Is the trend in chlorophyll-a in the Arabian Sea decreasing?



[1] Recent studies of satellite-derived Chlorophyll concentrations (Chl-a) in the western Arabian Sea (AS) have suggested an increasing temporal trend, but the length of the records used have typically been too short to resolve longer-term trends, if any. Our analysis of a long term satellite ocean color data shows a change of trend in the summer chlorophyll for the western AS before and after 2003; Chl-aconcentration was indeed increasing till 2003, but appears to be declining since then, indicating a secular multi-year trend in Chl-avariability. However, this trend is not uniform over the entire region. Analysis of wind, sea surface temperature (SST), Sea Level Anomaly (SLA) and thermocline depth, suggests that the declining summer monsoon chlorophyll-a(Chl-a) concentration may be due to increasing SLA in this region. The earlier observed biological changes in the western AS could be an artifact of the change in local winds and ocean dynamics, which may be a part of the natural long-term variability.

1. Introduction

[2] For more than a decade now, a series of satellites have provided high-quality ocean color data, creating an immense opportunity to monitor biological productivity on enhanced spatio-temporal scales, and have facilitated improved understanding of ocean biogeochemistry and physical-biological interactions at all time-scales [e.g.,Murtugudde et al., 1999; Waliser et al., 2005; Jin et al., 2012]. Analysis of satellite and in-situblended ocean chlorophyll (Chl-a) records showed a significant 6% decrease in the global net oceanic primary production since 1980 [Gregg et al., 2003] whereas analysis of 6 years Sea viewing Wide Field-of-view Sensor (SeaWiFS) data showed an increase by 4.1% (P < 0.05) in the global ocean chlorophyll [Gregg et al., 2005]. The observed change was not geographically uniform; it was prominent in the coastal regions (∼10.4%). A local maximum in the increase was observed off the Somali coast (>37%) in the western Indian Ocean. The northern and equatorial Indian Oceans have also contributed to the observed increase by 14% and 9%, respectively [Gregg et al., 2003].

[3] Some recent observations have shown contradicting trends in ocean productivity for the Arabian Sea (AS). It was reported to have increased by more than 350% in the western AS [Goes et al., 2005] owing to strengthening of the monsoonal winds in this region whereas there is no such trend in the eastern half of this relatively small regional sea [Prakash and Ramesh, 2007]. Since a trend in primary production, whether positive or negative, can have greater implications on the carbon cycle and oceanic source/sink of atmospheric CO2, it is necessary to ascertain whether the trend is of basin scale or a localized one and whether it is long-term or secular. Since the previous studies have revealed large seasonal-to-interannual variability in the AS, a sufficiently long record (at least a decade or longer) is required to identify any long-term or climate-driven trends [McClain, 2009]. Since AS is a basin of global significance, any trends in its productivity would imprint a global signature in the carbon budget, and likely also in ocean circulation, ecosystem processes, biogeochemical cycling and could have significant implications for marine fisheries. The present study analyzes SeaWiFS surface Chl-adata to investigate long term changes, if any, in the Arabian Sea. SeaWiFS has been providing continuous data on ocean optical properties since late 1997 and may be sufficient to resolve any long-term trends [McClain et al., 1998]. This study investigates whether the observed Chl-a trend is basin scale or region specific. We also delineate some of the possible mechanisms which can explain the observed change.

2. Data and Methodology

[4] Throughout this analysis, we have focused on the variability of bio-physical properties over the AS bounded by 0–25°N and 45–80°E. The monthly composite Level −3 Standard Mapped Images (SMI) data of SeaWiFS (September 1997–December 2010) at 9-km spatial resolution were obtained from the National Aeronautics and Space Administration (NASA) Ocean Color Website ( Though we are showing 1997 data in our plots, it is not included for trend analysis since summer 1997 was poorly sampled by SeaWiFS. Tropical Rainfall Measuring Mission/Microwave Imager (TRMM/TMI) data were used to study Sea Surface Temperature (SST) for period January 1998 to December 2010. We used monthly SST data at a pixel resolution of 0.25° obtained from university of Hawaii web-site (

[5] We also used the QuikSCAT scatterometer winds (July 1999 to November 2009) and Cross Calibrated Multiplatform (CCMP) ocean surface winds (January 1998 to November 2009) [Ardizzone et al., 2009; Atlas et al., 2009]. Monthly maps of wind and wind stress curl (proportional to the vertical upwelling velocity produced by Ekman pumping), were derived from these data. Ocean temperature analysis data (for 1997–2010) are obtained from the “ENSEMBLES” (EN3) dataset provided by the Met Office Hadley Centre [Ingleby and Huddleston, 2007]. A large fraction of the profiles in that dataset were originally from the World Ocean Database (WOD) 2005, a collection of more than 7 million ocean measurements gathered globally from ships, moored buoys and Argo profiling floats. The WOD 2005 measurements are supplemented in the EN3 data by temperature profiles from the Global Temperature Salinity Profile Program as well as by temperature data collected from profiling floats provided by Argo. The mean depth of the 23°C isotherm (D23), a representative depth of thermocline depth, in the Arabian Sea is derived from these dataset. The SSHA data used in this study were derived from the weekly and monthly merged data from multi-satellite sensors (TOPEX/Poseidon, ERS and Jason) over the 12 year period from January 1997 to December 2010. The data have spatial resolution of 1/3 degree resolution and provided by Aviso.

3. Results and Discussion

[6] The basin average time-series analysis of satellite derived monthly Chl-a concentration in the Arabian Sea during 1997–2010 is shown in Figure 1. The summer (June-September) peak surface Chl-aconcentration appears to have increased during the years 1998–1999, but there is no such secular increase after 1999. The increase during 1998–1999, therefore, may be a pseudo trend influenced by the Indian Ocean Dipole/Zonal Mode (IODZM) and the El Niño-Southern Oscillation (ENSO) events during 1997–98, described in detail later. A weak basin-scale increasing trend in the monthly mean Chl-a (r2 = 0.01) for AS has been reported for the period 1997–2007 [Prasanna Kumar et al., 2010] but our analysis for 1997–2010 shows no such trend (slope: −1.4*10−4 ± 3.6*10−4; r2 = 0.001). To understand whether this observed trend has been the same for the entire basin or restricted to some part of region, we select four most productive regions, namely R1 (Southeastern AS), R2 (Northern AS), R3 (Oman Coast) and R4 (Somali Coast) in the Arabian Sea (shown as boxes in Figure 2a) and analyze area-averaged time-series of Chl-a for each regions. Region R1 experiences a low intensity upwelling during the summer monsoon which makes this zone significantly productive during the season [Wiggert et al., 2005]. R2 is more productive during the winter monsoon when the convective mixing brings ample nutrients into the surface layer and triggers high biological production [Madhupratap et al., 1996]. This zone is also quite productive during the summer monsoon. Regions R3 and R4 are the most productive regions in AS during the summer monsoon; southwesterly monsoon wind causes an intense coastal and open ocean upwelling in these regions [Wiggert et al., 2005]. The spring and fall inter-monsoons in AS are characterized by weaker winds which cause stratification and result in oligotrophic conditions in the surface [Murtugudde et al., 1999; Banse and English, 2000; Wiggert et al., 2005]. The present analysis shows that there is no appreciable secular change in surface Chl-a in regions R1, R2 and R3 (Figure 2b). However, region R4 shows two distinctly different trend regimes in the summer Chl-a (before and after 2003). Since R4 shows a unique trend in summer surface chlorophyll peak values, we select this region for detailed analysis. Interestingly this is the same region in which an earlier study has shown a remarkable increase in productivity [Goes et al., 2005].

Figure 1.

Area averaged monthly time-series of SeaWiFS derived Chl-a for the Arabian Sea basin (45 to 80°E & 0 to 25°N) for years 1997–2010.

Figure 2.

(a) Box shows the areal limit for regions R1 (73–78°E & 8–14°N), R2 (61–66°E & 22–26°N), R3 (52–58°E & 14–20°N) and R4 (47–55°E & 5–10°N). The base image shows composite (1998–2010) surface chlorophyll for the Arabian Sea basin. (b) Area averaged monthly time series of Chl-a for the region R1, R2, R3 and R4. The trend lines shown in R4 depict the increasing and decreasing trends during 1998–2003 and 2004–2010, respectively.

[7] The time series of Chl-a in R4 has two significantly distinct trends (Figure 2b): the summer peak chlorophyll concentration increased consistently from 1998 to 2003 (slope: 0.24 ± 0.06; r2 = 0.85; p = 0.02) but has been decreasing since then (slope: −0.06 ± 0.03; r2 = 0.44; p = 0.10). An earlier study [Goes et al., 2005] reported a monotonic increase in the summer Chl-aconcentration (>350%) in the western Arabian Sea. They also found a corresponding decrease in sea surface temperature (SST) and intensification of surface winds over the region. Since the strengthening of sea surface wind is manifestation of any change in the land-sea temperature contrast, they attributed observed increase to the global change. These results have since been disputed by several authors [Prakash and Ramesh, 2007; Kahru and Mitchell, 2008; Prasanna Kumar et al., 2010]. The observed increase may not be a trend but just an impact of the 1997–1998 El-Niño/IODZM [Sarma, 2006; Kahru and Mitchell, 2008], one of the strongest in instrumental records. The year 1997 was also unique in a sense that, in this year the Indian Ocean was also influenced by an extreme climatic event known as IODZM [Saji et al., 1999; Webster et al., 1999]. During an IODZM, a pattern of zonal (east-west) variability across the Indian Ocean basin is established with anomalously cold SST in the eastern (off Sumatra) and a warm SST in the western Indian Ocean. The Arabian Sea productivity was reduced by 30% during the 1997 IODZM compared to a normal year [Sarma, 2006]. However, if we exclude the year 1997 and estimate the change from year 1998, the observed change was 2-fold in region R4: the summer peak Chl-a concentration in 1998 and 2003 was 0.91 mg/m3 and 1.77 mg/m3, respectively. The observed increase during 1997–2003 was also amplified due to presence of a cold core eddy in 2003 (Figure 3). It cannot, however, be refuted that there was considerable increase in productivity and intensification of bloom even after 1997–1998. But the increase was not monotonic and persistent. The time series plot (Figure 2b) suggests a distinct decrease, from 1.77 mg/m3 in 2003 to 1.07 mg/m3 in 2010.

Figure 3.

Weekly SSHA in the western Arabian Sea during the third week of August for years 1997–2010. Box represents region R4. The presence of a cold core eddy, marked by negative SSHA, can be clearly seen in year 2003.

[8] Since it is well established that summer productivity in the western Arabian Sea is mainly governed by wind-driven upwelling along the western coast [Wiggert et al., 2005], we analyzed CCMP wind data to see if there is any significant change in wind speed/stress over this region during the past 13 years. We also analyzed SST and sea level anomaly (SLA) to understand whether it is only the wind that governs productivity or other physical processes too play a significant role. The time series plot of CCMP winds (Figure 4), averaged over R4 does not show any appreciable change in wind strength from 1997 to 2003. Analysis of QuickSCAT wind also shows a similar result. We have also computed wind stress curl over the study region; similar to wind speed, wind stress curl also did not show any significant change. Despite any change in the wind strength/stress curl, the Chl-aconcentration has declined significantly since 2003. This contradicts the earlier hypothesis that strengthening of wind during 1997–2003 had caused enhanced upwelling and had increased the productivity. Therefore, the observed change in Chl-a concentration in R4 would need a different explanation and for that, we explored the possible role of ocean to explain the varying trend regime.

Figure 4.

Time series of area averaged monthly wind speed by CCMP (black, solid line), Quick Scat (blue, long dash), TMI (green, short dash) and monthly wind stress Curl (X 10−7) for the region R4 for years 1997–2009 (red bar). Trends lines fitted to individual data are also shown in the same color of time series. TMI wind data has been plotted for the period of 1997–2003 (following Goes et al. [2005]); Slopes of the trend lines are insignificant (up to two decimal places).

[9] In order to understand the responsible mechanism for the observed Chl-achange in the region under discussion, i.e., R4, we analyzed co-variability of the SST, SLA, thermocline depth (23°C isotherm) and Chl-a. Although the area averaged SST does not show any secular trend, the minimum SST does show a marginal decrease during 1998 to 2003 and has been increasing since 2004 (Figure 5b). The average SST during the summer (June to September) monsoon has decreased by 1°C during 1998 to 2003, while it has risen since then by 1°C. Since SST is one of the indicators for the strength of upwelling, the observed decrease in SST is possibly due to increase in upwelling in this region and vice versa. The warming of SST also causes an increase in the mean sea level [Domingues et al., 2008]. In the northern Indian Ocean, the sea level rise is estimated to be occurring at an average rate of 1.3 mm/yr [Unnikrishnan and Shankar, 2007]. It is consistent with the rate of global sea level rise (1.7 ± 0.5 mm/year), although not uniform over the basin [Bindoff et al., 2007]. Satellite data show that SLA along the Somali coast shows a negative trend (Slope = −0.06 cm/month, P-value = 0.02;Figure 5, blue line) during the period 1997 to 2003 and a positive trend since 2004 (slope = 0.07 cm/month, P-value = 0.03). An analysis of satellite altimeter andin-situ data, in conjunction with the model studies, indicated that the SLA in the northern Indian Ocean has an oscillatory character and may be a natural decadal variability [Han et al., 2010]. Sea level is a good indicator of the vertical movement in the thermocline [Rebert et al., 1985]; an increase in the sea level is associated with deepening whereas a decrease indicates shallowing of the thermocline. The depth of the 23°C (D23) isotherm, which is a proxy for thermocline depth (Figure 5, red line), was shallower than climatology during 1997 to 2004, deepening thereafter. The average depth of D23 during the summer (June to September) was 130 m in the year 1997 which gradually shoaled up to 85 m in 2003 and 71 m in 2004, respectively. After 2004 it gradually deepened to 109 m in 2010. Similar observations have also been reported from the Lakshadweep Sea where the stronger upwelling was associated with shallower thermocline while a weaker upwelling was associated with a deeper thermocline in 2002 and 2005 summer monsoons, respectively [Gopalakrishna et al., 2008]. In the Arabian Sea, the thermocline and the nutricline are found to be closely associated [Wilson and Adamec, 2002], and therefore any change in the depth of nutricline has a direct impact on the supply of nutrients to the surface layer and the associated productivity. The deepening of the thermocline in the western Arabian Sea is associated with a reduced nutrient concentration below the euphotic zone and thus causing a decrease in surface chlorophyll. Our analysis clearly suggests that it is the change in the sea level and its associated effect on the thermocline that governs the productivity trend in the western Arabian Sea. It has been shown through studies in other basins that the remotely forced planetary waves may modulate the thermocline, and thereby without any local forcing from atmosphere such as winds, it may cause upwelling/downwelling [Girishkumar et al., 2011a, 2011b]. The sea level changes may also occur due to processes such as local and remote forcings, steric component of sea level due to thermal expansion of water column, ocean mass component from ice melting, rivers, etc. In the study area (R4) it is well known that remote forcing (Rossby waves) is also contributing to the sea level changes and thermocline movement [Brandt et al., 2003]. Though it has been widely debated that productivity over large part of the world ocean, including the Arabian Sea, has been increasing due to human-induced global warming [Goes et al., 2005; Gregg et al., 2005; Henson et al., 2010], the present analysis using 13 years long data indicates that the observed variability may be a manifestation of the decadal oscillation in sea level anomaly. We need a longer time record to identify the effect of recent climate perturbations, if any, on marine productivity; it has been demonstrated, using ocean color data and model studies, that time series data of ∼40 yrs are required to distinguish anthropogenic trend from a natural variability [Henson et al., 2010]. Clearly, understanding the mechanisms, local and remote, of sea level variability in the region would need further analysis and is beyond the scope of this study.

Figure 5.

(a) Area averaged monthly time series of sea level anomaly (blue) and thermocline depth (D23- red) along with their trends for the region R4 during 1997–2010. (b) Area averaged monthly time series of TMI SST for the region R4. Trend lines shows decrease (slope = −0.17) from 1998–2003 and increase (slope = 0.10) from 2003 onwards in the minimum SST.

4. Conclusion

[10] In conclusion, our analysis of satellite derived chlorophyll suggests that despite the lack of any appreciable change in the wind over the western Arabian Sea, the productivity has significantly varied during the past 13 years. A comprehensive analysis of SLA and D23 suggests that it is the change in sea level that governs productivity in this basin on a longer time scale. The present study also suggests that the observed variability in productivity is not an effect of the global warming but may be a part of the decadal oscillation. This should drive future efforts towards understanding of natural variability in the World Ocean and not just on the effects of global warming. Natural variability is clearly a baseline without which unreliable warning messages will continue to be issued about human impacts on the oceans.


[11] The encouragement and facilities provided by the Director, INCOIS are gratefully acknowledged. We also thank various organizations who have contributed significantly toward collection of data and making it available for users. This is INCOIS publication 126.

[12] The Editor thanks two anonymous reviewers for assistance evaluating this paper.