Reconstructed streamflow for Citarum River, Java, Indonesia: linkages to tropical climate dynamics
- First Online:
- Cite this article as:
- D’Arrigo, R., Abram, N., Ummenhofer, C. et al. Clim Dyn (2011) 36: 451. doi:10.1007/s00382-009-0717-2
- 362 Views
The Citarum river basin of western Java, Indonesia, which supplies water to 10 million residents in Jakarta, has become increasingly vulnerable to anthropogenic change. Citarum’s streamflow record, only ~45 years in length (1963-present), is too short for understanding the full range of hydrometeorological variability in this important region. Here we present a tree-ring based reconstruction of September–November Citarum streamflow (AD 1759–2006), one of the first such records available for monsoon Asia. Close coupling is observed between decreased tree growth and low streamflow levels, which in turn are associated with drought caused by ENSO warm events in the tropical Pacific and Indian Ocean positive dipole-type variability. Over the full length of record, reconstructed variance was at its weakest during the interval from ~1905–1960, overlapping with a period of unusually-low variability (1920–1960) in the ENSO-Indian Ocean dipole systems. In subsequent decades, increased variance in both the streamflow anomalies and a coral-based SST reconstruction of the Indian Ocean Dipole Mode signal the potential for intensified drought activity and related consequences for water supply and crop productivity in western Java, where much of the country’s rice is grown.
Monsoonal droughts and floods in southeast Asia have profound impacts on human populations, particularly in rural settings where crop productivity is highly vulnerable to such extremes (Naylor et al. 2007). In such settings, the need for adequate and reliable water resources is of increasing concern. Some experts forecast severe water shortages, along with drought and flood extremes, for much of Asia in coming decades, with potentially profound impacts for human populations and the environment. One likely cause of such change is the release of anthropogenic greenhouse gases, which are expected to result in substantial alterations in the distribution and variability of rainfall (Bates et al. 2008). Circulation changes driven by expected warming are anticipated to impact the Inter-tropical Convergence Zone (ITCZ) and monsoonal rainfall patterns over much of southern Asia, although there is considerable uncertainty in future projections (Overpeck and Cole 2007; Abram et al. 2008; Vecchi et al. 2008). Water-related issues are thus of critical interest to those involved in sustainability and climate risk management efforts in drought and flood prone regions of the globe (e.g. Boer 2007).
In Indonesia, integrated action is needed by (local, regional, national) government, climate specialists and other planners to mitigate against the increased threats of extreme drought and related wildfires, as well as severe flood events, in areas where water demand, land use and development are rapidly on the rise (Aqil et al. 2007; Overpeck and Cole 2007; Abram et al. 2008; Field et al. 2009). Indonesia’s most populated locations, including the island of Java (where much of the nation’s rice is grown), are particularly vulnerable to extremes in rainfall and water shortages (Boer 2007).
Historical records of streamflow are typically limited in length for many areas of Asia and the globe. The Citarum streamflow record (Figs. 1, 2), one of the longest such time series available across monsoon Asia, is only about 45 years in length. This is clearly insufficient for understanding the potential range of hydrological variability and extremes in this region, and how this variability may be changing in a warming world. There is therefore a need for additional information on the behavior of the Citarum River and other hydrologically-important locations in monsoon Asia. Long records of streamflow variability would thus be highly useful for assessing past activity, evaluating model simulations, and other applications relevant to climate risk management (Asian Institute of Technology 2005, http://portal.iri.columbia.edu/portal/server.pt, NOAA 2009).
Alternatively, high-resolution, well-dated proxy records from tree rings, corals and other data archives are of considerable value for developing longer time series for analysis. Tree-ring based hydrometeorological reconstructions have been developed for numerous sites around the globe. Notably, this research includes much work in the United States (e.g. Stockton and Jacoby 1976; Cleaveland 2000; Meko et al. 2001; Woodhouse 2002; Woodhouse et al. 2006), as well as Turkey and other locations (e.g. Akkemik et al. 2008,) However, paleostreamflow records are quite limited for Asia (e.g. for China—Gao et al. 2006, and Mongolia—Pederson et al. 2001), particularly near the equator.
In this paper, we present a tree-ring reconstruction of Citarum River streamflow that extends available instrumental observations by more than two centuries, and discuss its linkages with both tropical Indian and Pacific atmosphere–ocean climate variability. The tree-ring based Citarum River basin streamflow reconstruction was generated using a tree-ring chronology network from Tectona grandis (teak) using multiple sites across western Java and Sulawesi, Indonesia. Based on the closely coupled relationships between tree growth, rainfall, and Indo-Pacific climate for this region, this reconstruction was used to quantitatively assess past hydrometeorological variability over the pre-instrumental period. We compare this record to other observational and proxy records for the region, specifically to a coral-based index of Indian Ocean SSTs (Abram et al. 2008). Along with tropical Pacific variability linked to ENSO, Indian Ocean climatic conditions have been implicated as a dominant factor forcing drought and wetness extremes across much of Australasia (Abram et al. 2008; Ummenhofer et al. 2009).
2 Materials and methods
2.1 Instrumental data
We use the instrumental gauge record of monthly streamflow for the Citarum Basin, west Java (Figs. 1, 2) to calibrate the tree-ring reconstruction. The Citarum River has experienced considerable human modification due to land usage and other factors (Santoso and Warrick 2003; Lasco et al. 2004; Boer et al. 2004, 2005, 2007). Despite this, it correlates very strongly with large-scale climate variables, e.g. the Southern Oscillation Index (May–July, r = 0.49, 0.0007, n = 44) and Dipole Mode Index or DMI (Oct-Nov HADISST, r = −0. 75, 0.0000, n = 44). Albeit short, it is one of the longest and highest quality streamflow records for Indonesia, spanning from 1963-present (data from the Ministry of Public Work, Republic of Indonesia; kindly provided by Dr. R. Boer, Bogor National University, Bogor, Java, Indonesia). This reconstruction is only relevant to the Citarum station; other streamflow records for Java are typically 10 years or less in length (R. Boer, Bogor Agricultural University, pers. comm..). Gridded HADISST (global ocean surface temperature, monthly 1 degree area grids, 1870 to present, Hadley Center, UK; Rayner et al. 2003), rainfall data from the Vasclim (Variability Analysis of Surface gridded monthly rainfall dataset, Climate Observations, analysis, 1951–2000, 0.5° spatial resolution (Beck et al. 2005), and the Kaplan Nino-3.4 SST index (1856-present, anomalies in 5S–5N, 120–170W, Kaplan et al. 1998; Knutson et al. 1999) were also used for comparison to the reconstruction, using simple and spatial field correlation analyses. Analyses were performed using KNMI Climate Explorer (http://climexp.knmi.nl/).
2.2 Indonesian proxy data
Tree-ring site information for six candidate chronology predictors used in nested regression models to reconstruct Citarum River, Java streamflow
Correlation year t (t + 1)
2. Donoloyo Cagar Alam
3. Pagerwunung Darupono
5. Sakla (combined Saradan, Klangon Natural Forest)
6. RBGB (combined Randublatung. Gubug Payung, Bekutuk, Begin)
We also utilize the coral-based reconstruction of the Indian Ocean Dipole Mode Index (DMI SST) (Abram et al. 2008; Fig. 1), which includes coral oxygen isotope data for the Mentawai Islands, off Sumatra, western Indonesia, for independent validation of the Citarum streamflow reconstruction. This record is based upon a suite of coral isotopic records from across the Indian Ocean with a common period since 1846. These data were used to develop a basin-wide gradient index of SST variability linked to the IOD system, ENSO and the Asian monsoon.
2.3 Reconstruction development and analysis
Calibration and verification statistics for four nested reconstruction models. Ar2: variance explained in calibration period, accounting for degrees of freedom
1. 6 Proxy Series: 1834–2000
2. 5 Proxy Series: 1830–2000
3. 4 Proxy Series: 1820–2000
4. 3 Proxy Series: 1759–2000
The reconstruction is most robust at higher frequencies, since the instrumental streamflow time series for the September–November season has no autocorrelation, and since the proxies were prewhitened to remove persistence (there are varying degrees of coherency at low frequencies for some series (Cook and Kairiukstis 1990). Lags t and t + 1 were used in the regression (Cook and Kairiukstis 1990).
Trends in the recurrence rate of low flow and high flow streamflow events were tested against the null hypothesis of constant recurrence rate. A Gaussian kernel analysis was employed for this purpose (Mudelsee et al. 2003), for which the event threshold was set at −1 sigma for low flow and +1 sigma for high flow events, with the data normalized for the 1963–2000 period of overlap between the observed and reconstructed streamflow data. The kernel window was 35 years and 10,000 bootstrap simulations were used to estimate the 90% confidence interval.
3 Results and discussion
A low-variance period in the early-mid 20th century is nearly unique over the full length of the reconstruction (Fig. 4; variance based on normalized data is 1.01 for entire 1759–2006 period, vs 0.64 for 1905–1960 and 0.68 for 1920–1960). The latter interval of low variability has been described as a pronounced weakening in the interannual variance of the monsoon-ENSO-IOD systems, primarily detectable within the 4–7 year peak bandwidth (Kestin et al. 1998; Torrence and Webster 1999; Allan 2000; Abram et al. 2008). The cause or causes of this low variance period are not known, but it is believed to be a system-wide phenomenon that is also linked to the behavior of the Indian monsoon, and may involve a shift in internal atmosphere–ocean circulation variability, external forcing, or some combination of the two (see references above). Interestingly, reconstructed streamflow during an earlier period of comparable length (1780–1820) also features lower than average variance (0.71), and may indicate another time of relatively quiescent ENSO and/or IOD activity; although we caution that this event occurs during the earlier, less replicated part of the reconstruction.
Percentage of years in ~50 year reconstruction increments in which normalized departures of reconstructed streamflow exceed ±1 standard deviation
Time interval (years)
High flow (years) (%)
Low flow (years) (%)
A long-term perspective on past streamflow variability has been provided by the tree-ring reconstruction for the Citarum River, west Java, Indonesia, one of the first such high-resolution proxy reconstructions for Indonesia and for monsoon Asia as a whole (Figs. 1, 2). Although the instrumental observations used for calibration and verification testing are limited in length, and proxy availability and model quality decline back in time, model statistics are reasonably robust over the more recent period of the reconstruction and provides some climatic information even back to the weaker, earlier part of the record. The reconstruction substantially extends the short streamflow record for the region, by several centuries. It displays sensitivity to both tropical Pacific and Indian Ocean climate conditions, as illustrated by the spatial correlation fields with large-scale SST and rainfall data for the region (Fig. 8). Over the common period, comparison with Niño 3.4 SSTs indicates a very weak correlation from 1920 to 1960, a time of unusually low (nearly quiescent) ENSO activity which is also evident in the trans-Indian Ocean coral DMI (Abram et al. 2008) and in tropical Pacific coral records (Urban et al. 2000). Our results indicate that the apparent weakening of the ENSO-IOD system that occurred at this time was unusual, and possibly unique, in the context of the past two centuries. It has been hypothesized that future greenhouse warming may lead to further intensification of tropical climate variability (Abram et al. 2008), and our results demonstrate that this could have serious implications for water resources in the Indonesian archipelago, including the Citarum basin.
This streamflow reconstruction adds to the network of proxy records and reconstructions being generated from across monsoon Asia (e.g. Buckley et al. 2006). The temporal and spatial coverage of this network is improving and will prove useful for eventual multiproxy studies that investigate ENSO evolution and behavior in the tropical Indo-Pacific region. In particular, this record might provide a useful verification tool for historical climate data rescued from historical archives (e.g. Atmospheric Circulation Reconstructions over the Earth, or ACRE, http://www.met-acre.org/Home). A multiproxy approach using tree rings, corals, and other proxies can be integrated with historical data to considerably aid understanding of the spatiotemporal evolution of ENSO, its stationarity over time and its teleconnections.
This project was funded by the National Science Foundation’s Paleoclimate program, Grant No. OCE 04-02474. We thank Dr. Rizaldi Boer of Bogor National University, Java Indonesia, for providing the streamflow data. We very much appreciate the comments of the reviewers, which have significantly improved the manuscript. LDEO Contribution Number 7312.