantas catchment area. the decrease happens during the trend of each station data using the Mann–Kendall the dry as well as the wet period. ALDRIAN AND Y. The annual rainfall pattern (mm/day) of the Brantas catchment area showing regular monsoon session that consists of dry and wet seasons.8E 112E 112. the strongest and III. The decadal spatial pattern changes as shown in and May (Table I).8E 112E 112.6S 8S 8. DJF and JJA Rainfall Monthly Average 7. as we observe that the mountain wavelet method will be used for comparison. respectively.1002/joc . we expect. Then the decrease in the lowland. but with a larger trend test directly on monthly and annual data.4S 112E 112. which wet season and longer dry period over the lowland.8E 112E 112.8E 112E 112.4S 112E 112.8E January February March April May June 7. Individual station climatic trend phenomena occur in the lowland as well as in the Before looking at the spatial analyses.6S 8S 8. Copyright  2007 Royal Meteorological Society Int.2.

32 −0.33 – −0.81 76.05.58 – 0.05 −1.00 −1.36 Wates Kediri −0.001.59 −15.43 −1. 0.43 – – 1.60 Average −0.31 ∗ −15.85 strong influences from large climate phenomena in the Birowo −2.02 −16. there is no significant trend during the Kediri −2.20 Kertosono −0. J.54 −1.71 – 6. Station name Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Tangkil −3. Period 1955–1964 1965–1974 1975–1984 1985–1994 1995–2005 DJF 368.91 −1.52 −17.86 – −1. April and May or the transitional Dampit −0.52 50.47 0.12 −0.45 −0.12 which is located between two southern mountains and Table III.60 −1.06 −1.25 3072.13 −0. there is no supporting data from nearby stations to the formula: Annual rain = m × year + b.59 0.23 −1.79 2642.04 −1.76 −3.01.43 −0.12 Copyright  2007 Royal Meteorological Society Int.99 −0.38 −0.36 Doko −1. bold and bold with a dark background represent significant levels at 0.17 – −1.37 3176. Table II.00 – −6.1.66 50.16 −0.13 −1.40 −0.67 320.35 −1.70 0.40 −0.44 2737. and the lowest decrease in August. 0. The Station name Mann–Kendall Linear regression trend during August is mostly undetected because of the inconsistent trend all over the study period and Z Signific.10 0.86 −2.63 −0.00 (March to May) is more stable with a gradual and Semen 0.50 0.95 −0.41 0.04 −0.63 0.50 ∗∗ Furthermore.11 −1.41 −0.83 −0. Wates Kediri −2.02 persistent decreasing trend than the other transitional period.62 2388. 28: 435–448 (2008) DOI: 10.38 −1.36 May 154. 0.02 −1.51 −1. Linear regression of the long-term (51 years) annual also receives shadow rainfall in dry and wet seasons. m (mm/year) b (mm) because of the large variability possibly from external ∗∗∗ Tangkil −3.89 −1.12 JJA 81.36 2336.23 2160.70 −2.41 −0.18 −0.38 −1.44 Pujon −0.19 −0.92 Poncokusumo −1. the period) than the extension of the wet period to the first largest decrease occurs in May followed by April or transitional period: while for the first transition period the boreal spring.84 ∗∗ and the wet season. which shows large inter-annual variability.66 −0.45 −0.22 44.33 −3.59 −2.04 35.26 386.56 – 0.1002/joc .1.34 0.92 Wagir 0.55 −1. respectively.94 – – – −0. where there is no consistent trend over the whole period.73 −0.56 Semen 1. Aldrian and Susanto (2003) classify this region Poncokusumo −2.04 −0.79 299.47 −18.21 0.71 −1. ∗∗∗ that have similar characteristics.14 1.71 −2. The Mann–Kendall trend test Z-values for monthly rainfall. where the positive (negative) means an increase (decrease) in trend over 51 years.83 2479.01 −21.79 −0.16 1.30 86.67 −2.23 −0.45 −0.11 −2.22 −0.36 −1. Figure 4 illustrates what exactly occurs with the the possibility of the extension of the dry period to the Tangkil station with the most significant monthly trends second transitional period (extension of the dry spell and explains the above hypothesis.18 Dampit 0.44 −1. there is no wet spell period but only a steady or stable September and October.06 −1.05 −0.37 ∗∗ transition period (August–November) between the dry Kertosono −3.12 −0.05 and 0. respectively.94 2405. SPATIO-TEMPORAL CLIMATIC CHANGE OF RAINFALL IN EAST JAVA 439 Table I.94 −0.13 326.05 Wates Sawahan −0.39 −0.69 0.12 −0. because it receives Wagir −1.29 ∗ −17.90 −24.44 −1.49 −1.17 −2. The Mann–Kendall trend test is not applicable to the same cases especially in August. The linear regression was calculated according However. ∗ .18 −1.18 −0.49 −2.01 −2.48 ∗∗ Pacific whose impact will be most pronounced in JJA.79 −2.46 −0. Climatol.47 175.07 0.00 −0.22 −0. On the other hand. Moreover.27 −0. 0.60 – – 0.54 −1. many of significant Tugu −3.+.31 −2.00 −1.85 −2.11 0.17 −0.50 0. The instability of trend from August to November indicates trend. The latter two ∗ Pujon −2.53 0.42 −2.89 −0. Tangkil has the most significant decreasing trend.43 3316.33 1. On the contrary.00 369.35 −1.38 −0.81 – −0.32 −0.74 Kediri −0.92 2432.22 −0.06 1.42 −1.46 −1.42 268.68 forcings.80 −1.91 0.51 −0.05 Birowo −0.61 −1.64 – −10.74 −1.56 December 359.98 values occur in March.001. ∗∗ .04 −2.94 360.33 −0.40 −1.32 −0.15 0.76 −0.98 Tugu −2.70 Doko −0.82 −2.43 −0.45 period between the wet and dry period.08 −1. most of the indicate Mann–Kendall significant levels of 0.62 2576.43 – – −0. The decadal area average of seasonal and monthly trends (mm/month).68 as the southern monsoonal region. rainfall trends.73 −1.38 −1.07 −1.63 −1.15 0. Underlined.05 −0.48 3161.58 −12.00 −1. increase in trends are insignificant. On average.20 148.46 −0.87 327.87 facts indicate that the boreal spring transitional period ∗ Wates Sawahan −2.27 −1.06 −0.23 0.82 −1.83 −1.

has many missing data and therefore of each eigen map.3. All stations The result of the EOF analyses for six data groups except Semen have decreasing trends.19 – accumulation regardless of the frequency. the trend ∗∗∗ analysis on the annual frequency will correspond to Tangkil −0. respectively.28 above-significant levels.00923 −0. this station. which most significant as well as the largest decrease among always explains more than 64.9 + Pujon −0. Copyright  2007 Royal Meteorological Society Int.00848 −1. Table IV. with results quite similar to Wates Kediri −0. this Baliunas et al. (1997) to investigate trends in the central eigen map contour follows the land morphology contour. This means that. The EOF–FFT analyses mulated rainfall trend as shown in Table III. The eigen map climatic trend over the Brantas catchment using the of the first pattern (PC1) of the inter-annual data in wavelet method. However. From Figure 5 (right panel). DJAMIL Tangkil St (395m) 12 10 600 8 Month 400 6 4 200 2 0 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 Rainfall (mm) Year Figure 4. not the wet season. The wavelet method indicators of the monsoon signal Here we present a slightly different application to detect weakening.72 direct Mann–Kendall test we look at the total annual Dampit −0. (2001) to analyse variability in European temperatures. the dry season.01467 −2. The monthly versus annual contour plot of Tangkil rainfall showing the stability of the boreal spring in contrast to the instability of the boreal fall.01457 −1. the annual signal is the most pronounced Station Linear Mann–Kendall Mann–Kendall one and in fact the only one above the threshold of name regression Z significant 5% significance level. order of pattern according to the principal component In this study we present another method for detecting and insignificant spectral intensities.18 is tabulated in Table IV. trend in the long-term rainfall data. 0. quite a substantial amount.01865 −3. this area is gradient dominated by the monsoonal signal.01068 1.1002/joc .01369 −2. ∗ . while Semen has an increasing trend. which indicates detecting climate trend has been used in many studies. The trend Doko −0. the power spectrum of the principal component data as well as Doko. for the second and the third eigen maps do not seem to have any largest. the over the area. The decreasing trend of corresponds to the annual or the monsoonal pattern.92 – those from the direct application of the Mann–Kendall Kediri −0. Climatol.1. The use of the wavelet method for Figure 6 always shows positive values. 0. From that table we notice the persistent the monthly trend analysis as in Table II.55 ∗∗ the most significant trends in the data. J. ALDRIAN AND Y. but the about ten times more than other eigen maps.0006a −1.440 E. respectively. Tangkil has the dominance of the first eigen vector (eigen map). S.76 + analysis using the wavelet method has a more significant ∗∗ Wates Sawahan −0. 3. The eigen values in the mountain areas are higher than England temperature series. 28: 435–448 (2008) DOI: 10. This method is different in the sense that we Kertosono −0. Similar analysis is applied for the total annual accu. Similar to in Table V. + .05. which is the Poncokusumo −0.01. ∗∗ . Therefore. In the total 51-year period for the largest power spectrum intensity from PC1 in the annual signal and the lowest decrease.03699 −0. between mountainous and lowland areas.02536 −3. Park and Mann (2000) to detect shifts in global temperature and Datsenko et al. In fact. we find that the first eigen map their results are quite unreliable. (2004) used the inverse wavelet to the monsoonal variability with some local differences transform to calculate trends of the Czech temperature.89 ∗∗∗ are only looking at the annual signal. the annual rainfall will contribute to the total rainfall loss which is the most dominant climate pattern.01252 −2. ∗∗∗ indicate Mann–Kendall significant level of 0.001. that the whole Brantas basin has synchronized patterns For example. Results of applying the wavelet method ∗ Birowo −0.02597 −4. From lowest decrease occurs in Dampit.56 – monsoonal trend. while with the Tugu −0. will be likely extended to the boreal fall transition period. 0.44 ∗ ∗∗∗ trend test.93 level than its counterpart and has different stations with Semen 0. Moreover. Thus.75 Wagir −0. and 9 out of 13 as defined above (‘Data and Method’ section) is given stations have significant level of their trends. The between 1236 and 62 mm.8% of all variances or all stations.00610 −2. Pisoft et al. there is an annual rainfall loss reaches about 6 times the second largest signal. many missing data. which is. One possible conjuncture for the latter phenomenon is our choice to analyse the annual a The results of Doko station is quite unreliable since there are too frequency only.

018652 0 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 Time (year) Figure 5. we present two outcomes explains about 72% of all variances (the inter-annual of applying these methods. Table V. Climatol. where it dominance or percentage.79 3. From the FFT power spec- turns out that 1-year is the most dominant period exceed. Period → 1955–64 1965–74 1975–84 1985–94 1995–05 1955–05 Eigen map 1 (%) 73.1002/joc . where the annual peak is above the significant threshold curve (dashed curve).57 5. the PC1 power spec- data). Here. i. as shown in the right figure.27 Eigen map 2 (%) 5.90 71. The analyses of the power density spectrum trum after applying the FFT method and the variance (Figure 6 right panel) also confirms this result.76 6. SPATIO-TEMPORAL CLIMATIC CHANGE OF RAINFALL IN EAST JAVA 441 (a) Kertosono Station Monthly Precipitation (1955−2005) Precipitation (mm) 1000 500 0 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 Time (year) (b) Monthly Precipitation Wavelet Power Spectrum (mm2) (c) Global Wavelet Spectrum 4 0. change.20 64.Gradient = −0.55 Eigen map 3 (%) 3. The 1-year period of the wavelet power spectrum is then extracted (bottom figure) and calculated for its linear regression trend.81 65.52 72.77 3. we used the PC data for their spectrum analy- intensity contrast between the dry and the wet seasons ses through the FFT method and investigated the strength over the mountain areas is higher than that over the of their components.08 those of the coastal areas. we obtain the monsoonal strength of each sec- ing ten times the others.77 4.e. Annual pattern has the most significant power spectrum. This condition explains the previous explanation decadal dataset.23 3. the most coastal area. As we understood before. J.5 2 Period (years) 1 2 0 4 8 −2 16 −4 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 0 1 2 3 Time (year) Power (mm2) × 105 (d) One Year Period Wavelet Time Series 6 Power (mm2) 4 2 Reg. An example of the application of the wavelet method in detecting annual trend changes.10 4. After the principal component tion or sub-period. Then. Decadal and inter-annual variances of eigen maps.29 72. This contrast comes from the persistent dominant pattern is the annual or monsoonal pattern. In orographic effect during the dry period in the mountain Figure 6 we show the spectrum of the first PC of the area.87 4.38 5. Full trend information in 13 continuous stations is given in Table III. trum. 28: 435–448 (2008) DOI: 10. which explains why the rainfall analyses.89 5.25 0. which lies in the center. The two side curves in the wavelet spectrum map indicate the border of the significant region. we used the combination of the two methods The first principal component from the EOF method to detect the climatic trend in the annual signal or the yields yearly variability or the monsoonal pattern that monsoonal dominance. as represented by the power density Copyright  2007 Royal Meteorological Society Int. there is no other about the orographic effect in mountain areas and that clear pattern in lower (above the 1-year period) or higher the coastal area is more susceptible to the annual pattern frequencies (below the 1-year period). Beside the 1-year period.

data length of each time series as 10 years starting from Hence. indicate the trend of the mon- sists of a 10-year time series with a certain central year soonal dominance.5 5 −3 8. Since the second leading EOF as the trend signal for a gridded eigen map 1 has a significant annual signal (the mon.5 −2 10 −2. The inter-annual eigen map of PC1. then.1002/joc .. we can consider that The results of applying the combined EOF–FFT this eigen map exclusively signifies the monsoonal dom. PC2 and PC3 along with their corresponding spectra after FFT. spectrum at the annual frequency. Eigen map 1 has similar positive sign all over. Climatol.5E 113E 0 1 2 3 4 5 6 7 8 9 10 11 12 Period (year) Eigen Map (1955−2005) for PC2 7.6S 1. soonal signal) above the others.5E 113E 0 1 2 3 4 5 6 7 8 9 10 11 12 Period (year) Eigen Map (1955−2005) for PC3 7. or the variance. S.6S 1.5 8S −1 15 −1.5 −2 10 −2. and a strong annual signal. Their respective variances are given in Table IV.5 8S −1 15 −1. In preparing the second approach. J.5 5 −3 8.2S 3 35 FFT PC2 1955−2005 2.5 1 25 Intensity 0.5 −2 10 −2. where for reasons of inance.5 2 30 7. On the contrary.5E 112E 112. only decadal values of the power spectrum are Copyright  2007 Royal Meteorological Society Int. Thus in total we have about 46 time series. Dai et al.5 20 −0.2S 3 35 FFT PC3 1955−2005 2.5 20 −0. DJAMIL Eigen Map (1955−2005) for PC1 7. changes in this value over time will correspond 1955 onwards. which indicates a homogeneous monsoonal influence. to changes of the monsoonal strength. method is illustrated in Figure 7. therefore.6S 1. 28: 435–448 (2008) DOI: 10. We applied the EOF–FFT method for decadal conducted a windowed or running EOF method with the sub-sections and calculated their power density spectra.5E 112E 112. We shall call this method. the running EOF variance sure of the monsoonal dominance is in the monsoonal method.5E 112E 112.2S 3 35 2. Changes of variances from these running eigen clarity.5 1 25 Intensity 0.5 8S −1 15 −1. maps in time. we (Figure 7).5 1 25 Intensity 0. The second mea.442 E. global precipitation analysis. ALDRIAN AND Y.4S 0 111.5 FFT PC1 1955−2005 2 30 7.5 20 −0.4S 0 111.5 2 30 7.5E 113E 0 1 2 3 4 5 6 7 8 9 10 11 12 Period (year) Figure 6. Each sub-period con.4S 0 111. (1997) used directly contribution to the whole pattern.5 5 −3 8.

right ENSO cycles in the west Equatorial Pacific anomaly y-axis).5 Annual PC1 −2.5 −1. Mann–Kendall and wavelet) is that the result to aSST deviation in this period are high in comparison of EOF–FFT method stands for the whole domain or for to other periods. the strength of the monsoonal pattern second leading EOF. The trend of the monsoonal dominance from power spectra of the indications of significance. we now examine the mode of the global EOF pattern.001.01). SPATIO-TEMPORAL CLIMATIC CHANGE OF RAINFALL IN EAST JAVA 443 Monsoonal Dominance from PC1 using EOF analyses (Kutzbach. Another proxy of quasi-decadal ENSO no spatial variation.119 Variances (%) 80 of the temporal and spatial patterns of the EOFs.79 with a significant level below 0.2 and -0. 1967..0. We performed EOF analyses to the 0 0 leading annual and seasonal modes. (1997). found the global rainfall trend from the second leading Following Dai et al. This conclusion is not clearly periods as the rainfall ENSO relationship above. The gridded anomalies are first normalized by their standard deviations prior to 10 20 the EOF analyses. the variance percentage gives a clearer picture. The region (150 ° W–90 ° W.1002/joc . in ° C) for the PC1. Hence. correlations decline. value of 0. 1974–1988 and 1989–2005.6 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 Figure 8. 2006).0 0. the decadal variability although does not have similar ing the last five decades. The strength. ratios of rainfall deviation others (i. The leading annual eigen coefficients of PC1 and PC2 (black lines) along with the NINO3 aSST (grey lines.0 −0.0 PC1–Coeff 1.4 PC2–Coeff 0.5 0.e. respectively. of annual precipitation anomalies are shown in Figure 8. The linear regression trend line (bold straight line) for the sea surface temperature (aSST) in the so-called NINO3 variance percentage is given by the formula shown in the graph. The monsoonal trend is represented The time component of EOF 1 of annual modes.365 (p < 0. This seen from the information of the eigen map variances result is different from that of Dai et al.001). Each EOF was components of the first of two leading EOFs (EOF 1) performed on a 10-year basis with the central year given in the x-axis.0 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 0. the linear straight grey lines highlight the bi-decadal variability at level 0.4 Annual PC2 −0. who alone but also from the FFT analyses of their PCs. the PC1 power spec. 0. In fact. Copyright  2007 Royal Meteorological Society Int. their shown along with smooth running EOF variances of PC1. Before and after this period.18 during the period 1955–1973. left y-axis) and the variance percentage of PC1 (line. In other regions. 28: 435–448 (2008) DOI: 10.0 −1. rainfall and ENSO there is a quasi-decadal relationship. Furthermore. The stability of the 1960 1970 1980 1990 2000 modes in EOF analyses of subsets of the dataset was Central Year also examined with some statistical measures as one Figure 7. ing trends. exhibits over the Brantas catchment is significantly reduced dur. Wallace et al. 100 1993) after removing the annual signal.2 0. The time and space and variances of PC1 from the moving EOF method. For the PC2. 5 ° S–5 ° N) with a correlation Mann–Kendall trend test Z-value for this is −5..5 NINO3–aSST 1.9% of the total variance. we 30 discuss the likely physical processes behind the leading 60 20 40 modes of precipitation fields. J.2 −0. which here with a declining trend of the power spectrum after FFT (bar accounts for 49.29% of total variances. replicates the graph. climate variability has been observed from a tree-ring trum also suggests a decadal variation of the monsoonal record of central Java (D’Arrigo et al.0 −0. On the basis Spectrum Intensity 40 y = –0. The eigen value 1 is considered as monsoonal with a significant annual pattern from its power spectrum.2438x + 75.79 (p < 0. The correlation value and its significance increase between 1963 and 1982 to about 0. 8. between the Brantas basin Although both methods agree on indications of decreas. The temporal and spatial structures of precipitation fields changes in ENSO could not account for the significant 2. (1997). and the possible ENSO impact for this period is the One disadvantage of the both methods compared to the largest in this area. Hence. Climatol.

ALDRIAN AND Y. ∗∗∗ SON 66. DJAMIL decrease over southern Chile. whereas a larger decreasing trend the ENSO signal is present within the first leading EOF in occurs mostly during the latter. that the first period in the modes in global precipitation fields. the inter-annual variabilities are large. especially the annual trend.177 0.01.554 0. The mountain area is represented by a grid near comes from the ENSO influence.70 7. Thus.74 8.85 5. ∗ . and therefore their ENSO relationship (by correlation values) is given a comparable decrease between highland and lowland in Table VI.01). we also found instability of the decided to investigate more on the dry period and second transition period. the mountain area and the coastal areas. Pujon (>1000 m. there is an increasing trend of the ratio 0. Comparing between two that in some years. ∗∗ .280 0. DJF 29. the rainfall rainfall and the Equatorial Pacific aSST. The summary amount and data variances over the mountainous areas of the annual and seasonal leading modes along with are larger than those over the lowland areas. Climatol. Since as an ENSO index. As shown in Figure 9. then the ratio of rain falls in the wet season to present within the JJA leading mode (p < 0. 0. Like the total and also the number of the dry months (dry the annual leading mode. season or have 100% ratio. the whole rain fall during the wet transition periods. they will be subjected to similar Anomaly monsoonal exposure.81 ° S).05 0.85 0. Furthermore. Kelud) Mountains.60 5.444 E. If the rate of decline of September/October/November (SON) with a significant the rainfall in the dry season exceeds that of the wet correlation 0. we variability. 112. J. 7. 2007 the station height. Variances of leading EOF after removing the annual (<100 m. Less significant signal is season. However.18 −0. we do not derive our comparison dominant influence for the annual variability as well..365 0. We used a threshold value for PC1 PC2 PC3 PC1 to aSST variances an extreme or total dry month of 5 mm/month rainfall in two locations.56 7. (2007) concluded that between total rain fall in the wet season to the total annual the SON rainfall over south Indonesia could be used rainfall. MAM has less variances than SON. Aldrian et al. this result is in orography. with this type of annual variability of the first leading mode of seasonal trend detection approach we rarely see a significant trend precipitation correlates well with the ENSO signal in level.. 7.05). the inter. we notice amount. it is likely to miss boreal spring is more stable than the one in the boreal many regional modes.001.003 0.48 ° E. From Table III we notice that there is an imbalance analysis. which the latter indicating more unstable inter-annual On the basis of all the previous explanations. Semen have SON and followed by JJA.455 is assumed as a no rainy day month (extreme dry ∗ JJA 73. while March/April/May an increasing trend of the ratio. and the result can be seen in Figure 11. From Figure 10. October to March) and that over the dry period (from EOF analyses of seasonal precipitation revealed that April to September).656 Mojokerto had more number of extreme dry months. Moreover.554 (p < 0.470 Thus a month with rainfall amount of less than 5 mm/day MAM 47. 112. The increase of the rainfall ratio such as this implies the The variances of anomaly data are calculated after increasing trend of the dry spell period and the increasing normalizing by their standard deviation.45 4. Note that the variances of anomaly data areas still implies susceptibility of the lowland areas.4. respectively. whereas the observed trend 3. Nevertheless. the three largest decreasing trends agreement with that of Aldrian and Susanto (2003) by m values (the trend’s slope) occur in stations below for the seasonal correlation strength between Indonesian 1000 m. again. Moreover. and the coastal area is represented by a grid near the Mojokerto city Table VI.20 9. this method may not be appropriate between the decreasing trends over the wet season (from in detecting regional climate trend. Therefore.732 month). all stations except. which turns to be the For this purpose. Therefore during risks of drought in the dry season as well as flood and the peak of the dry period (JJA) with small rainfall landslide in the wet season. which is clearly dominant in our fall. we are interested to changes in ENSO relationship (Haylock et al. sensitive to the ENSO variability. there seems (MAM) has the most insignificant correlation owing to be no relationship on the rate of change in the ratio to to the persistent spring barrier (Aldrian et al. on the basis of station data but from the gridded data Thus. see what happens to the annual pattern. and no shadowing or Fohn effect Variances (%) Correlation series will be taken into account. 2006).001). the correlation of SON rainfall spells) will increase. We already know While the global analysis can capture the most important by both transitional periods. ∗∗∗ indicate correlation at the northern side of the Kelud and Arjuna (north of significant to the level of 0. there is no role of and references therein).56 −0.1002/joc . Both stations are located signal along with their correlations to aSST of NINO3 region on a similar phase of the eigen map (positive side) and and the data series variances.05.47 ° E.16 12. or in other words.65 (p < 0. ∗∗ Annual 49.37 ° S).53 0. Here. 28: 435–448 (2008) DOI: 10. instability of the second transition period mainly instead. series are highly correlated to the leading EOF variances. S. 0. the number is large. as mentioned earlier. The seasonal imbalance in Brazil towards wetter conditions in the southwest and After using several analysis methods for detecting the drier conditions in the northeast could be explained by trend.264 which increased for the last five decades in comparison Copyright  2007 Royal Meteorological Society Int. as shown in Figure 10 and in Table VII.29 6.90 8. SON variability is highly compare the condition of rainfall data from two different nearby locations. to aSST increases between 1963 and 1982 to about Actually. Previously.

Curve on each graph is polynomial regression on the fifth power.0 −0. 1960 1960 DJF Dry month period Mojokerto (lowland) JJA 10 9 1955 1955 8 2. SPATIO-TEMPORAL CLIMATIC CHANGE OF RAINFALL IN EAST JAVA 445 100 90 80 70 Ratio (%) 60 50 y = 0.0 0. The leading seasonal eigen coefficients of PC1 (black lines) after removing the annual signal along with the NINO3 aSST (grey lines. both had a similar number of dry months. Climatol. which is about 2 months.0 1.0 −1. Two examples of increasing trends of ratio of the rainfall in wet season to total annual rainfall during the last five decades.0 7 months 6 5 2005 4 2005 NINO3–aSST NINO3–aSST 3 PC1–Coeff PC1–Coeff 2 1 2000 0 2000 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 Year 1995 1995 10 Dry month period Pujon (highland) 9 1990 1990 8 7 months 6 5 1985 1985 4 3 2 1 1980 1980 0 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 1975 Year 1975 Figure 11. 1965 1965 to Pujon which is more stable.0 1.0 0.0 1.1002/joc . 2000 2000 0 1955 1965 1975 1985 1995 2005 Year 1995 1995 100 90 1990 1990 80 70 Ratio (%) 1985 1985 60 y = 0. in the last 10 years −1.0 1.5 MAM −1. the Pujon area has a more stable number of dry months of about 1–2 months for the last 10 years with a maximum of 6 months.1 40 30 2005 NINO3–aSST NINO3–aSST 20 PC1–Coeff PC1–Coeff 10 Birowo (195m) Figure 9.5 0.0 2.5 1.0 −0.0 a record of 8 months of dry season. This result supports the previous explanation about the orographic rain that occurs in the mountainous area.0 −2.5 −2. J.0 0.5 1.0 −0.69 50 40 1980 1980 30 20 1975 1975 10 Wates Sawahan (620m) 0 1955 1965 1975 1985 1995 2005 1970 1970 Year 1965 1965 Figure 10.1614x − 241.0 −0.5 SON the number increased to 4 months and in 2002 it reaches 1955 1955 2. Meanwhile. which appears to be the longest dry season for the whole five decades. Number of extreme dry months (<5 mm) for Mojokerto 1970 1970 (top) and Pujon (bottom) from 1955 to 2005. and reveals that the monsoonal Copyright  2007 Royal Meteorological Society Int.5 −1. 28: 435–448 (2008) DOI: 10.0 0.0 −1.5 −1.5 0. in ° C). However. In the early years of 1960 1960 investigation.0 −2.5 1.5 −1.1994x − 315.5 −1.5 −2.0 2.5 0.5 0.5 1.

The critical SST over the Indonesian maritime continent may also be responsi.4772 1. Roder. the global climate change 0. surface warming should increase the surface evapora- 2001. a higher SST will reduce the Brantas catchment area. The global dimming phenomena may not be the only phenomenon respon.45 ± 0. which has dence of increasing water vapour in the troposphere since been reduced persistently during the last five decades.78∗ 6.65 – Tugu 18 0.e.7571 1. though amount received by the surface of the earth. However.11 −1. we introduced new methods of the use of rainfall as well as evaporation will be reduced since the annual spectrum of the wavelet and EOF vari.61 – Wagir 1106 0. 28: 435–448 (2008) DOI: 10. ∗∗ . 1991.3997 0. J. the lowland areas are around 29. As a matter of fact. Long-term and persistent air Global dimming and warming have opposite effects. Station name Station m (%) b Mann–Kendall Total Missing height (m) Z changes (%) data Tangkil 395 0.6 ° C.02 0. Concluding remarks sible to the reduced evaporation.16 −1. DJAMIL Table VII.11 −1.08 1984–1989 Doko 325 0.2071 1.20 −3.53 7. 1991). A higher SST above that critical value does not bring tial and temporal variation and trend of climate over about more evaporation.72∗ 6.12 – Wates Sawahan 620 0. ALDRIAN AND Y. the explanation and investigation of the probable cause ple. 2002). 0. the expansion of the dry area has shifted and Aldrian and Susanto (2003) the critical SST dur- the balances between the wet season and dry season ing the peak of the wet season or December–January is especially over the lowlands. Complementarily.10 −1. The results show the sea surface and eventually cool it.1002/joc .56 6.81 4.15 ° C (Hansen and Lebedeff. According to Bony et al.12 −1. The analyses Copyright  2007 Royal Meteorological Society Int.10 1980–1990 Pujon 1258 0.16 – Semen 625 0. 0.72 1.2 – Wates Kediri 175 0.3 ° C since the 1970s (Krishnan and Ramanathan.001. i. 4. Jones and Briffa. 1992. respectively. 0.01.71∗ 8. periods with lower rainfall amount may also indicate 1990. Thus.09 −0. S.13 −1. season in a 1-year period of the monsoon. ∗ .9277 0. more evaporation will take more latent heat away from ances to detect the annual trend. Gaffen Persistent reduction of evaporation may be associated et al.02 5. For exam.12 −1.67∗ 10.63 – Poncokusumo 1120 0. apparently has changed the local climate in this catch- Air pollutants hinder the solar radiation from reaching ment. 2002. Climatol. Liepert. Linear regression of the ratio of rainfall in wet season to total annual from 1955 to 2005 according to the formula: P ercentage ratio = m × year + b. 1996). all other things being equal. Longer dry Vinnikov et al. In this formed. Atmospheric thermodynamics suggests that with the global dimming phenomena (Stanhill and Cohen. (1997a. Walliser and Graham more susceptible to the climate change. As a result. IPCC. tion and evapotranspiration.1177 1.13 −1.61 – Kertosono 47 0. (1993) concluded that at SSTs over 29.4062 1.. 1988. Lau et al.08 −0.8194 0. East Java. The warming is also supported by evi- a global change in the evaporation supply. the absorbing aerosols over the Indian subcontinent of the evaporation reduction is beyond the scope of this have led to a statistically significant cooling of about study.6341 1.1. As it.5296 1.02 negative trend had more effect in the coastal area.446 E. Analyses of station ances between the wet and dry season that usually lasted records have shown that during the last 100 years the about 6 months in the early decade.63 – Dampit 450 0..05.14∗∗ 5.59 – Birowo 195 0. the ground. The cause of that there is no contradiction among methods that we the SST rise comes from the global surface temperature used. ick and Farquhar. increase or the global warming.27 5. 1987. 2002). the mid-1970s in the Tropics (Elliott et al.2833 2.b).+. ics but spill over to the sub-tropics.5 ° C deep atmo- Our results also suggest that the methods employed spheric convection decreases and clouds are no longer here are suitable for climatic trend analyses. the increase The decrease of the monsoonal strength as detected of the evaporation amount will not occur in the trop- by the EOF and FFT methods was caused by the imbal. The annual pattern globally averaged surface air temperature has increased turns out to be at a different state based on longer dry by ∼0. pollution in the lower and middle atmosphere due to which may lead to the same consequence.79 4. Ohmura and Wild. 2002.12 – Kediri 70 0. We have investigated and analysed the long-term spa- ble. then study.3448 1. the reduc- traffic over land.8679 0.. If the SST rises above that critical value. instead.15 −2. ∗∗∗ indicate Mann–Kendall significant level of 0. 1992. (1997) a result.93 6. local evaporation especially over large water bodies will be reduced. sea and air has reduced the radiation tion of local surface evaporation in the tropics. 1990.

Trebejo I. The climate trend detected by the EOF method shows Haylock MR. Frick P. Varney SK to an increased risk of extreme climate condition such as (eds). 1988.. Elliott WP. and the lowest in JJA. Greenhouse-Gas- highest amount of rainfall in DJF. 1921. Ngkoimani LO. Identification of three dominant rainfall Mann–Kendall trend test. Meira Filho LG. Lau KM. Climate Change 1995. wavelet. Cambridge University Press: Cambridge. 1979. Journal of Climate 4: 989–1008. accordance with the results of Dai et al. An operational objective analysis scheme. 2006. 1938. Contribution Acknowledgements of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change (IPCC). which leads to Geophysical Research Letters 24: 1351–1354. Gaffen DJ. Journal half century. Sea surface temperature and large- scale circulation influences on tropical greenhouse effect and cloud fall ratio during the wet season and an increase of the dry radiative forcing. Barnett TP. Corradi V. IPCC. a decrease of the monsoonal strength in the catchment for Baez J. temporal and seasonal details. Widodo FH. 364. Krusic P. drought over the dry season and flood and landslide over IPCC. signal represents a long-term decrease of precipitation. Karoly D. Alves LM. 2001. Surface observed global land although globally there is an increase in the precipitation precipitation variations during 1900-88. Almost all evidences support a decreasing Bony S. the last five decades. Harris N. Hansen J. Zulaikah S. Lebedeff S. Briffa KR. Secondly. 1996. Verhandelingen. Griggs DJ. Sud YC. 28: 435–448 (2008) DOI: 10. global tropospheric moisture. project 94/01/CW and by the INSIDE Project from Jones PD. Marengo JA. Maskell K (eds). especially in sub-tropical regions. Rebello E. The imbalance Houghton JT. Journal of Climate 10: trend. regions. Copyright  2007 Royal Meteorological Society Int. In other tropical 2943–2962. This trend is due to the imbalances Garcia VJ. 1987. Geophysical Research Letters 15: 323–326. Katternberg A. EOF and FFT analy. Global surface air temperature variations ASIA Pro ECO-ASI/B7-301/2598/20-2004/79071. a decrease of the monsoonal dominance. 200. which preserves the high rainfall area. The ENSO influence on this imbalance is temperature. 1997b. Research Letters 33: L04709. Monthly Weather Review 114: 745–757. Palmer J. 584. Smith ME. Saha S. Time scales and trends lated rainfall monthly and annually. temperature. The Science of Climate Change. spell period. Angell JK. The mountains of the Brantas catchment area during the dry Netherlands. Soon W. UK. Spe. Elsevier: Amsterdam. UK. Sakulich J. Sonechkin DM. Holocene 2: 165–179. Firstly. Schlessinger ME (ed). Cambridge University Press: Cambridge. climatic purposes. Seasonality of multidecadal and centennial variability in European temperatures: the catchment area is dominated by the monsoonal pattern wavelet approach. We report two major findings in this study. Theoretical and Applied Climatology 87: 41–59. Geophysical rainfall since the 1950s. The coastal areas are more influenced by the anomalies. more extensions of the dry Haylock M. 2003. Dai A. 81–93. Curtis A. 1959. Peterson TC. Magnetic Meteorology Observation: Batavia. the mountain areas. The IPCC Scientific Assessment. of the wet and dry seasons. Space and time scales of season. 1997. Bidegain M. at a 1-year period. The decreasing trend of rainfall in this region is in Braak C. Berlato MA. and the Mediterranean). This research was partially supported by the SARCS Cambridge University Press: Cambridge. Induced Climate Change: A Critical Appraisal of Simulations and The local orographic effect regularly happens over the Observations. Ding Y. the mountainous area is less susceptible to the climatic Hackert EC. Vincent LA. Climate Change. The spatial and temporal climate pattern in the Brantas Datsenko NM. Gates LD. 1991. areas. Methods of comparing annual rainfall records for that over some parts of the low latitudes (the Philippines. UK. Ambrizzi T. DelGenio AD. Indonesia. Monitoring tropospheric water vapor changes using radiosonde data. Shabalova MV. Journal of Geophysical Research 92: 13345–13372. Het Climaat van Nederlandsch Indie. IPCC. International Journal of Climatology 23: 1435–1452. This pattern consequently brings the Elliott WP. Craddock JM. we note in central England temperature data (1659–1990): a wavelet analysis. 1992. Van der Linden PJ. which can be seen by an increase in the number Hansen J. Fung IY. In fact. 1997. Spatial coherence and predictability of comparison to the low and a more stable dry period over Indonesian wet season rainfall. (993) D’Arrigo R.1002/joc . SPATIO-TEMPORAL CLIMATIC CHANGE OF RAINFALL IN EAST JAVA 447 have been performed for the 51-year period (1955–2005) References from the monthly rainfall data with several methods: the Aldrian E. Mechanisms of Java Rainfall change. 1991. Noguer M. changes in the annual pattern and an increase of the rain. spatial. UK. Journal of Climate 19: 1490–1512. Nicholls N. Callander BA. 2001. 2001 The Scientific Basis. Biometrika 30: Tirta I for providing rainfall data. 1986. by which the dry season gets Santos JL. A new measure of rank correlation. who found vol. Sokoloff D. Houghton JT. International Journal of Climatology 20: 1533–1541. Quintana J. Journal of Climate 10: 2055–2077. falls in the wet season to the total annual rainfall and leads IPCC. the wet season. Global trends of measured surface air decrease of the monsoonal strength than the mountain temperature. cial thanks to Sutopo Purwo Nugroho and the Perum Jasa Kendall MG. Cambridge University period brings as a consequence an increasing ratio of rain Press: Cambridge. Climate Change 1992: The Supplementary Report to the IPCC Scientific Assessment. Lau KM. Coronel G. Trends in total and extreme longer and spills into the second transitional period (the South American Rainfall in 1960–2000 and Links with sea surface boreal fall). 2000. the EOF 2 or the trend Cressman GP. Trends in extreme rainfall indices for an areas are detected over the lowland areas. 311–328. 1997. Moncunill DF. which resulted in a decrease in the accumu. during the past two centuries. Sud Y. Bijaksana S. 1910–1998. 2006. Barros VR. Susanto RD. Houghton JT. Journal of Climate 14: 3882–3887. Hence. Wilson R. 1990. regions within Indonesia and their relationship to sea surface ses. Susskind J. obvious for the second transitional period. J. Climatol. updated high quality data set for Australia. Hastenrath S. Climate Change. Weather 34: 332–346. Grimm AM. Journal of Geophysical Research 116: 449–463. 2001. McBride JL. Comparison and rainfall trend over the catchment area during the recent satellite assessment of NASA/DAO and NCEP-NCAR Reanalysis over tropical ocean: Atmospheric hydrology and radiation. Bony S. Jenkins GJ. over the twentieth century: Part I. 1997a. Houghton JT. Global surface air temperatures: Update of dry months in recent decades in the coastal area in through 1987. Aldrian E. Malaysia. 2007. Monsoon drought have established a declining trend of the sub-Saharan over Java. 1992. Callander BA. 8. Seasonal variability of there was a significant climatic change temporally and Indonesian Rainfall in ECHAM4 Simulations and in the Reanalyses: spatially. Lebedeff S. Ephraums JJ (eds). Monthly Weather Review 117: 765–783. of Climate 10: 1441–1462. Anuncia¸ca˜ o YMT. Haylock M. Lamb and Peppler (1992) and Nicholson. Xiaosu D (eds). Marino MB. Rusticucci M. Baliunas S. Nechet D. Roles of ENSO.

1993. 1998. Gunawan D. Vincent L. 227–258. Nicholson SE. Park J. Trewin B. Hennessy KJ. Copyright  2007 Royal Meteorological Society Int. Empirical data Econometrica 13: 245–259. International Journal Krishnan R. Graham NE. Navarra A (eds). Climate change: Is the hydrological cycle regulation. 1967. Peppler A. Jones PD. Boehm R. Lau KM. London. Journal of Climate 10: 381–392. ALDRIAN AND Y. Fast algorithm for spectral analysis of America. Journal of Climate 3: 662–677. Cycles and trends in the Czech 2325–2339. Lamb PJ. Peterson TC. Leyu CH. Non-parametric test of randomness against trend. Tuomenvirta H. Convective cloud systems and the last decade. precipitation). of Climate 6: 2063–2082. A practical guide to wavelet analysis. Plummer N. Science 298: 1410–1411. Journal of Climate 6: 1463–1466. Journal of Climate 5: 476–488. Maitrepierre L. Journal 269–284. Wu HT. Further case studies of tropical Atlantic Roderick M. Hanssen-Bauer I. Br´azdil R. and Tahitian meteorological readings. Statistical Methods in the Atmospheric Sciences. International Journal Liepert B. Bony S. Tibig I. The role of large-scale atmospheric Agricultural and Forest Meteorology 107: 255–278. Haylock MR. 1997. Alexandersson H. Easterling DR. 1992. 1970. 2000.448 E. Spatial patterns: EOFs and CCA. International Journal of Climatology 21: temperature fields in the Northern Hemisphere troposphere. 1993. Nicholls N. Allan RJ. Mann ME. 2004. Storch H. Journal of Applied Meteorology 6: 791–802. surface temperature and precipitation complexes over North Press WH. Finet A. USA. 2001. 2002. Parker D. evaporation over the past 50 years. Charles Griffin: Heino R. Trends in Wallace JM. temperature series using wavelet transforms. Chambers LE. Lugina KM. Vinnikov KY. Observed reductions of surface solar radiation at of Climatology 18: 1141–1164. 59. Empirical eigenvectors of sea-level pressure. 272. 1995. 467. 1945. Lefale P. Auer I. Karl TR. Structure and seasonality of extreme daily rainfall and temperature in southeast Asia and the interannual and interdecadal variability of the geopotential height and South Pacific: 1961–1998. Trends in total rainfall. 28: 435–448 (2008) DOI: 10. Geophysical Research Letters 29: 1340. Suppiah R. Journal of Geophysical Research 98: 12881–12893. Evidence of surface cooling from of Climatology 24: 1661–1670. Groisman P. Ramanathan V. In Analysis Isobe H. Lwini T. Application of Statistical Techniques. Cohen S. Wilks SD. sites in the US and Worldwide. Homogeneity adjustments of In Situ atmospheric climate data: extensions of the southern oscillation index using early Indonesian A review. of Climate Variability. Szentimrey TS. Torok S. 1995. Saharan drought. Heidelberg. events and numbers of dry days in Australia. Geophysical Research Letters 29: Torrence C. global temperature: a multiple wavelet correlation approach. 1998. 2002. Preisendorfer RW. Climatol. Inape K. unevenly sampled data. von Storch H. Wild M. Elsevier Science: New York. von Ouprasitwong N. 1421–1433. 1990. Interannual temperature events and shifts in International Geophysics Series. 425. Kalvov´a J. Springer: Berlin. International Journal of Climatology 18: 1493–1517. Interactions 4-1: 1–53. Førland EJ. DJAMIL Kendall MG. absorbing aerosols. Earth MA. Stanhill G. Yee D. S. Mestre O. Academic Press: Burlington. Pahalad J. 2002. Tran VL. Plummer N. 2002. Jones P. Lau KH. accelerating? Science 298: 1345–1346. J. Principal Component Analysis in Meteorology Kutzbach JE. Bulletin of American Meteorological Society 79: 61–78. Gullett D. Page CM. 1989. The Astrophysical Journal 338: 277–280. 2001. Zhang Y. Collins DA. Kaltofen MH. Groisman PY. 1988. Salinger MJ. Farquhar G. Pre-1866 1998. An overview of African rainfall fluctuations of Walliser DE. 1998. Mann HB. Daw G. Konnen GP. Rank Correlation Methods. and Oceanography. Kestin TS. on contemporary global climate changes (temperature and Manton MJ. Hennessey KJ. Journal of Climate 11: Pisoft P. Global dimming: A review of the evidence. 1993. Della-Marta PM. UK.1002/joc . Warm-Pool sea surface temperatures: coupled interactions and self Ohmura A. Compo GP. Nicholls N. The cause of decreased pan surface atmospheric and oceanic patterns associated with sub. heavy rain surface temperature. circulation in the relationship between tropical convection and sea Suppiah R. Salinger J. Rybicki GB.