Decreasing surface chlorophyll in the tropical ocean as an indicator of anthropogenic greenhouse effect during 1998–2020

Available satellite data reveal a decreasing trend in surface chlorophyll (SChl) over the entire tropical ocean until 2020. Where contributions by internal variability and external forcing remain unclear. Here, state-of-the-art climate model simulations are analyzed to show that external forcing significantly contributes to the decreasing SChl trend. In contrast, internal variability plays a weak or even offsetting role. As for the underlying processes, anthropogenic greenhouse emissions lead to a remarkable reduction in SChl over the tropical oceans, whereas industrial aerosol load facilitates a considerable increase in SChl in the western tropical Pacific. In addition, the negative phase of the interdecadal Pacific variability during 1998–2020 contributes to an increase in SChl, while the impact from the Atlantic multidecadal variability is relatively weak in facilitating a decrease in SChl. Overall, these results imply that the impact of anthropogenic forcing has emerged as indicated in the tropical marine ecosystem.


Introduction
The tropical ocean is characterized by a permanently stratified density structure and relatively high primary production, which are attributed to strong upwelling driven by the trade winds. Under the greenhouse warming scenario, primary production tends to decrease in the global ocean, with a significant reduction of 4%-11% in the tropical ocean (Fu et al 2016, Kwiatkowski et al 2017, Moore et al 2018. The surface chlorophyll (SChl) concentration is a relatively direct proxy for photosynthetic biomass in the ocean surface layer, albeit with substantial variations in the chlorophyll-carbon ratio existing in this region (Wang et al 2009, Behrenfeld et al 2016, Siemer et al 2021. As for the role of chlorophyll within the earth system, changes in SChl not only represent a response of the marine ecosystem to climate change but also feedback to the climate system by affecting the vertical distribution of shortwave penetration in the tropical ocean (Lewis et al 1990, Falkowski 2012, Zhang et al 2019, Doi and Behera 2022, Khandare et al 2022, Sein et al 2022. Therefore, determining how the SChl changes over the tropical ocean in response to global warming is important to climate and marine ecosystem predictions. Due to the limitation of continuous observations, long-term trend in SChl is subject to substantial uncertainty resultant from external forcing and internal climate variability effects, with the former being considered to play a leading role (Martinez et al 2009, Frölicher et al 2016. In terms of external forcing, greenhouse gases (GHGs) and aerosols act to significantly modulate decadal climate variability and long-term trends (Fyfe et al 2013, Wang et al 2015, Hua et al 2018, Qin et al 2020. For example, some evidence suggests that the recent global warming hiatus is related to the negative phase of the interdecadal Pacific variability (IPV), which can be attributed partially to anthropogenic aerosol change (Smith et al 2016). The corresponding enhancement of the trade winds associated with this cooling condition strengthens the equatorial upwelling and subsequently increases SChl in the tropics (Martinez et al 2009). In addition, as a natural factor, major volcanic eruptions also could impact the multiyear biological production in the tropical Pacific (Chikamoto et al 2016). However, the impact of external forcing on the SChl trend remains unclear at this stage due to the limitation of observational length. For example, previous studies have widely suggested that the time of emergence (ToE) in chlorophyll signals due to anthropogenic activity needs more than 50 years in the tropical oceans as revealed using large ensemble (LE) experiments from four Earth system models (ESMs) (Schlunegger et al 2020). Meanwhile, Henson et al (2010) pointed out that the emergence time of climate change-driven SChl trend is ∼20-30 years in the equatorial region. Dutkiewicz et al (2019) found that remote sensing reflectance has an earlier ToE than chlorophyll, suggesting that climate change affects the marine ecosystem more quickly than previous estimations.
Internal climate variability also plays an important role in determining the recent decadal changes in SChl over the tropical oceans, such as the IPV and the Atlantic multidecadal variability (AMV) (Zhang et al 1998, Martinez et al 2009, Taucher and Oschlies 2011, Wernand et al 2013, Gregg and Rousseaux 2014, Krumhardt et al 2017, Hammond et al 2018, Lim et al 2022. The surface cooling over the eastern tropical Pacific associated with the cold phase of the IPV acts to weaken the stratification and consequently increase SChl. Also, the intensified trade winds during the cold phase of the IPV further facilitate the shoaling of the thermocline (analogy to nutricline) in the eastern tropical Pacific (England et al 2014), which contributes to the upwelling of cold nutrient-rich water from below into the mixed layer (or mixing with these deep waters). The warm phase of AMV can lead to a sea surface temperature (SST) cooling in the eastern tropical Pacific, while the cold phase of IPV results in an SST warming in the equatorial Atlantic through the atmospheric bridge (Yao et al 2021). Meanwhile, the warm phase of AMV also led to a warming in the Indian Ocean during recent decades (Li et al 2016). Additionally, the variations of SChl are highly correlated with interannual climate variability in the tropical ocean (e.g. El Niño and Southern Oscillation, Indian Ocean Dipole, or Atlantic Niño) (Park et al 2011, Currie et al 2013, Chenillat et al 2021, Tian et al 2021a, Zhang et al 2022. Therefore, the internal climate modes may further modulate the recent trend in the tropical SChl by affecting inter-basin air-sea interactions and subsequent ocean dynamics, which has not been well investigated. Interestingly, the available longest record for SChl data reveals a robust decrease of SChl in the tropical ocean (figures 1 and S1). This phenomenon suggests that the reducing effect on the SChl due to anthropogenic activity-induced warming may have emerged in recent decades, although the IPV is still indicated in its negative phase. Furthermore, the future trend in SChl or net primary production has been widely investigated at a global scale using the CMIP simulations, but the observed trend in SChl in the entire tropical ocean and its related controlling mechanisms remain unclear. Whether the anthropogenic activityinduced SChl trend has emerged or not is a controversial issue at present.
In this study, we employed a hierarchy of climate model experiments from the CMIP6 to investigate the contributions of internal variability and external forcing to the observed decreasing trend in SChl (tables S1 and S2). Specifically, we used the Scenario Model Intercomparison Project (ScenarioMIP) and the Detection and Attribution Model Intercomparison Project (DAMIP) experiments to quantify the contribution of external forcing (GHG and aerosols). Meanwhile, we adopted the Decadal Climate Prediction Project (DCPP) experiments to evaluate the impacts of AMV and IPV on the recent trends in SChl in the tropical ocean (see table S1). The rest of the paper is organized as follows. Section 2 outlines the datasets and methodologies. Section 3 describes the observed and simulated chlorophyll patterns in the tropical ocean, and clarifies the controlling processes related to external forcing and internal climate variability. Section 4 provides the discussion and conclusions.

Observational data
SChl concentration data are from the GlobColor Project, derived from the merged satellite data products using a bio-optical model (Maritorena et al 2010). Because the coverage of the original data set from the GlobColor was from September 1997 to June 2021 when we began to analyze, we only analyzed the complete 23 annual cycles for our analysis from 1998 to 2020. To avoid interference from the extremely high productivity in the coastal region, we purposely excluded the effect from the coastal ecosystem by removing the SChl in the continental shelf (topography depth less than 200 m); thus, the effect of terrestrial organic materials in the coastal region due to anthropogenic activity cannot affect the decreasing trend in SChl over the open oceans (figure S2). Monthly SST data are from Extended Reconstructed Sea Surface Temperature, Version 5 (ERSSTv5) during 1850-2020 (Huang et al 2017), which are used to calculate the climate index.

CMIP6 experiments and data processing 2.2.1. Historical and SSP5-8.5 experiments
In this work, 14 ESMs from CMIP6 were used (table S2), consisting of the historical (1850-2014) and Shared Socioeconomic Pathways 8.5 (SSP5-8.5) simulations (2015-2100). Monthly mean SChl concentration, potential temperature, salinity, and net primary production were used for analyses. Additionally, mixed layer depth (MLD) is defined as the upper-ocean layer depth at which density changes relative to surface density by a threshold (∆ρ); here, we used ∆ρ = 0.03 kg m −3 to calculate MLD. Upper 200 m stratifications are defined as the potential density difference between 200 m depth and the surface (de Boyer Montégut 2004, Capotondi et al 2012. All data are interpolated onto 1 • × 1 • horizontal grids with the Climate Data Operators (CDOs). Due to the significant difference in modeled SChl concentration among different ESMs and observational data (figure S3), standardized SChl was first calculated for intermodel comparison in the historical, SSP5-8.5, and DAMIP experiments, respectively.

DAMIP experiments
To further quantify the contributions of GHG, aerosol forcing, and natural forcing to the SChl change, we adopted the DAMIP experiments from four available ESMs including chlorophyll outputs with 18 members (Gillett et al 2016), such as ACCESS-ESM1-5 (3 members), CanESM5 (10 members), CESM2 (2 members), and NorESM2-LM (3 members). We used outputs from three historical experiments: well-mixed greenhouse-gas-only (hist-GHG), anthropogenic-aerosol-only (hist-aer), and natural solar irradiance forcing-and volcanic forcing-only (hist-nat) simulations. Note that we first compute the ensemble mean within one single model, then calculate the multimodel ensemble mean (MME) with equal weights.

DCPP experiments
The DCPP is a coordinated multi-model investigation program into decadal climate prediction, predictability, and variability (Boer et al 2016). In the DCPP Component C, four experiments are adopted to examine the responses of the global climate to negative and positive phases of AMV (IPV) impacts, including idealized AMV (+), AMV (−), IPV (+), and IPV (−) pacemaker experiments ('+' represents the positive phase). Here, we adopted the experimental results from the Institute Pierre-Simon Laplace Climate Modeling Center coupled climate model (IPSL-CM6A-LR (Boucher et al 2020). Specifically, an idealized Pacific (Atlantic) pattern derived from the DCPP standard method was imposed into the daily climatological SST field and then run for 10 years. Outside the nudging regions, the rest of the oceanatmosphere interaction is fully coupled and free to evolve. All radiative forcing agents were held constant at the pre-industrial levels. Among them, IPV (+) and IPV (−) experiments are performed by 10 members, while AMV (+) and AMV (−) are performed by 50 members by adjusting initial conditions to have different atmospheric perturbations.  (Meehl et al 2021). The global mean SST time series was removed at each grid when it was used to calculate the AMV and IPV indices. Furthermore, we adopted a regression method proposed by Meehl et al (2021) to analyze the tropical response to internal climate variability in historical simulations. Specifically, a lagged regression of yearly mean SST anomalies on the simulated AMV and PDV indices by three years was performed in each model simulation. The choice of one to five lagged years in the calculations cannot significantly affect the regression pattern according to observational data (figure S4) and CMIP6 simulations as shown in Meehl et al (2021).

SChl trends in the observation and CMIP6 simulations
Observational data reveal that SChl in the Northern Hemisphere exhibits a decreasing trend. In contrast, SChl exhibits an increasing trend south of 20 • S, especially in the northern flank of the Subantarctic Front and the southeastern tropical Pacific. In the tropical oceans, SChl exhibits a consistently decreasing trend, with a remarkable decrease emerging in the northern tropical Pacific and the southern tropical Indian Ocean (figure 1(a)). Quantitatively, SChl over the tropical oceans exhibits a decreasing trend by −0.009 (−0.007) mg m −3 per decade (or −7.1% (−5.6%) per decade relative to the annual mean) when the 1998-2001 successive La Nina events are included (excluded) by 2020. To exclude the impact of extreme climate events on the trend, we performed a moving-year trend analysis to examine the sensitivity of the SChl trend to the start year and end year for calculating the linear trend. The results show that the SChl exhibits a decreasing trend in most selected periods, indicating that the calculation is insensitive to the selections of the end-point years (figure S1).
In the CMIP6 simulations during the period of 1998-2020, the MME also reveals a decreasing trend in the tropical Oceans ( figure 1(b)), especially in the northern tropical Pacific, the southern tropical Indian Ocean, and the southern equatorial Atlantic Ocean; these modeling results are coincident with the observed trend ( figure 1(a)). Generally, MME results can be used to represent the anthropogenic impact on the climate system by canceling out the internal climate variability among intermodel results. Note that the trend in SChl from the CMIP6 simulations is relatively smaller than that in the observations ( figure 1(b)), which is due to the potential contribution from extreme events to the SChl reduction. In addition, the available samples are limited in the CMIP6, which may further weaken the SChl trend calculation due to some overestimated SChl trends in some ESMs (e.g. CESM2-WACCM simulates an increasing trend in SChl). Over the longer term, SChl exhibits a relatively consistent reduction during 1850-2020 or 1998-2020. Under the global warming scenario, almost all models exhibit decreasing SChl trends during 2051-2100 (figure S5). Correspondingly, SST gradually increases, the mixed layer becomes shallow, and upper ocean stratification enhances, which is accompanied by a reduction in net primary production (figure S5). The consistent spatial patterns in the SChl trend between the observation and MME further confirm that surface warming due to anthropogenic activity may contribute to the recent decreasing trend in SChl.
SChl trends in several tropical sub-basins exhibit distinct features. SChl in the tropical Indian Ocean has exhibited a decreasing trend since the 1960s in association with the rapid warming in the upper Indian Ocean (Liu et al 2016, Roxy et al 2016. This feature is well captured by the observational SChl data, especially in the southern tropical and northwestern Indian Oceans ( figure 1(a)), and is well simulated by the ESMs in the northwestern tropical Indian Ocean ( figure 1(b)). Additionally, interannual to decadal variabilities are remarkable in SChl (figure S6), which may be related to extreme Indian Ocean Dipole events during 2004-2007(Ma et al 2015. In the tropical Pacific, ESMs well simulate the decreasing trend in SChl, albeit with a relatively weak decreasing trend in the eastern equatorial Pacific (figures 1(d) and S6), where the persistent 1998-2001 La Niña event significantly contributes to the decreasing trend in SChl by 22%; correspondingly, the observed negative trend is larger than that in the ESM simulations, indicating that internal climate variability tends to modulate the SChl trend. Interestingly, the SChl trend in the MME during 1998-2020 in the tropical Pacific is comparable with the observational trend (figure 1(e)), suggesting that anthropogenic influence may efficiently reduce SChl in the longer term. In the tropical Atlantic, the decreasing trend in SChl from observational data is also remarkable in the MME time series, which emerged at the beginning of the 1990s ( figure S6). This reduced SChl in the northern tropical Atlantic may be related to the positive phase of the AMV or due to the anthropogenic effect (Mann et al 2021).
Overall, although MME underestimates the recent decreasing trend in the tropical ocean (figure 1(g)), a consistent trend pattern in the SChl emerges in the three ocean basins. In addition, the pattern of ToE also exhibits that the signals of anthropogenic effects in the northern tropical Pacific and Atlantic Ocean are less than 30 years (figure S7). These pieces of evidence suggest that the impact of anthropogenic activity may have emerged.

Contribution from external forcing to the SChl trend: DAMIP experiments
To further quantify the contribution from external forcing, we adopted DAMIP analogous 'single forcing' experiments (Methods) to examine the influences of historical GHGs, industrial aerosols, and natural forcing on the recent SChl in the tropical ocean. Because the available samples for ESMs are relatively small, the change in SChl from 1998 to 2014 exhibits extremely substantial uncertainty (figure S8), we examined the relatively long-term trend from 1950 to 2014 to illustrate the possible impact of diverse external forcing. Over the entire tropical ocean, GHG forcing (hist-GHG) tended to reduce SChl from 1950, while the industrial aerosols (hist-aer) acted to increase SChl during this period (figures 2(d) and S8). Additionally, SChl has decreased under natural forcing due to volcanic eruption and solar forcing since the 1950s, which plays a secondary role in reducing SChl (figure 2(c)).
It is well known that GHG-induced surface warming tends to enhance the upper ocean stratification, leading to a decrease in SChl. However, the negative effect of GHG on chlorophyll only emerges in the western-central equatorial Pacific, the eastern equatorial Atlantic, and the northwestern Indian Ocean. This spatial pattern is distinct from the MME one from all external forcing, which can be partially attributed to small samples with a large model spread (i.e. only four ESMs are used). SChl in MME from 14 ESMs exhibits an evenly decreasing trend in the tropical ocean during 2015-2100 under SSP5-8.5 ( figure  S9).
The impacts of the industrial aerosol (hist-aer) on SChl are diverse among different ocean basins. Specifically, under the industrial aerosol forcing, SChl in the western tropical Pacific exhibited a rapid increase from 1950 to 2014. A slight decrease in the tropical Atlantic and the northern Indian Oceans can be seen in figure 2(b). Because the distribution of aerosols is heterogeneous, this diverse response of chlorophyll can be related to a reduction in sulfate aerosol optical depth over the United State of America and Europe but an increase over China and India during recent decades (Klimont et al 2013). Therefore, the counterbalancing influences from aerosols in the western tropical Pacific and tropical Atlantic Ocean led to a relatively slow increase in SChl levels during recent decades (figures 1(a) and 2(d)). In the equatorial Indian Ocean and eastern tropical Pacific, the change in SChl is relatively tiny under industrial aerosol forcing (figures 2(a) and (c)). Note that industrial aerosol forcing not only affects ocean dynamics (e.g. surface cooling) but also alleviates the nutrient limitation in the oceans by exerting a fertilizing effect (Wang et al 2015). Clearly, there exist interplays between ocean physics and biogeochemistry, and a more comprehensive understanding of the role played by industrial aerosol forcing in SChl change is beyond the scope of this study.
As shown in figure 2(d), the GHG effect plays a crucial role in determining the decreasing trend in chlorophyll over the tropical ocean. At the same time, industrial aerosol acts to increase SChl due to its surface cooling effect, especially in the western tropical Pacific (figures 2(b) and S8). It is noted that a substantial uncertainty in SChl projection can exist among these DAMIP simulations (figure S8). For example, SChl increases in the tropical ocean are evident in the CanESM5 simulation under GHG forcing, which coincidently corresponds to the slight increase under the SSP5-8.5 scenario (figure S10). This opposite response of SChl is rare in the CMIP6 simulations, indicating that the diverse responses of the marine ecosystem need to be investigated in the future.

Contribution of the internal climate variability to the SChl change
As presented in the previous section, the GHG forcing reduces the SChl, while industrial aerosol forcing has increased the SChl in recent decades. At this stage, the contribution of the internal variability to the SChl trend remains unclear. Due to the prominent roles played by the IPV and AMV in the internal climate changes on decadal timescales, we adopted a statistical method to examine their effects on the SChl trend under historical forcing. Then, we used the DCPP idealized pacemaker experiments to separate the impacts of IPV and AMV on SST and SChl changes. Figure 3 shows the lagged regression pattern of the SST and SChl anomalies on the simulated AMV and IPV indices by three years, averaged over the 14 ESM historical runs (figures S11-S14). Consistent with a previous study (Meehl et al 2021), positive SST anomalies emerge in the equatorial and southern Atlantic Ocean when IPV takes a lead by three years (figure 3(c)). However, the northern Atlantic exhibits a slight warming or even cooling pattern. Specifically, some ESMs (e.g. CESM2, GFDL-ESM4, MIROC-ES2L) can simulate a positive SST response in the northern tropical Atlantic as found in Meehl et al (2021) using CESM1, indicating the intermodel uncertainty is still substantial in the northern tropical Atlantic response to Pacific influence (figure S12). Additionally, the Indian Ocean exhibits consistent warming when IPV leads by three years; the tropical Pacific warming leads to an enhanced Walker Circulation and intensifies the Indonesia Throughflow, which subsequently contributes to the heat uptake in association with the change in wind stress curl (England et al 2014, Maher et al 2018. Therefore, the positive phase of IPV leads to tropical warming, further contributing to a decreasing SChl through stabilizing density stratification in the upper ocean ( figure 3(e)). In turn, it is further demonstrated that the recent decreasing trend in SChl is not caused by the negative phase of the IPV ( figure 3(a), the shaded region).

The SChl-SST relationship analyzed by statistical method
As for the response to the positive phase of AMV, the tropical Pacific and southern Indian exhibit a cold state, while the northern tropical Indian Ocean exhibits a warm state, albeit with large model uncertainty in the Northern Hemisphere (figure 3(d)). Three ESMs (CESM2, CESM2-WACCM, and NorESM2-MM) simulate the well-defined IPV phase in the Pacific. However, other ESMs tend to simulate consistent cooling in the tropical and subtropical Oceans or NH warming and SH cooling (figure S11). Ruprich-Robert et al (2021) suggested that inter-model uncertainty is mainly caused by different amounts of moist static energy injection from the tropical Atlantic surface into the upper troposphere. Meanwhile, the southern tropical Pacific still exhibits consistent cooling due to enhanced Pacific trade winds (Li et al 2016, Meehl et al 2021. Correspondingly, the SChl exhibits large inter-model uncertainty in response to the AMV forcing. Most ESMs exhibit an increase in SChl in the tropical Indian Ocean and the western tropical Pacific. A remarkable increase in SChl emerges in the southern tropical Pacific associated with an SST cooling (figures 3(f), S11 and S13). Thus, the impact of the positive phase of AMV may hardly affect the recent SChl decreasing trend over the entire tropical ocean (figures 3(f) and S15), but can exert a reducing effect on SChl in the tropical Atlantic and Indian Ocean. Further analyses of all the possible combinations of different phases in AMV and IPV suggest that the outof-phase between IPV and AMV accounts for 58.3% of all conditions. The positive phase of AMV and the negative phase of IPV act to induce surface warming in the tropical ocean, which is also accompanied by a reduction in SChl (figure not shown). This condition well fits for the current stage, which indicates that the combined effects of positive AMV and negative IPV phases indeed contribute to the decreasing SChl, especially in the northern Atlantic and Indian Ocean being supported by the IPSL-CM6A-LR pacemaker experiments as follows.

DCPP experiments
We further analyzed the idealized IPSL-CM6A-LR experiments described in the methodology section. Spatial patterns of SChl and SST responses to AMV and IPV are shown in figure 4, which are similar to the regression method as shown in figure 3. As the response to the positive phase of AMV, SST in the Pacific exhibits a horseshoe-shaped pattern with a cold anomaly center located in the tropical Pacific and warm anomalies located in the western subtropical Pacific from the pacemaker experiments. Previous studies have demonstrated that AMV-induced SST warming can lead to enhanced vertical motion over the tropical Atlantic and anomalous descent over the tropical Pacific with the enhanced easterly wind (Meehl et al 2021). The significant SST cooling in the equatorial Pacific exceeds the 70% significance level, associated with an enhanced easterly wind and a decrease in Sea Surface Height (SSH) (figure S16). Furthermore, the nurtricline was shoaled due to the enhanced Ekman divergence, which subsequently increases SChl, albeit with large model uncertainty (figures 4(c) and (g)). Additionally, the response of the Indian Ocean SST exhibits a warming pattern in the tropical and northern Indian Ocean (figure 4(a)) because AMV-induced SST warming drives easterly wind anomalies over the Indo-western Pacific as Kelvin waves. This wind effect further contributes to SST warming through wind-evaporation-SST feedback (Li et al 2016). However, the response of SChl to the AMV-induced SST warming is not pronounced in the tropical Indian Ocean (figure 4(c)), which further suggests that the internal climate variability from the Atlantic Ocean cannot affect the SChl trend in the Indian Ocean. As a whole, the positive phase of the AMV plays a relatively weak role in determining SChl trend over the entire tropical oceans (figure 4(g)).
Consistent with the result from the regression method, the positive phase of IPV tends to induce an SST warming in the entire tropical Oceans ( figure 4(b)), except for the southeastern tropical Indian Ocean. During the positive phase of IPV, surface warming over the eastern tropical Pacific contributes to the enhanced upward vertical motion, and anomalous descent branches over the tropical Atlantic and Indo-western tropical Pacific basin (figure S16). This forcing further increases SSH over the western tropical Indian and Atlantic oceans, which corresponds to a depression of nutricline (thermocline). Thus, the positive phase of IPV tends to reduce the SChl over the tropical oceans (figure 4(d)). It is noted that SChl in all sub-basins exhibits a substantial decrease in its response to the positive IPV phase ( figure 4(h)). This response, in turn, suggests that the recent negative phase of IPV cannot explain the recent decreasing trend in SChl.

Discussion
As presented from our analysis, GHG-induced warming acts to dominate the current decreasing SChl trend in the tropical ocean, while internal variability plays a weak or even opposite role in facilitating the SChl trend. To further demonstrate this conclusion, the probability distributions of SST and SChl trends based on the internal variability condition and climate change condition are shown in figures 5(a) and (b). Here, the internal variability is obtained from the historical run when the quadratic linear trends are removed (similar distributions are found in the pre-industrial simulations) and climate change conditions are from the SSP5-8.5 scenario. Probability distributions of the SST trend show that the tropical SST trend during 1998-2020 fits well outside the 95% confidence bound of the internal variability and at the lower end of the probability distribution with climate change included. A similar condition occurs in the SChl trend, which fits well outside the 95% confidence bound of the internal variability and is well located in the distribution of climate change (i.e. GHG effects) ( figure 5(b)). This result demonstrates that the decreasing SChl trend in combination with SST warming during 1998-2020 can be well explained by anthropogenic activity. Nonetheless, there are some caveats that need to be considered in our analysis. Because of the limitation of samples from DAMIP, the more accurate estimation of GHG contribution is difficult at the current stage due to model biases. In addition, the impacts of GHG warming on phytoplankton growth have not been parameterized in the current ocean biogeochemical model, which may further affect the projection of the SChl trend.
Previous studies have widely used various CMIP experiments with different scenarios and extensive ensemble experiments to detect the ToE of anthropogenic signals, suggesting that the ToE of SST and anthropogenic carbon are 20-30 years. However, for the SChl and salinity, their emergence timescales are longer (50+ years), which is highly dependent on the performance of ESMs (Elsworth et al 2020). We also noted that the ToE is less robust across the ESMs, and more sensitive to the forcing scenario considered (Schlunegger et al 2020). Therefore, the emergence times for physical and biogeochemical components in CMIP6 experiments still need to be examined (Schlunegger et al 2019, Ying et al 2022, which can identify the anthropogenic signals discussed in this study. As for the underestimation of the SChl trend in MME of CMIP6, we suggest some possible reasons that are responsible for this issue. (1) We used the standardized chlorophyll field to remove the simulated differences in chlorophyll among ESMs and used the equal weight to calculate MME, which may lead to an underestimation of results from wellperformed ESMs (e.g. GFDL-ESM4).
(2) During 2013-2020, marine heatwaves frequently occurred in the northern subtropical Pacific and further extended equatorward, which made a contribution to a decrease in SChl in the tropical Pacific (Laufkötter et al 2020). A similar condition occurs in the 1998 La Nina event when a phytoplankton bloom occurs in the eastern Pacific. Therefore, the impacts of extreme events on the relatively short period of observed SChl cannot be ignored (Zhang et al 2022).
The decreasing trend in SChl under anthropogenic activity-induced GHG warming can further affect climate change in two ways: (1) the decreased chlorophyll enhances the penetration of solar radiation into the subsurface layer in combination with the shoaling of the mixed layer (Zhang et al 2018), which further affects the heat redistribution and watermass transformation (Gnanadesikan and Figure 5. (a) and (b) Probability distribution (PDF) curves for tropical SST and SChl due to internal variability (blue curves) and external forcing (red curves); (c) schematic diagram of the physical mechanism responsible for the decreasing trend in SChl during 1998-2020. Specifically, the observed tropical chlorophyll trend is partially attributed to anthropogenic greenhouse gas (GHG) forcing. In contrast, internal climate variability (AMV and IPV) and industrial aerosols play offsetting roles in facilitating an increase in chlorophyll during 1998-2020. Negative phase of the IPV acts to induce an SST cooling over the tropical oceans, which further increases SChl; as for the positive phase of the AMV, positive SST anomalies over the northern Atlantic lead to an SST cooling and enhance the trade winds over the equatorial Pacific, which subsequently leads to a strengthening of the upwelling and a slight increase in chlorophyll.

Anderson 2009).
(2) The decreased chlorophyll corresponds to a reduction in phytoplankton biomass, which contributes to less uptake of anthropogenic carbon (Kwiatkowski et al 2017). Therefore, an indepth analysis needs to be conducted in the future to elucidate the effects of decreasing SChl on the tropical climate under a warmer climate scenario.

Summary
Available satellite data reveal a decreasing trend in SChl by −7.1% per decade over the tropical ocean. Here, we adopted a series of coupled model experiments from CMIP6, demonstrating that the observed tropical chlorophyll trend is partially attributed to GHG forcing (summarized in figure 5(c)). In contrast, internal climate variability (AMV and IPV) and industrial aerosols play offsetting or weak roles in facilitating an increase in chlorophyll during 1998-2020. By adopting the CMIP6 historical simulations and the DCPP experimental results, we found that the negative phase of the IPV acts to induce an SST cooling over the tropical oceans, which further increases SChl; as for the positive phase of the AMV, the positive SST anomalies over the northern Atlantic lead to an SST cooling and enhance the trade wind over the equatorial Pacific, which subsequently leads to a strengthened upwelling and a slight increase in chlorophyll. Therefore, the internal climate modes or variability cannot fully explain the recent chlorophyll trend over the tropical oceans. In contrast, the DAMIP and SSP5-8.5 experiments further demonstrate that GHG-induced warming leads to an enhanced stratification in the upper ocean and consequently a decrease in SChl over the tropical oceans, albeit with a small positive contribution from the aerosol cooling forcing.