Impacts of Physical and Biological Processes on Spatial and Temporal Variability of Particulate Organic Carbon in the North Pacific Ocean during 2003–2017

The North Pacific Ocean is a significant carbon sink region, but little is known about the dynamics of particulate organic carbon (POC) and the influences of physical and biological processes in this region at the basin scale. Here, we analysed high-resolution surface POC data derived from MODIS-Aqua during 2003–2017, together with satellite-derived sea surface chlorophyll and temperature (SST). There are large spatial and temporal variations in surface POC in the North Pacific. Surface POC is much lower in the subtropical region (<50 mg m−3) than in the subarctic region (>100 mg m−3), primarily resulting from the south-to-north variability in biological production. Our analyses show significant seasonal and interannual variability in surface POC. In particular, there is one peak in winter-spring in the western subtropical region and two peaks in late spring and fall in the western subarctic region. Surface POC is positively correlated with chlorophyll (r = ~1) and negatively correlated with SST (r = ~−0.45, P < 0.001) south of 45°N, indicating the strong influence of physically driven biological activity on the temporal variability of POC in the subtropical region. There is a significantly positive but relatively lower correlation coefficient (0.6–0.8) between POC and chlorophyll and an overall non-significantly positive correlation between POC and SST north of 45°N, reflecting the reduction in the POC standing stock due to the fast sinking of large particles. The climate modes of the Pacific Decadal Oscillation, El Niño–Southern Oscillation and North Pacific Gyre Oscillation have large impacts on POC in various seasons in the subtropical region and weak influences in the subarctic region. Surface POC was anomalously high after 2013 (increased by ~15%) across the basin, which might be the result of complex interactions of physical and biological processes associated with an anomalous warming event (the Blob).


Regional Setting
The North Pacific Ocean consists of two distinctive regions: the subtropical region and the subarctic region (Fig. 1). The former can be represented by the North Pacific Subtropical Gyre (NPSG), which is influenced by two major currents: the North Pacific Current to the north and North Equatorial Current to the south 27 . In addition, there exists the Kuroshio in the west and the California Current (CC) in the east. The California Current is the predominant flow in the eastern portion of the NPSG 28 , which has weaker physical variability than the western portion of the NPSG 27 . Overall, the surface water in the NPSG is known as an "ocean desert" due to the conditions of strong stratification and severe nutrient depletion in the upper layer 29 .
The subarctic Pacific, which is known as a high nutrient low chlorophyll (HNLC) region, consists of two major cyclonic gyres: the Western Subarctic Gyre (WSG) to the west and the Alaska Gyre (AG) to the east. The WSG is fed by cold nutrient-rich coastal currents, such as the southward Oyashio and the eastward Okhotsk Current 30 ,

Seasonal variability.
To examine the seasonal variability of chlorophyll and POC in the subarctic and subtropical regions, we focus on two representative latitude bands (i.e., 45-55°N and 25-35°N), analyse the seasonal mean and seasonal variability (coefficient of variation, CV) and compare the differences between the west and the east (Table 1). Overall, both chlorophyll and POC show modest seasonality in the subarctic region, with two peaks (i.e., in late spring and fall) to the west of 155°W (Fig. 3a). Clearly, the seasonality of chlorophyll is stronger in the west (33% for CV) than in the east (10% for CV) ( Table 1). POC shows a similar pattern, with larger CV in the west (19%) than in the east (6.7%). The eastern subarctic Pacific has small ranges of both chlorophyll (0.33~0.4 mg m −3 ) and POC (89-98 mg m −3 ), and as a result, the POC:Chl ratio (257-276 g:g) has a narrow range during all four seasons ( Table 1). The POC:Chl ratio in the west is lower in summer-fall (~200 g:g) than in winter-spring (~270 g:g, Fig. 3b).
For the subtropical band, the seasonal variabilities of chlorophyll and POC are more pronounced (Fig. 4a), with higher CVs in this region than those in the subarctic region although the POC and chlorophyll concentrations are much lower in the former than in the latter (Table 1). Seasonal variability is also stronger in the west than in the east for both POC (CV = 29% vs. 12%) and chlorophyll (CV = 48% vs. 22%) in the subtropical region. Over the seasonal cycle, the western section has higher POC values (~50 mg m −3 ) in winter-spring than in summer-fall (~31 mg m −3 ) (Table 1). In contrast, the POC:Chl ratio in the west is lower (~400 g:g) in winter-spring than in summer-fall (~500 g:g) (Fig. 4b). The seasonal variability of the POC:Chl ratio is less pronounced in the eastern subtropical Pacific than in the western subtropical Pacific (Table 1), with narrow ranges for POC (30-40 mg m −3 ) and the POC:Chl ratio (400-500 g:g).

Interannual variability.
We also evaluate the interannual variability of the SST, POC, chlorophyll and POC:Chl ratio over the two bands. Figure 5a illustrates that the SST variation is much larger in the east (with anomalies varying from −1 °C to +1.5 °C) than in the west (from −0.6 °C to +0.4 °C) in the subarctic band There are clear differences in the SST anomalies between the west and the east in the subtropical region (25-35°N), e.g., there are negative anomalies during 2011-2014 in the west and during 2006-2012 in the east (Fig. 6a). Seemingly, there is an eastward propagation of positive SST anomalies, starting in 2008/9 along 150°E and reaching 130°W in 2015. Overall, the POC and chlorophyll anomalies are almost opposite to the SST anomalies, i.e., higher POC (chlorophyll) values corresponded to colder SSTs (Fig. 6b). However, the POC:Chl ratio shows a similar interannual variation as the SST anomaly ( Fig. 6c) but has some degree of time lag in the west. Almost opposite to the POC anomaly, the highest POC:Chl ratio is found during 2008-2010 and 2012-2013, and the lowest POC:Chl ratio is found during 2014-2015.

Discussion
Spatial variability. An earlier study showed large spatial variability in primary production in the North Pacific 32 , i.e., an overall increase from the south to the north and an increasing trend from the east to the west, similar to the spatial pattern of POC (Fig. 2). The high production in the subarctic region is due to nutrient-rich conditions 17,29,31 . The zonal variability of primary production in the subarctic Pacific is suspected to be largely attributed to iron availability, i.e., iron is more available in the WSG than in the AG 30,33 .
The standing stock of POC is a balance between production and loss processes (e.g., the export of POC). There is evidence of more large-sized phytoplankton species (e.g., diatoms) in the subarctic region than in the subtropical region 32,34 . In general, particles generated from large phytoplankton and their pellets have relatively fast sinking velocities 4,34-37 , which can reduce the POC standing stock in the surface water, causing a relatively small POC:Chl ratio. Indeed, a field study showed a relatively small POC:Chl ratio (160~170 g:g) in the western subarctic region 38 . Overall, the observed POC:Chl ratios are close to the value (~200 g:g) representing the predominance of local biological activities in POC production 39,40 , indicating that biological processes are largely responsible for the higher levels of surface POC observed in the subarctic region.    www.nature.com/scientificreports www.nature.com/scientificreports/ The subtropical region reveals much lower levels of POC (<50 mg m −3 ), which are results of lower biological production 32 . A few studies indicate that chlorophyll variability in the subtropical Pacific is mainly driven by physiology, i.e., the small phytoplankton in this region have a high phytoplankton carbon to chlorophyll (C:Chl) ratio (150-250 g:g in the subtropical region, compared with 50-100 g:g for larger phytoplankton in the subarctic region) associated with low growth rates due to nitrate limitation and light acclimation [41][42][43][44] . Such a high C:Chl ratio in phytoplankton can be partly responsible for the increased POC:Chl ratios (300-600 g:g) in the subtropical region. On the other hand, the high POC:Chl ratios can also be attributed to the slow downward flux of POC in the surface water due to the relatively small sizes of phytoplankton and their pellets suspending in the surface water of the subtropical region 4,5,35 .

Seasonal variability.
Previous studies showed a clear seasonality in satellite-derived primary production in the North Pacific, i.e., higher values in summer than in winter in the subarctic Pacific but higher values in winter than in summer in the subtropical Pacific 32,43 . Notably, this seasonality is similar to the seasonality of the satellite-derived POC in our study (Figs 3 and 4). Furthermore, we found similar seasonal variation between in situ primary production and satellite-derived POC in the northwest Pacific (Fig. 7). The agreement in the seasonality of POC and primary production indicates that biological production is largely responsible for the seasonal variability of POC.
The seasonality of POC is stronger to the west than to the east in the subarctic Pacific (Table 1). An earlier study showed much stronger seasonal variation of primary productivity in the western subarctic Pacific than in the eastern subarctic Pacific 45 , which was due to the larger seasonal variation in the iron supply in the west than in the east 30,46 . Apart from biological regulation, the relatively high POC:Chl ratio in winter in the west may indicate a supply of detrital POC, since there is evidence that more Okhotsk seawater flows into the Northwest Pacific through the southern Kuril Straits during the winter season 47 . On the other hand, the higher winter POC:Chl ratio is also probably associated with the predominance of small phytoplankton due to the low temperature and light limitation 48 .
The POC in the subtropical Pacific shows remarkable seasonal variability, with a pattern opposite to that of the POC:Chl ratio, i.e., a lower POC and higher POC:Chl ratio occur in summer-fall than in winter-spring. An earlier study demonstrated that the highest phytoplankton C:Chl ratios in this region typically occur in summer and fall 41 , suggesting that, apart from reductions in biological production, there might be further reductions in chlorophyll due to physiological responses. Based on the seasonal increases in POC (~20 mg m −3 ) and chlorophyll (~0.1 mg m −3 ) from summer-fall to winter-spring, we found that the change in the POC:Chl ratio is rather low (~200 g:g), indicating that biological processes are largely responsible for the seasonal increase in POC. www.nature.com/scientificreports www.nature.com/scientificreports/ Interannual variability. Our analyses show a significantly positive correlation (averaged r = ~0.75, P < 0.001) between the anomalies of chlorophyll and POC (Fig. 8a), suggesting that the interannual variability of POC in the North Pacific is primarily linked with biological processes. Interestingly, the correlation coefficient is smaller (r = 0.6~0.8) in the productive regions (e.g., subarctic region, coastal areas and marginal seas) than in the oligotrophic subtropical region (r = ~1), indicating that other processes influence POC variability. Indeed, there is evidence that large particles sink relatively fast in the subarctic region, particularly during seasons of high productivity 4,35 , which can cause a reduction in the POC standing stock in the surface water, thus altering its relationship with chlorophyll. There is a significantly negative correlation (averaged r = ~-0.45, P < 0.001) between chlorophyll/POC and SST south of 45°N (Fig. 8b,c), implying that physical processes play a large role in regulating the interannual variability of primary production and POC. Warmer SSTs reflect weaker vertical mixing, which causes poor nutrient conditions, lower phytoplankton production 49,50 and thus less POC.
The relationship of POC with SST varies in the subarctic region, with a positive correlation in the majority of the basin. In general, rising temperatures and improved light availability (due to shoaling MLD) are beneficial for phytoplankton growth at high latitudes 38,50,51 . However, the overall non-significantly positive correlation between POC and SST and the relatively weak correlation between POC and chlorophyll in the subarctic region indicate that changes in the POC stock may not reflect changes in biological activity. First, the POC standing stock in the surface water can be reduced by downward export of large POC in the subarctic region 4,35 . Second, current systems can transport high-POC waters from nearshore to offshore and open ocean regions 52,53 , which is largely influenced by both local (e.g., winds and fresh water fluxes) and remote (e.g., basin-scale circulation) forcings 54 . These processes may have individual change patterns over an interannual time scale, thus altering the relationship between POC and biological production.
impacts of climate forcings. It is generally recognized that physical and biogeochemical processes are largely influenced by climate forcings in the North Pacific 10,18,55 . Here, we further analyse the interannual variability of POC and its relationships with climate modes (ENSO, PDO, NPGO) in different regions. Figure 9 illustrates large differences in the POC-climate mode relationship between the subarctic and subtropical regions. Based on the correlation coefficient estimated with data over 2003-2013 (excluding the influence of the Blob after 2014), it is clear that the POC anomaly showed no relationship with the climate modes in the subarctic region (Table 2). Similarly, an earlier analysis based on SeaWiFS data (over 1997-2010) revealed that chlorophyll had an insignificant correlation with PDO and NPGO in the subarctic Pacific 56 . However, a more recent study using SeaWiFS/ MODIS data over a longer period (1997-2012) showed various responses in different seasons, e.g., increased chlorophyll in winter but decreased chlorophyll in spring and summer during the positive phase of PDO in the On the other hand, POC had significant correlations with climate modes in most seasons in the subtropical region ( Table 2). As a result, POC was significantly correlated with climate modes in all seasons, and the strongest correlation was with PDO in both the west (r = 0.38, P < 0.001) and the east (r = −0.56, P < 0.001). It appears that the correlation of POC was stronger with NPGO than with ENSO in the west but stronger with ENSO in the east. Clearly, there were opposite POC-climate relationships between the west and the east, e.g., POC was higher in the west but lower in the east during the warm phases of PDO and ENSO (see Fig. 9b), suggesting different  www.nature.com/scientificreports www.nature.com/scientificreports/ physical and/or biological responses in these regions. There is evidence of stronger westerly wind in the west but northward advection of warm air in the east during the warm phases of PDO and ENSO 57 , resulting in higher levels of nutrients in the west than in the east 9,10 . Indeed, field studies showed lower rates of primary production in the eastern subtropical Pacific during the warm phases of PDO 58 and ENSO 58-60 , as a consequence of increased  www.nature.com/scientificreports www.nature.com/scientificreports/ stratification and reduced nutrient mixing 61 . Similarly, the POC-NPGO relationship was also opposite between the west and the east (Table 2), which can be ascribed to downwelling-favourable conditions of the Subtropical Gyre and upwelling-favourable conditions of the California Current during the positive phases of NPGO 57,62 .
A number of studies have reported that the record-high SST anomaly (the Blob) triggered an unprecedented bloom of warm-water planktonic species in the coastal regions from southern California to the Aleutian Islands during 2014-2015 20,63,64 , and our analyses show an overall increase in both chlorophyll and POC during the same period in the north Pacific (Figs 5b and 6b). Here, we further assess the basin-scale anomalies of chlorophyll and POC (i.e., deviation from the mean over 2003-2017) during this extreme warming event. Given the fact of inconsistent responses of chlorophyll to climate forcing in different seasons 18 , we compare the anomalies of chlorophyll and POC with the SST anomaly in the main seasons (spring, summer and fall) (Fig. 10). Clearly, SST shows an overall increase in the subarctic region, except for in summer in the west, but a decrease in some  (170°E-180°E, 25°N-35°N and 135°W-125°W, 25°N-35°N). The scales of POC in (b) are different from those in (a) (smaller). All time-series are smoothed with a 13-month running mean.  www.nature.com/scientificreports www.nature.com/scientificreports/ parts of the subtropical region (Fig. 10a,d,g). The positive SST anomaly in 2014-2015, mainly occurring in the Northeast Pacific, has been described elsewhere 19,65,66 . Clearly, the spatial patterns of warming and the changes in chlorophyll and POC are not similar. On the other hand, there is an overall similarity between the spatial patterns of the chlorophyll and POC anomalies, except for in fall. In particular, both chlorophyll and POC increase in most sections of the subarctic region in spring and summer but decrease in the eastern subtropical region in spring. The increase in POC is greatest in fall in terms of magnitude and spatial coverage, with a large area showing a >15% increase, while the change in chlorophyll in fall shows no similarity.
There is evidence of strengthened northwesterly winds in the western-central part of the subtropical region in winter-spring during the Blob 65 , which could increase the nutrient supply due to enhanced vertical mixing and horizontal transport 67,68 , thus increasing the primary production and POC level. For the subarctic region, while the overall increase in chlorophyll may be associated with warming that can enhance biological production in this region 51,69 , the insignificant correlation between chlorophyll and SST (Fig. 8b) indicates that there may be other mechanisms causing the enhanced biological activities and/or elevated POC. A field study reported increased phytoplankton biomass in the eastern subarctic region associated with the Blob, with a significant increase in chlorophyll (~100%, relative to 2013) during 2014, and it is speculated that this increase might result from improved light and nutrient conditions 21 . In addition, there is evidence of extreme sea ice melting event in the Okhotsk Sea in 2015 70 , which may supply iron into the surface water and thus enhance primary production in the subarctic region [71][72][73] . In addition, there might be increased secondary production through grazing in association with warming 74-76 , leading to an increase in POC in the subarctic region. Indeed, a recent study showed evidence of increased zooplankton biomass in the Gulf of Alaska during 2014-2015 77 . On the other hand, the overall increasing trend in POC from spring to fall (Fig. 10c,f,i) may reflect the accumulation of detrital POC through the food web following the spring phytoplankton bloom 78,79 . In summary, there may be complex physical processes associated with the Blob event, which could have impacts on biogeochemical processes and ecosystem dynamics, with implications for the sources/sinks of POC in the north Pacific. Therefore, process-based physical-ecosystem models are needed in future studies to quantify the key physical and biogeochemical processes that regulate the oceanic carbon cycle due to changes in environmental conditions.

conclusions
We analysed the spatial and temporal variability of surface POC in the North Pacific Ocean, together with surface chlorophyll and SST. Overall, the POC level varies by a factor of two from the subtropical to the subarctic region, i.e., there are much higher values of POC in the subarctic region (>100 mg m −3 ) than in the subtropical region (<50 mg m −3 ), similar to the spatial distributions of chlorophyll and primary production. However, the POC:Chl ratio is lower in the subarctic region (or when productivity is high), which might be partly attributed to the decreased phytoplankton C:Chl ratio and rapid sinking of large-size particles. The seasonality of POC manifests large meridional and zonal differences, with one peak in winter-spring in the western subtropical region and two peaks in late spring and fall in the western subarctic region. Surface POC has a significantly positive correlation with chlorophyll (r = ~1) and a negative correlation with SST (r = ~−0.45, P < 0.001) south of 45°N, indicating that physically driven biological activity is responsible for the temporal variability of POC in the subtropical region. There is also a significantly positive but lower correlation coefficient (0.6-0.8) between POC and www.nature.com/scientificreports www.nature.com/scientificreports/ chlorophyll and an overall non-significantly positive correlation between POC and SST north of 45°N, which reflects the reduction of the POC standing stock in the surface water due to the faster sinking of large particles. Our analyses demonstrate that the climate modes of PDO, ENSO and NPGO have large impacts on the POC dynamics in the subtropical region, showing opposite correlations between the west and the east. The Blob may have complex influences on physical and biological processes, leading to an increase in POC (by ~15%) during 2014-2015 in the majority of the basin.

Data and Methods
Satellite data descriptions. In this study, we used MODIS-Aqua-derived chlorophyll-a, POC and SST over the period of 2003-2017. The level 3 chlorophyll-a data were derived from the standard NASA algorithm, i.e., the OC3M band ratio algorithm from O'Reilly et al. 80 merged with the colour index (CI) algorithm from Hu et al. 81 . The CI algorithm was a three-band reflectance (R rs ) difference algorithm, using the difference between the R rs in the green band and a reference formed linearly between R rs in the blue and red bands, i.e.: where λ blue, λ green and λ red represent the instrument-specific wavelengths closet to 443, 555 and 670 nm, respectively. The chlorophyll derived from the CI algorithm is defined as: (2) 10 The OC3M algorithm is a fourth-order polynomial relationship between the ratio of R rs and chlorophyll-a, i.e.: where a 0 -a 4 are sensor-specific factors, which can be found in the algorithm description file of NASA (https:// oceancolor.gsfc.nasa.gov/atbd/chlor_a/). The OC3M algorithm was used when the chlorophyll retrievals were above 0.2 mg m −3 , and the CI algorithm was used for clearer water (chlorophyll < 0.15 mg m −3 ). In between these values, the CI and OC3M algorithms were both used by a weighted approach. The level 3 POC data were based on the algorithm from Stramski et al. 22 , an empirical relationship derived from in situ measurements of POC and blue-to-green band ratios of spectral remotely sensed reflectances: Match-up analysis. Satellite-derived chlorophyll product has been widely validated and showed quite good performance 26,[82][83][84] . There were also some validations of satellite POC 23,24,26,85  We downloaded the level 3 daily MODIS Aqua POC products (4 km spatial resolution) from https://oceandata.sci.gsfc.nasa.gov/MODIS-Aqua/Mapped/Daily/4km/. Satellite POC data were paired with in situ data with a spatial and temporal window 23,26,82 . The spatial window in our study was set as a 3 × 3 pixel box centred on the location of the in situ measurements. The box was only selected when at least 6 of 9 pixels were valid and the coefficient of variation was less than 15%. For the temporal window, we first tried the approach of Swirgon and Stramska's match-up 23 , i.e., the time difference between the satellite passing time (13:30 P.M local time) and the in situ sampling time was less than 2 h, which yielded only 19 match-ups. Following a more recent validation work on satellite-derived chlorophyll and POC 26 , we applied a temporal window of 24 h and found 41 match-ups that were from 2003 to 2015.
We adopted the Mode II reduced major axis regression approach following the previous validation studies 23,26,82 , accounting for the uncertainties in both in situ and satellite data, which showed a significantly good regression (R 2 = 0.67, P < 0.001), with a slope of 1.12 (Fig. 11). The differences between in situ and satellite-derived POC were also evaluated by standard methods 82 , namely, the mean normalized bias (P bias ), the root mean square error (RMSE) and the mean absolute percentage error (MPE), which were calculated as follows: (2019)  time-series analyses. We obtained MODIS-Aqua monthly means of chlorophyll-a, POC and SST over the period of 2003-2017 with a 4 km spatial resolution from https://oceandata.sci.gsfc.nasa.gov/MODIS-Aqua/ Mapped/Monthly/4km. We adopted linear interpolation from the nearest surrounding points to fill the missing data in the monthly MODIS POC and chlorophyll and re-gridded to the spatial resolution of 1° × 1°. We calculated anomalies by subtracting the climatological monthly mean (2003-2017) from monthly data. The Niño3.4 index, defined as the SST anomaly for the Niño3.4 region (i.e., 5°N-5°S, 120°−170°W), was obtained to represent the ENSO event 88 (https://www.cpc.ncep.noaa.gov/data/indices). We also downloaded the PDO (http://research. jisao.washington.edu/pdo/) and NPGO (http://www.o3d.org/npgo/npgo.php) indices. The PDO and NPGO indices are defined as the first and the second corresponding principal components of SST anomalies and sea-surface height anomalies over the northeast Pacific (180°W-110°W; 25°N-62°N) 62 . Correlation analyses were carried out to examine the relationships between the POC anomalies and the anomalies of other variables or climate indices.