Increasing cloudiness in Arctic damps the increase in phytoplankton primary production due to sea ice receding

Introduction Conclusions References

Increasing cloudiness in Arctic damps the increase in phytoplankton primary production due to sea ice receding S. B élanger1 , M. Babin2 , and J.-E.Tremblay 2

Introduction
The impacts of environmental changes on Arctic and Sub-Arctic marine ecosystems are already detectable from field- (Grebmeier et al., 2006;Li et al., 2009) and satellitebased measurements (Arrigo and van Dijken, 2011;Arrigo et al., 2008;Kahru et al., 2011).While the overall increase in primary productivity (PP) has been attributed to a longer growing season, due to enhanced light availability for photosynthesis (Arrigo and van Dijken, 2011;Arrigo et al., 2008), changes in environmental forcing of nutrient supply to the surface have been proposed as the main driver of PP in seasonally icefree waters (Tremblay and Gagnon, 2009).
At high northern latitudes, photosynthetically active radiation (PAR) is known to be an important limitation for marine photosynthesis.Environmental factors affecting the amount of PAR include cloud cover and the presence of sea ice and associated snow cover, which strongly attenuate shortwave radiation.While the ice and snow cover have decreased significantly in recent decades (Comiso et al., 2008), cloud cover has increased (Eastman and Warren, 2010;Wang and Key, 2005).In fact, the shortwave radiation reaching the sea surface during summer months dropped at a mean annual rate of 0.66 W m −2 yr −1 between 1982 and 1999 due to increasing cloudiness (Wang and Key, 2005).Climate models predict both a reduction in sea ice and an increase in cloud cover for the 21st century as the Arctic warms (Vavrus et al., 2010).During the open water season clouds typically cover 90 % of the sky, attenuating significantly the incoming shortwave radiation.Using satellite lidar measurements of cloud properties Palm et al. (2010) found negative correlations between cloud fraction and sea ice coverage, suggesting that increasing temperature and moisture fluxes during icefree period favors cloud formation.They also found greater low cloud frequency and cloud optical thickness above ice-free water areas (Palm et al., 2010).Although previous studies suggested that the Arctic region will become more productive overall due to a decline in the duration and extent of sea ice (Arrigo and van Dijken, 2011;Arrigo et al., 2008), the net effect of the opposing trends in the evolution of sea ice and cloud Figures cover on PAR, and consequently on the marine primary productivity has never been assessed.
The environmental forces that control vertical mixing in the upper ocean (e.g.wind, freshwater input) and, consequently, the supply of nutrients to the sea surface, may counteract (increase stratification) (Li et al., 2009) or amplify (increased frequency of upwelling events that brings nutrient-rich waters to the surface) (Carmack and Chapman, 2003;Tremblay et al., 2011) the positive influence of increasing PAR.From the perspective of space-borne observation, a change in ocean stratification affecting nutrient supply should be detectable from the ensuing change in chlorophyll a concentration (CHL), a proxy for phytoplankton biomass routinely derived from remote sensing of ocean color (Behrenfeld et al., 2006).
Photosynthesis decreases when absorbing materials from terrestrial origin increase the blue light attenuation (Smyth et al., 2005).Shortly after the spring freshet, the Arctic shelves receive massive amount terrigenous colored dissolved organic matter (CDOM) and suspended particulate material (SPM).Recent studies have reported increasing coastal erosion along several arctic coastline (Rachold et al., 2000) and river runoff at northern latitudes (Peterson et al., 2006).Models also predicted an important release of DOM into the Arctic Ocean (+700 %) from the carbon-rich siberian peatlands (Frey and Smith, 2005).Diffuse light attenuation coefficient (K d ) can, therefore, change due to variations in allochtonous material input.In constrast, removal processes of optically active constituents (e.g.CDOM photobleaching) can increase light penetration and favour PP.To detect such changes from space, PP models need to consider K d independently from CHL (e.g.Smyth et al., 2005).
The objectives of this study were (1) to assess the trends in PAR reaching the sea surface versus PAR penetrating the sea surface after considering sea ice cover, and (2) to quantify the relative contribution of changing PAR conditions due to sea ice, clouds and ocean optical properties, respectively, to the observed trends in PP.To achieve these objectives we developed a PAR model that assimilates satellite-based cloud properties and a fully spectral PP model for optically complex Arctic waters.The Figures

Back Close
Full implementation of the model was initiated as part of the Malina project in order to develop long-term monitoring capabilities of the Arctic's marine ecosystems productivity.We show that, over the 1998-2010 period of ocean color observation, PAR increased only slightly (but not significantly) in Arctic waters as a result of opposing trends in cloud and sea-ice cover.Nevertheless, positive PP trends were found.They are explained by concurrent changes in PAR, CHL and K d , and the contribution of each factor varied strongly spatially.

Primary production model
Daily PP rates were calculated using a photosynthesis-irradiance model (i.e.P vs E curves): chlorophyll a concentration (CHL; in mg m −3 ), photosynthetically usable radiation (PUR, in Einstein m −2 s −1 ), light-saturated CHL-normalized carbon fixation rate (P B m ; in mg C (mg CHL) −1 h −1 ), and saturation irradiance (E k , Einstein m −2 s −1 ) are needed for the calculation of PP at each depth.Ocean color data, binned at a 9.28 km resolution on a equal-area grid, were used for CHL and to calculate PUR at each depth in the water column.Briefly, monthly CHL data retrieved using a semi-analytical algorithm (GSM01) (Maritorenna et al., 2002) were obtained from the Ocean Color MEa-SUREs project (v6) at UCSB http://wiki.icess.ucsb.edu/measures/Products). GSM01 was found to perform better than standard empirical algorithms (e.g.NASA's OC4v6) in Arctic waters dominated by CDOM absorption (Ben Mustapha et al., 2012).Next, Sea-WiFS Level 3 monthly water-leaving reflectance (R rs ) at 412, 443, 490, 510, 555 and 670 nm were obtained from the NASA GSFC (reprocessing 2010.0).Spectral IOPs, namely the total absorption (a) and backscattering (b b ) coefficients, were estimated from R rs (λ) using a quasi-analytical algorithm (QAA) (Lee et al., 2002).The accuracy of the QAA in Arctic waters remains to be assess, but a preliminary validation indicated an excellent performance of this algorithm for the retrievals of the total a and b b (absolute relative difference < 18 %) (B élanger, 2006) attenuation coefficient K d (λ) averaged over the euphotic zone was estimated following the approach of Lee et al. (2005) using the QAA-derived IOPs as input.K d (λ) was used to propagate E d (z, λ, t) throughout the water column.This was achieved at twelve optical depths from the sea surface to a level of 0.1 % of the incident light.PUR(z, t) was calculated at each time step and depth using : where a ph (λ) is the spectral phytoplankton absorption coefficient (in m −1 ), and E 0 is the spectral scalar irradiance.The later was calculated by dividing E d by the mean cosine of downwelling irradiance (Morel, 1991), approximated by the expression (a+bb) K d (Sathyendranath et al., 1989).a ph (λ) was calculated using an empirical statistical relationship established between a ph (λ) and CHL derived from measurements made in the Western Arctic Ocean by Matsuoka et al. (2011).E k was modeled using the Arrigo et al. (1998) model developed for high latitudes, while P B m was assumed to be constant at 2.0 mg C (mg CHL) −1 h −1 , an averaged value based on field measurements in Arctic waters (Harrison and Platt, 1986).
PP calculations using the above model were made for each day of the year using the monthly means of the IOPs and CHL.Gaps in the monthly fields, due to presistent cloud or sea-ice cover, were filled with monthly climatology of IOPs and CHL.When OC data were available, the daily PP was computed for each day of the month.The daily PP rate of the pixel was adjusted as a function of the daily fraction of open water (1 − SIC).PP was assumed nil where no IOPs and CHL data are available (i.e.pixels never documented by SeaWiFS).This method allowed us to consider the exact same surface area from year to year for the whole time series, and thereby minimizing the possible bias introduced by the increasing number of ocean pixels documented by SeaWiFS through time as open-water area increases.Introduction

Conclusions References
Tables Figures

Back Close
Full

Trends analysis
The trends in yearly PAR and PP over the 13 yr SeaWiFS time series were calculated for each pixel using a nonlinear trends estimator as described in Zhang et al. (2000).This is a non-parametric method that removes autocorrelation and outliers from the time series before calculating the trend using the Theil-Sen approach (TSA; Sen slope).
The zyp.zhang function implemented in R was used.The Mann-Kendall non-parametric test was then run on the resulting time series to test the significance of the trends.

Conclusions References
Tables Figures

Back Close
Full

Below-surface photosynthetically available radiation trends
In the seasonally and permanently open water, sea ice is the primarily factor controlling the penetration of PAR in the ocean (Fig. 2a) (Perovich et al., 2007).As expected, PAR(0−) has increased significantly since 1998 at rates reaching up to +150 Einstein m −2 yr −1 ) at the margin of the Central ice pack (Fig. 2b) resulting in a relative increase of 0.34 % yr −1 in PAR for the circum-Arctic waters (red bars on Fig. 3a).Interestingly, the increase in the high Arctic (≥ 70 • N) was largely counterbalanced by a general decrease in PAR(0−) between 55 to 65 • N (Fig. 2b).Integrating total PAR(0−) over Arctic and sub-Arctic Seas (including Okhotsk, Bering, Labrador, White Seas and Hudson complex; see Fig. S1) resulted in slightly negative trend of −0.30 % yr −1 (Fig. 3b).

Total circum-Arctic primary production
The PAR model was used to drive the primary production model as detailed in Sect.2.2.Total circum-Arctic estimate of primary production was more than two fold smaller (203 ± 15 TgC yr −1 ) than previous satellite-based estimates (i.e.441 to 585 TgC yr −1 ) (Arrigo et al., 2008;Arrigo and van Dijken, 2011).This departure arises partly from the choices of ocean color algorithms and photosynthetic parameters used in the model.Our PP model explicitly accounts for the fact that Arctic waters are optically complex (Matsuoka et al., 2007;Siegel et al., 2005;B élanger et al., 2008).Firstly, total light absorption and scattering coefficients of seawater constituents were assessed from ocean color reflectance measurements to estimate the diffuse attenuation (Lee et al., 2002(Lee et al., , 2005)).Secondly, CHL was retrieved using a semi-analytical approach that minimize the effect of colored detrital mater (CDM), which is dominant in the Arctic (B élanger et al., 2008;Matsuoka et al., 2007Matsuoka et al., , 2011;;Wang et al., 2005).The most important factor explaining the lower PP estimation is the lower CHL values obtained employs the spectral diffuse attenuation coefficient (K d ), estimated using a QAA (Lee et al., 2002(Lee et al., , 2005)), rather than the relationships between CHL and IOPs, and between IOPs and K d , as usually employed in PP models developed for clear case 1 waters.Interestingly, when PP is calculated using OC4v4 CHL estimation and the Kd parameterization employed by previous studies, the total PP rises by ∼ 85 % to 357±15 TgC yr −1 .
This value is still lower than previous estimation of Arrigo et al. (2008) probably because our model does not provide PP sectors that have never been documented by SeaWiFS.Finally, the light-saturated CHL-normalized carbon fixation rate, P B m , which usually varies as a function of sea surface temperature in PP models, is fixed to a constant in our model (2.0 mg C (mg CHL) −1 h −1 ) due to a lack of robust parameterization for arctic waters.

Primary production trends
Figure 4b shows that changes are spatially heterogeneous and relatively small over most of the circum Arctic Ocean (≤ 3 gC m −2 yr −1 ).The overall positive trend in PP is mostly driven by the historically productive regions of inflow (Barents and Chukchi) and interior (e.g.Kara, Laptev, East Siberian and Beaufort Seas) shelves (Fig. 4b).Because small relative changes in these productive areas can dwarf large relative changes in unproductive areas, the standardized trends presented in Fig. 4c provide a better assessment of intra-regional changes than do absolute trends.After standardization, positive trends are found on seasonally ice-free inflow and interior shelves, as well as in permanently open waters (Southern Iceland Shelf, Western Bering Sea).Negative trends are found along major exit routes of Arctic water over outflow shelves in the Eastern Greenland Sea and Canadian Archipelago (e.g.Northwestern Baffin Bay, Lancaster Sound), and in the northern part of the Bering Sea.

Trends in ocean optical properties
The above results show that changes in PAR(0−) alone cannot explain the general increase in PP (Fig. 2 versus Fig. 4).We therefore examined whether changes in ocean optical properties can be detected and if they explain PP trends.We first calculated the trends in monthly chlorophyll a concentration since it is arguably the most important parameter driving the PP estimation.We found modest positive CHL trends in May and June when the increased in PP was highest (< +0.5 % yr −1 ), and even negative trends during the productive months of July and August (< −0.3 % yr −1 ) (black versus green bars in Fig. 3a).The increase in CHL, however, was more important when considering both Arctic and sub-Arctic Seas, with standardized trends reaching 1 % yr −1 in June (Fig. 3b).
Our model uses satellite-derived spectral diffuse attenuation (K d (λ)) to propagate PUR through the water column.The diffuse attenuation of PUR, K PUR , which integrates the spectral effect of light avalailability for photosynthesis, is thus an appropriate quantity to examine the impact of spectral diffuse attenutation trends on PP.Because PP in Eq. ( 1) is proportional to CHL and inveresly proportional to K PUR (i.e. 1 K PUR ), the ratio CHL/K PUR is strongly correlated to PP.In essence, this ratio expresses the the phytoplankton biomass, CHL, relative to the total diffuse attenuation background.The later varies when CDOM, SPM or phytoplankton pigments characteristics (e.g.pigment packaging) and concentration vary.In the context of a PP trends analysis, trends in ocean optical properties is best represented by CHL/K PUR .
The seasonal variation in CHL/K PUR is shown on Fig. 5a-d blooms, such as Baffin Bay and the Barents and Chukchi Seas (Fig. 5a).In contrast, CHL/K PUR remained low during the whole summer season in strongly stratified arctic waters (e.g. the Canada Basin).In August, most arctic waters are strongly stratified except in regions where deep-water upwellings are dominant (e.g.NOW polynya, Northeast Greenland, Hudson Strait, Siberian coastal waters) (Fig. 5d).
The monthly trends depicted in Fig. 5e-l, showed a general increase in CHL/K PUR in May, particularly in the Southern Labrador Sea.In June, the positive trends are found in the Northern Labrador, Barents and Bering Seas (Fig. 5f, j).In July, positive trends are found in Hudson Bay, arround Southern Greenland and in the Southern Barents and Chukchi Seas (Fig. 5g, k), whereas relatively strong negative trends occurred in coastal waters of Northern Greenland.When CHL/K PUR increases, the phytoplankton biomass increases relative to light attenuation, thus resulting in higher PP.Therefore, as for PP, we observed an overall increase in CHL/K PUR for all months in both the circum Arctic and the combined Arctic and sub-Arctic Seas (Figs. 3 and 5e-l).

Discussion and conclusions
The decrease in PAR(0+) over open water was nearly ubiquitous between 55 • N and • N, with decadal reduction of 8 % for the integrated Arctic and sub-Arctic Seas and up to 20 % in specific areas.This trend is consistent with several studies reporting strong positive anomalies in cloud amount above newly opened waters during summer and early fall (Eastman and Warren, 2010;Vavrus et al., 2010;Palm et al., 2010;Wang and Key, 2005).Increased cloudiness thus partly counteracts the positive influence of declining sea ice on PAR(0−) and renders the change in PAR(0−) non significant over the 13-yr study period (+3.4 % per decade in the circum Arctic and −3 % over both Arctic and sub-Arctic Seas; Fig. 3).Warmer temperature and moisture fluxes are likely the main drivers of the increase in cloudiness (Eastman and Warren, 2010;Vavrus et al., 2010;Palm et al., 2010;Wang and Key, 2005).The positive trends over the perrennally ice-covered ocean is, however, more uncertain due to the lower quality of cloud Figures

Back Close
Full properties retrievals over ice covered conditions (Schweiger et al., 1999;Chernokulsky and Mokhov, 2012).Previous analyses of satellite-based PP time series revealed significant variability in spatial patterns and temporal trends across Arctic marine ecosystems (Arrigo and van Dijken, 2011;Arrigo et al., 2008;Kahru et al., 2011;Perrette et al., 2011).Here we also found a significant (p < 0.05) PP trend of +2.8 TgC yr −1 (or +1.4 % yr −1 in relative terms; Fig. 3a) between 1998 and 2010 for the circum-Arctic.Longer growing season and light availability (i.e.PAR(0−)) was previously identified as a main driver of change in the high Arctic (Arrigo and van Dijken, 2011;Arrigo et al., 2008).We also found a correlation between annual PAR and PP anomalies across the circum-Arctic (r 2 = 0.42, p < 0.001; not shown).However, changes in light availability do not alone explain the positive PP trends in many Arctic sectors, except on Arctic interior shelves where PAR(0−) is largely driven by SIC (Figs. 2 versus 4).In many regions, such as Hudson Bay, Baffin Strait, Baffin Bay and the Labrador, Norwegian and Barents Seas, PAR(0−) decreased while PP increased.The pattern in Fig. 5 shows that this observation can only be explained by a change in ocean optical properties, here expressed as the ratio of chlorophyll a concentration to the diffuse attenuation coefficient of photosynthetically usable radiation (CHL/K PUR ) (Fig. 5).This ratio seems to provide a good proxy for trophic state of Arctic and Sub-Arctic waters.
In general, low CHL/K PUR values occurred in stratified, nutrients-depleted surface waters, while high values were found in productive waters sustained by nutrients inputs from rivers or the deep ocean (Fig. 5a-d).In several Arctic sectors, for example, the haline stratification limits nutrient supply, explaining the low surface CHL.In these waters, however, the diffuse attenuation remained relatively important due to the high background in colored detrital materials (CDM).The high CDM background in the Arctic polar mixed layer (PML) is due to the relatively high concentration of CDOM from both terrestrial and marine origin (Stedmon et al., 2011).In the Barents Sea in summer, low CHL/K PUR ratio resulting from concurent low surface CHL and high CDM background (Fig. 5b-d of the spring bloom that occured in May (Fig. 5a).Indeed, CDOM production from microbial activity is often decoupled from phytoplankton production and is delay after the spring bloom (Nelson et al., 1998).In sectors where river inputs are dominant, however, very high values of CHL/K PUR may be overestimated due to the difficulty to distinguish CHL from other optically active constituents.Improvements of ocean color algorithms for both CHL and K d remains a major issue in the coastal arctic (Ben Mustapha et al., 2012).CHL/K PUR remained relatively constant within Arctic interior shelves, possibly because sea ice and clouds limit the number of good quality data in these environments.The general rise in CHL/K PUR observed in permanently open waters during May could be related to increasingly early stabilization of the water column in spring, resulting in earlier phytoplankton blooms (Kahru et al., 2011).The positive PP trends in the Labrador Sea (Fig. 4), for example, are due to increasing values in CHL/K PUR from south to north in early summer (Fig. 5i, j).Similarly, high PP in the Barents Sea is mostly driven by the strong positive trend in CHL/K PUR in June and July (Fig. 5j, k).Despite a general positive trend in PP over much of the Arctic, we found strong indications that PP is declining in regions of prime ecological importance.Negative PP trends were mostly explained by a reduction of PAR due to increasing cloudiness (Fig. 1) in the Northern Bering Sea (Fig. 4) and by higher SIC and lower penetration of sun light in the water column (Fig. 2) near the location of the northeast water (NEW) polynya.The latter result may be due to an increase in sea-ice export trough Fram Strait since 2003 as reported recently by Kwok et al. (2009).The North Water (NOW) polynya, located in the Northern Baffin Bay, experienced one of the most severe drop in PP (5 gC m −2 yr −1 ; Fig. 4b).With an area of 80 000 km 2 , the NOW has been considered to be the most productive recurrent polynya north of 77 • N due to a long and intense diatom bloom that starts in May and fuels a rich marine ecosystem supporting polar cod, large aggregations of marine mammals, sea birds and polar bears (Tremblay et al., 2006).Our results indicate that the timing of this bloom may have changed over time.Indeed, a modest increase in CHL/K PUR in May was followed by a sustained decrease Introduction

Conclusions References
Tables Figures

Back Close
Full in June and July (Fig. 5), which was primarily responsible for the decline in the annual productivity of the NOW.These results raise questions about the ultimate drivers of changes in the NOW.We speculate that bloom dynamics are linked to changes in the quantity or properties (e.g.salinity, nutrients, CDOM) of the in-flow of cold, nutrientdepleted waters coming from the Arctic Ocean (Kwok et al., 2010;M ünchow et al., 2011).
Another limitations are inherent to our satellite-based PP trends assessment.First, our method cannot detect if changes occured in the subsurface chlorophyll a maximum (SCM), which can be important in the Arctic waters during the summer period (Martin et al., 2012).Recent studies, however, suggested that the error in the annual PP assessment resulting from the omission of the SCM is small, but more important during the post-bloom period in late summer (Ardyna et al., 2012;Arrigo and van Dijken, 2011).Second, under-ice phytoplankton blooms, which are undetectable from space, may be more important than previously thought.These blooms have been observed in different Arctic sectors and can locally represent most of the annual PP (Arrigo et al., 2012;Mundy et al., 2009).
High spatial resolution maps of PP trends, together with those in PAR(0+), PAR(0−) and CHL/K PUR , provide new insights about the main drivers of changes in primary productivity across the Arctic and Sub-Arctic marine ecosystems.Future developpements of long-term monitoring capabilities of marine arctic ecosystems should, among other, (1) include improvements in the ocean color algorithms to reduce the uncertainty on the satellite-based PP estimation in optically complex waters, (2) address the problem of data continuity to produce consistent time series of ocean color and other sensors, (3) evaluate and compare methods used to estimate PAR reaching the sea surface under cloud and ice conditions and (4) examine in more details the environmental variability (e.g.SST, wind speed, storms frequency, etc.) to better understand the most important drivers of changes in the ocean optical properties and PP.Introduction

Conclusions References
Tables Figures

Back Close
Full  Full Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | and were obtained from the International Satellite Cloud Climatology Project (ISCCP) web site.The ISCCP global radiative flux data (FD) are distributed on a 280 km equal-area grid at 3 h intervals for dates between January 1984 and December 2009.Here, PAR(0+) refers to the integral of E d (λ) from 400 to 700 nm just above the sea and ice surface (in Einstein m −2 h −1 ), while PAR(0−) is the is the integrated irradiance just below the sea surface after considering air-to-sea interface reflection for both direct and diffuse components of the downwelling irradiance (ρ Fresnel ) and the daily sea ice concentration (SIC; in %) (i.e.PAR(0−) = PAR(0+) • (1 − ρ Fresnel ) • (1 − SIC)).Daily satellite-derived SIC data from the Defense Meteorological Satellite Program (DMSP) Scanning Multichannel Microwave Radiometer (SMMR), F8 and F13 Special Sensor Microwave Imager Discussion Paper | Discussion Paper | Discussion Paper | . The in-water spectral diffuse 13992 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | From 1998 to 2009, which corresponds approximately to the SeaWiFS era, PAR(0+) generally decreased at a rate raging from −100 to −50 Einstein m −2 yr −1 over seasonally and permanently open water (Fig. 1b), while it increased (∼ +50 to +100 Einstein m −2 yr −1 ) in over the permanently icecovered Central Arctic waters.The largest decrease in PAR(0+) were found between 55 • N and 70 • N (Hudson Bay, Gulf of Alaska, Bering Sea and Nordic Seas).The relative changes generally lied between ±2 % yr −1 (Fig. 1c).Monthly maps of PAR(0+) trends indicate that changes occurred mostly in June, May and July, respectively, depending on regions (Fig. S1, Supplement).If integrated over the whole circum Arctic waters (≥ 66.58 • N), PAR(0+) has decreased by ∼ 10 % over the SeaWiFS era (yellow bars on Fig.
Discussion Paper | Discussion Paper | Discussion Paper | using semi-analytical model as compared to CHL empirically derived from NASA standard algorithms such as OC4v4 (Ben Mustapha et al., 2012).In addition, our PP model Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | , where a general decrease is observed from May to August.Most Arctic waters exhibited very low CHL/K PUR (< 2 mg CHL m −2 ) relative to the North Atlantic and North Pacific.In May relatively high values were found in sectors known to host instense spring phytoplankton Introduction Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | ) could be due to the in situ production of CDOM during the consumption Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 3 .
Fig. 3. (A) Monthly trends in PAR(0+), PAR(0−), CHL, CHL/K PUR and PP for the circum Arctic waters (above the Arctic circle at 66.58 • N).Yearly trends are only shown for PAR(0+), PAR(0−), and PP.(B) same as (A), but for all regions considered as Arctic and Sub-Arctic seas, which include region below the Arctic circles infested by sea ice or melt glacier for at least on part of the year (e.g.Hudson Bay, Labrador Sea, Gulf of Alaska) (See Fig. S1 for the exact limits of the seas defined by the International Hydrographic Organization).

Fig. 5 .
Fig. 5. (A-D) Monthly climatology of the ratio between CHL and the diffuse attenuation of PUR (CHL/K PUR ); (E-H) absolute CHL/K PUR trends calculated using the TSA, and (I-L) the standardized CHL/K PUR trends (i.e.CHL/K PUR trends/climatological CHL/K PUR • 100).