4.1. Rates of NCP and GPP in WEP
 In this paper, we present NCP and GPP rates in a broad region of the equatorial Pacific, concentrating on the WEP, for which relatively little biological production data has previously been reported. The continuous nature of the NCP data allows assessment of spatial heterogeneity and the combination of NCP and GPP data yields information on how tightly the system is recycling carbon. Given the weak seasonality in the equatorial Pacific, the rates reported here probably reflect the annual average NCP and GPP though may be an underestimate since there is some suggestion that chlorophyll at least (and thus perhaps by extension production) shows some seasonality with highest values in the spring and lowest values in the fall [Strutton et al., 2008].
 A weak El Niño had developed by the end of the cruise, with a Multivariate Enso Index (MEI) [Wolter and Timlin, 1998] of 0.7–0.9. Turk et al.  found that new production between 165°E and 165°W decreased in El Niño events while new production between 145°E and 165°E increased during El Niño events. Furthermore, Strutton et al.  found that at MEI > ∼0.2, chlorophyll concentrations decreased as MEI increased in the region between 170°W and 180° but at MEI < 1, chlorophyll concentrations did not show a significant decrease with increasing MEI in the region between 125°W and 155°W (corresponding to our CEP). Thus, the NCP and GPP rates we measured in the CEP and WEP may be an underestimate of rates in a typical (i.e., non-El Niño) year, whereas the rates measured in PNG may overestimate rates in a typical year.
 The NCP estimates reported here are lower bounds since they assume no upward mixing of low-oxygen water. This assumption is supported by the fact that there is little vertical exchange in the western equatorial Pacific, as water of low salinity acts as a barrier between the surface and the thermocline [Lindstrom et al., 1987; Lukas and Lindstrom, 1991]. Nonetheless, we can estimate the maximum uncertainty associated with this assumption by calculating the maximum amount of O2 debt that could be mixed upward if this barrier layer were not present. To do this, we combine the concentration gradient of O2 between 100 and 300 m as measured in CTD profiles with a canonical estimate of vertical diffusivity coefficient Kz equal to 2 × 10−5 [Ledwell et al., 1993; 1998]. We find that in the WEP the O2/Ar method could be underestimating NCP by at most 0.5 ± 0.3 mmol O2 m−2 d−1. Future work could combine estimates of upwelling from other geochemical tracers, such as 3He [Jenkins, 1988; Jenkins and Doney, 2003], with O2/Ar measurements in order to correct for the upwelled O2 debt.
 We measured NCP only in the mixed layer, not the euphotic zone. In the western equatorial Pacific, the mixed layer depth is on average 30 m (range of 23–60 m) whereas the euphotic zone is approximately 60–90 m [Ryan et al., 2006]. Thus, if NCP was constant with respect to depth throughout the mixed layer and euphotic zone, our method would be only estimating half the euphotic depth integrated NCP. It is more likely, however, that NCP varies with depth, with NCP decreasing closer to the bottom of the euphotic zone, and thus this method underestimates total euphotic zone NCP by less than a factor of 2. Estimates of net primary production from 14C bottle incubations measured on the same cruise suggest that in the WEP mixed layer primary production accounts for 60% ± 20% of total euphotic zone production (Veronica Lance, personal communication).
 Our geochemical tracer rate, which averages over relatively large temporal and spatial scales, falls in the middle of the range of other estimates of WEP production. Turk et al.  measured new production rates by 15N uptake in bottle incubations and found 0.15–0.19 mmol N m−2 d−1 in regular to moderate El Niño conditions (1994 El Niño). Using the revised Redfield ratios of Anderson and Sarmiento , this new production flux is equivalent to 0.4–0.5 mol C m−2 yr−1. Our estimate of NCP is almost 3 times larger, which may have several causes. First, we sampled during El Niño conditions that were weaker than those prevailing when Turk sampled (MEI = 0.8 for this cruise versus MEI = 1.4 for Turk et al.). Second, new production may be temporally or spatially mismatched to NCP, although the two are expected to be equal at steady state or on long temporal and large spatial scales [Eppley and Peterson, 1979]. Third, rates may vary interannually because of changes in ENSO or to other factors. Finally, there may be biases or uncertainties in the different methodologies of 15N addition in bottle experiments versus dissolved gas tracers.
 A much larger flux of 3.3 mol C m−2 yr−1 was measured in September and October 1994, when the MEI = 1.3, by Rodier and Le Borgne  who estimated the export flux using shallow drifting sediment traps at 100 and 300 m. This flux is twice that of our measured fluxes, despite the fact that the MEI index was higher in 1994. Thus, we would have expected greater depression of production in 1994 than in 2006. The difference in our estimate and that of Rodier and Le Borgne may be due to the known problems associated with shallow sediment traps such as collection of swimmers (leading to overcollection) and hydrodynamic biases (leading to over or under collection) [Buesseler, 1991]. The discrepancy may also be due to the differences in NCP and export production, or to interannual variability.
 An additional relevant comparison to our results is that of Schlitzer , who estimated equatorial Pacific export production from adjoint modeling of dissolved oxygen, nutrient, and carbon data. Thus, his method, like ours, is based in part on the distribution of dissolved gases. However, Schlitzer's method is not tied to any given year as he used a historical data set spanning 50 years. Schlitzer did not estimate export production specifically in the WEP; rather, he estimated export production throughout the equatorial belt to be 1.1 mol C m−2 yr−1, which is in reasonable agreement with our estimate for WEP NCP of 1.5 mol C m−2 yr−1. However, it is surprising that the estimate for the entire equatorial Pacific, including the central and eastern equatorial Pacific, which are more productive, is similar to our estimate for only the WEP. The agreement is also unexpected considering our measurements took place in a moderate El Niño and, therefore, may be underestimating productivity.
 Our NCP estimates agree reasonably well with those of production from two biogeochemical models. Model estimates of WEP export production from a global three-dimensional marine ecosystem model of Moore et al.  are approximately 0.5–2 mol C m−2 yr−1, which compares well with our NCP estimate of 1.5 mol C m−2 yr−1. However, in the Moore et al. model, export production decreases rapidly toward the west, a pattern we do not observe in our data. In a different coupled physical-biogeochemical model [Christian et al., 2002], NCP estimates in the WEP were highest during El Niño and La Niña events, ranging in those events from 3 to 9 mol C m−2 yr−1. In years that were neither El Niño nor La Niña, NCP estimates ranged from 1.8 to 2.5 mol C m−2 y. This rate is slightly larger than this study's observation for a weak El Niño year.
 We cannot directly examine the east-west gradient in NCP given that we could not determine NCP in the CEP and eastern equatorial Pacific in this study. However, we can compare our NCP rates in the WEP to other rates measured farther east. First, the average NCP in the WEP (5.9 mmol O2 m−2 d−1, which is equivalent to 1.5 mol C m−2 yr−1) is approximately half the NCP rate calculated using a similar approach in the eastern equatorial Pacific at longitudes 95°W and 110 °W by Hendricks et al. . Hendricks et al. excluded the latitude band between 2°N and 2°S (which is likely the most productive) in their calculations, as upwelling there confounds the rate estimate. Since our estimates are made farther west (east of 180°) we are able to include data near the equator. Second, the average NCP in the WEP is roughly 65% of the rate of new production measured in the central equatorial Pacific during the JGOFS study [McCarthy et al., 1996], again suggesting that NCP follows the east-west gradient in production that has been observed in chlorophyll and other indices of biological activity in the equatorial Pacific.
 We cannot compare our estimates of GPP in the WEP to other studies as GPP has not previously been measured in that region. However, we can compare our estimates of GPP in the CEP to two previous studies. First, during the JGOFS experiments, Bender et al.  measured GPP from 18O labeled bottle incubation studies at 140°W and 0°N/S. The method of Bender et al. is fundamentally different from the method presented here since their method reflects a shorter time scale, a more local spatial scale, and is likely to be influenced by bottle effects [Peterson, 1980]. Bender et al. found that GPP was equal to 220 ± 6 mmol O2 m−2 d−1 (where uncertainty refers to 1 standard deviation) in the spring during El Niño conditions (MEI = 2.2) and equal to 330 ± 31 mmol O2 m−2 d−1 in the fall when the El Niño had weakened considerably (MEI = 0.6). It makes the most sense to compare our estimate of GPP to the Bender et al. fall estimate given the similarity of El Niño conditions and times of year. We find that GPP is equal to 220 mmol O2 m−2 d−1 at 140°W and 0°N/S, which is equal to the Bender et al. spring estimate but 33% lower than the Bender et al. fall estimate. This difference may reflect the difference between the in situ and in vitro methods or may reflect interannual variability. Second, Hendricks et al. , using triple oxygen isotopes, determined GPP at 95°W and 110°W (farther east than the CEP) between 8°N and 8°S but excluding the area near the equator, during November 2000 (MEI = −0.725) to be equal to 143 mmol O2 m−2 d−1. Our estimate in the CEP is considerably larger than the Hendricks et al. estimate, perhaps because of the difference in El Niño versus La Niña conditions, because Hendricks et al. excluded the region between 2°N and 5°S (probably the most productive region) from their estimate or because the Hendricks et al. estimate was made farther east where there might be smaller iron fluxes and thus smaller GPP.
 The f ratio reported here for the WEP is 11% ± 3%, significantly larger than an estimate of the biogeochemically equivalent pe ratio of 3.4% as calculated using the export production algorithm of Dunne et al.  in combination with an estimate of primary productivity from VGPM. However, Dunne et al. state their algorithm is only valid for pe ratios greater than 4%, and even though the f ratio from the data is larger than this, the pe ratio as estimated from the algorithm is below this limit. If the Dunne et al. algorithm that uses chlorophyll a is applied (rather than the algorithm that uses net primary production) in conjunction with shipboard measurements of chlorophyll a, then the Dunne et al. algorithm predicts negative pe ratios, a solution which is clearly unphysical.
 We can also compare the f ratio calculated in this study to the f ratio elsewhere in the Pacific. In the central and eastern equatorial Pacific, the f ratio is estimated to be between 6% and 30% [Dugdale et al., 1992; McCarthy et al., 1996; Rodier and Le Borgne, 1997; Raimbault et al., 1999; Aufdenkampe and Murray, 2002; Hendricks et al., 2005]. Thus overall, the WEP has lower NCP, GPP, and slightly tighter recycling than its central/eastern counterpart. This may be because the WEP has smaller amounts of major nutrients. Additionally, physical instabilities, such as tropical instability waves, have been found to be associated with increased production in the central and eastern equatorial Pacific [Murray et al., 1994; Dam et al., 1995; Dunne et al., 2000] and tropical instability waves are weaker or nonexistent in the western region [Chelton et al., 2000; Eldin and Rodier, 2003]. Finally, the warmer temperatures in the western region may contribute to tighter recycling as the degradation of organic matter may occur more quickly at warmer temperatures [Laws et al., 2000; Rivkin and Legendre, 2001].
 The WEP has been compared to the subtropical gyres because both regions are similarly low in nutrients. The WEP NCP rates determined in this study are smaller than NCP measurements made using tracer methods in the subtropical gyres [Spitzer and Jenkins, 1989; Gruber et al., 1998; Keeling et al., 2004; Emerson et al., 2008], although some of those estimates included NCP rates in the euphotic zone rather than only in the mixed layer as our study does. The dynamics of the subtropical regions and the WEP are quite different. The subtropical gyres are characterized by downwelling whereas the WEP has weak upwelling capped by a low-salinity barrier [Lindstrom et al., 1987; Lukas and Lindstrom, 1991] that prevents most of the nutrients from reaching the sunlit surface. Additionally, the temperature of the water is warmer in the WEP than in the subtropics.
 Whereas the average NCP is a useful metric for comparison with other regions in the world's oceans, it is also important to consider the variability of the NCP rates. The fact that the variance-to-mean ratio approaches unity throughout much of the WEP indicates that when assessing productivity, one must take into account the heterogeneity inherent in the ocean. Were measurements of NCP only taken at traditional “stations” then several high-productivity events would have been missed. Using this subkilometer scale method for measuring NCP, we find considerable variability in NCP (Figures 1 and 4). This is counter to the observation by Le Bouteiller et al.  that production, as inferred from pigment concentration, was relatively invariant during 5 day station occupations at 180°.
 Notably, the variance-to-mean ratio is smallest between 155°E and 165°E. This is the only region where there are no islands near the equator. Islands may lead to increased variability in NCP either because of direct influence of terrigenous material on the carbon cycle (for example, through the input of nutrients from river runoff or slope sediments) or because the islands disrupt the equatorial undercurrent, a major source of nutrients to the surface equatorial Pacific water [Murray et al., 1995]. ADCP data collected on the cruise show that, in the region between 173°E and 180 °E, the equatorial undercurrent is disrupted whereas it is cohesive in the region between 155°E and 165°E (Pierre Dutrieux, personal communication).
 To further investigate the nature of the variability of NCP, we performed spectral analysis on the NCP data, as well as on continuous records of sea surface salinity, sea surface density, sea surface temperature, zonal wind, and average wind speed. We examined different averaging periods ranging from 1 to 20 days. The conclusions were similar if the averaging window was between 5 and 20 days, and here we report results for averaging periods of 6 days. We detrended the data and also removed the station data from the record to eliminate aliasing by extensive observations from a single location. We performed the analysis as a function of distance along cruise track, rather than as time, so that the physical interpretation would be more obvious. We calculated the spectral density by Welch overlapping segment analysis on 70% overlapping segments of 500 km length. One prolate-spheroidal taper with a time bandwidth product NW = 4 was used per segment. The estimates are robust [Chave et al., 1987; Simons et al., 2000], and 95% confidence intervals were determined from jackknifed error estimates.
 In the WEP, plots of the spectral density (Figures 6a, 6b, and 6c) illustrate that NCP, salinity, and wind speed all have red spectra with no obvious peaks, suggesting there is no single oscillating process driving the variations. The spectral density plots for density and to a lesser extent temperature are similar to that of salinity and therefore are not shown. The time series of all properties are weakly, if at all, nonstationary. In the WEP, NCP, wind speed, and salinity show distinctly different spectra. First, the linear range of the power spectrum extends from roughly 1 to 20 km for NCP and wind speed whereas it extends from roughly 1 to 200 km for salinity. Additionally, the values of the spectral slopes are significantly different. These differences suggest that the mechanisms that are causing variability in NCP are in a large part different than the ones causing variability in density, salinity, temperature, and perhaps wind speed.
Figure 6. Logarithmic spectral density plots, as a function of distance along cruise track (in km) of (a) NCP, (b) sea surface salinity, and (c) 6 day average wind speed in the western equatorial Pacific and of (d) NCP, (e) sea surface salinity, and (f) 6 day average wind speed in the region near Papua New Guinea. The curve is the logarithmic spectral density, as calculated by Welch overlapping segment analysis. The filled circles indicate frequency resolution in the low-frequency portion. The light gray curves are the 95% confidence limits as determined by jackknifed error estimates [Chave et al., 1987; Simons et al., 2000]. The black line is a slope fit to the spectrum corresponding to wavelengths between 1 and 200 km. The equation for the least squares regression line and the spectral slope, with its 2σ error, are reported.
Download figure to PowerPoint
 A rough measure of length scale of correlation can be determined from calculating the e-folding length of the autocorrelation function for NCP, temperature, density, and salinity. For NCP, this correlation length scale is about 50 km. For salinity and density, the correlation length scale is about 130 km. For temperature, the correlation length scale is about 100 km and for wind speed the correlation length scale is 170 km. The correlation length scale of 100–130 km is on the same order as that for deep convective rain systems in the warm tropical Pacific [Masunaga et al., 2005]. The correlation length scale of NCP and the physical properties all fall within the range of length scale of eddies, which is 50–500 km. The correlation length scales are all much shorter than the roughly 1000 km length scales of tropical instability waves or Kelvin waves and orders of magnitude longer than the roughly 30–50 m length scale of Langmuir cells in the WEP. The difference in correlation length scales between NCP and the physical properties suggests that there may be fundamentally different processes controlling the distribution of NCP than are controlling the distribution of salinity, density, or temperature.
 Futhermore, coherence analysis, which assesses whether there are similar spectral properties (and thus perhaps common causes of variability) in different data sets, shows there is no strong coherence between NCP and salinity (Figure 7a), temperature, density, or zonal wind, whereas there was strong coherence (typically 0.6 < γ2 < 0.8 at most wavelengths) between salinity and density (Figure 7b). Of course, salinity is a fundamental control on density so it is not surprising that the two are coherent at all wavelengths. There is strong coherence (γ2 ∼ 0.5) between NCP and average wind speed at wavelengths between 100 and 200 km (Figure 7c). Notably, the strong coherence between NCP and the average wind speed only occurs in a narrow range of wavelengths (100–200 km) whereas the coherence between salinity and density occurs over a range of wavelengths. When different averaging windows for wind speed were chosen (5–20 days), the strong coherence at approximately 100–200 km persists. This suggests that there may be a fundamental link between variability in wind speed and in NCP in the equatorial Pacific, as has been suggested by Strutton et al. .
Figure 7. Square of the coherence, as a function of reciprocal distance along the cruise track (km−1) of (a) NCP and sea surface salinity, (b) sea surface salinity and sea surface density, and (c) NCP and 6 day averaged wind speeds, all in the western equatorial Pacific and (d) NCP and sea surface salinity, (e) sea surface salinity and density, and (f) NCP and 6 day averaged wind speeds, all in the region near Papua New Guinea. Light gray curves are the 95% confidence limits of the square of the coherence as determined by jackknifed error estimates [Chave et al., 1987; Simons et al., 2000].
Download figure to PowerPoint
4.2. Correlations in the WEP
 Although spectral analysis suggests that the causes for much of the variability are different for NCP than for the physical processes, correlations between NCP and sea surface density, salinity, and temperature (Figures 4a–4c) suggest that there may be, in part, a physical forcing to NCP. One can see intervals along the cruise track where a strong positive correlation of density and NCP is followed by a strong negative correlation (circled regions in Figure 4a). This may be a result of the complicated response of the tracer system to vertical mixing. An increase in vertical mixing, e.g., caused by increased winds, would result in an injection of dense, colder, fresher water. Such vertical mixing would also bring up nutrients, stimulating production and thus would be seen as an increase in NCP. However, the injection might also bring up an oxygen debt which would appear as a decrease in NCP given the limitations of this method. The ratio of nutrient concentration to oxygen debt may vary as C/N and C/P ratios have been shown to vary in space and time [Anderson and Pondaven, 2003]. If the water being transported to the mixed layer has a larger nutrient concentration than oxygen debt, then the nutrients could spur biological production and produce O2 in an amount that overcomes the O2 debt, and ΔO2/Ar and thus our estimate of NCP would increase, resulting in a positive correlation. Other possibilities for the positive and negative correlations include differing response of biological production to different types of horizontal fronts, potential time lags between physical changes and biological response, or differing averaging times of the methods to measure the physical properties (near instantaneous) and to measure NCP (averaging of 1–2 weeks).
 The correlations between NCP and zonal wind, wind speed, and gas transfer velocity (Figure 4d) support the suggestion that in the WEP, wind forcing is an important control on NCP [Strutton et al., 2008]. However, only the correlation with gas transfer velocity remains after detrending the data. This may be because the gas transfer velocity is the more relevant parameter since it is the weighted average of the wind speed over the prior several weeks (see section 2.2 and Reuer et al. ) whereas the wind speed is a simple 5 day average of wind speed (results were similar if other time periods for averaging wind speed were used). Another difference is that the gas transfer velocity is proportional to the square of the wind speed, but this alone is not enough to explain the stronger correlation with gas transfer velocity since the correlation of NCP with the square of the wind speed has R2 = 0.15 before detrending and R2 = 0.04 after, both significantly weaker correlations than that of NCP and gas transfer velocity.
 Surprisingly, no correlations were found in the WEP between the NCP or GPP rates and estimates of chlorophyll a from satellites or net primary production (NPP) as estimated from the VGPM algorithm (Figure 4e) in the WEP. The satellite data yields estimates of NPP, our gas measurements yield information on NCP and GPP, and we would expect a correlation between these properties. The lack of correlation between satellite data and the NCP rates is disturbing since satellite estimates are often used to form hypotheses about biological production and carbon cycling. However, it is true that NCP, NPP, and GPP are all assessing different aspects of biological production. GPP is the gross photosynthetic flux. NPP is photosynthesis minus autotrophic respiration, whereas NCP is photosynthesis minus autotrophic and heterotrophic respiration. Thus, the lack of correlation between satellite-derived NPP and gas tracer-derived NCP and GPP may be a reflection of the differences in spatial and temporal scales of the methods or a reflection of the differences in variability of photosynthesis, autotrophic, and heterotrophic respiration. Furthermore, while fluxes of GPP from triple oxygen isotopes and NPP from VGPM were not correlated within the WEP itself, we did find a significant correlation between NPP from VGPM and GPP (R2 = 0.32, p = 0.004) if we used the data from the entire equatorial region covered by this study (i.e., used all the data points on Figure 2). Both the VGPM algorithm and our GPP measurements predict higher biological productivity in the CEP than in the WEP, and it is this broad-scale pattern that leads to the correlation.
 Additionally, no correlations were found in the comparison of NCP and GPP rates in the WEP to the biogeochemical parameters collected at the stations (concentrations of nitrate, phosphate, silicate, and iron at the surface and at depth; photosynthetic efficiency Fv/Fm; chlorophyll concentrations; 14C primary productivity). The lack of correlation with discrete properties measured at stations may be because of the difference in temporal scale of the measurements. For example, nutrient concentrations reflect the state of the ocean at the moment the measurement is taken whereas the NCP and GPP measurements average over the time scale of exchange of the gases, approximately the previous 2 weeks. Alternatively, it could be because the standard set of parameters believed to influence biological production is not truly the driving force behind production in the western equatorial Pacific. Other work has also shown a lack of correlation between surface nutrients and production in the western equatorial Pacific [Rodier and Le Borgne, 1997]. It is unclear why NCP is not correlated with chlorophyll, photosynthetic efficiency, 14C primary productivity, nutrient concentration at depth (75, 100, or 125 m) or depth of the 20° isotherm.
 14C primary productivity is likely a measure of productivity between NPP and GPP [Marra, 2002]. A lack of correlation between the 14C-PP and GPP in the WEP may reflect the difference in time scales of the methods and/or may be a result of variability in autotrophic respiration. We do see a correlation between 14C-PP and GPP if we consider all the data in the equatorial Pacific, suggesting that both 14C-PP and GPP may be registering the same large-scale differences in the regions but are not responding to the small-scale variability in the same way. Researchers sometimes assume GPP is directly related to 14C-PP, often using a factor of 2.7 to convert between the two [Marra, 2002]. However, more recently, Luz et al.  found that in the Sargasso Sea the GPP/14C-PP ratio changed throughout the year, ranging from 3.5 to 7.9. The lack of correlation we observe here suggests there can be spatial as well as temporal differences in the GPP/14C-PP ratio. Indeed, we see such differences on larger scales, as the ratio of GPP/14C-PP is higher in the WEP than it is in the CEP or PNG.
4.3. Rates of NCP and GPP in the PNG Region
 Both the highest and the lowest values of NCP found in the equatorial region (outside of the CEP where the apparent low values are actually due to upwelling) are measured in the PNG region. Rates of GPP at one of the stations near the coast are moderately high, similar to rates in the CEP, whereas rates at the other coastal stations are on the low side, similar to rates in the WEP. Negative values of NCP are found near the mouth of the Sepik River, marked by a red star in Figure 1b. The Sepik River is known to contribute significant loads of organic matter, sediments, and nutrients to the coastal waters [Burns et al., 2008]. The negative NCP values are consistent with net heterotrophy, due possibly to remineralization of organic matter input by the river. Additionally, sediment plumes from the Sepik river may be shading phytoplankton, decreasing productivity. This shading hypothesis is consistent with the low GPP rate calculated at the mouth of the Sepik river as compared to the higher GPP rate calculated slightly south of the river. Although coastal upwelling could cause negative values, such upwelling is unlikely in September since upwelling along this northeastern coast of Papua New Guinea predominantly occurs in December through March during the Northwest monsoon [Fine et al., 1994; Cresswell, 2000]. Moreover, no shoaling of pycnoclines was observed at the time of sampling.
 Similarly, the negative values of NCP (∼−20 mmol O2 m−2 d−1) found in the area just to the northeast of the Vitiaz Strait are probably not attributable to coastal upwelling. Surface currents bring waters from the Vitiaz straits to the northwest [Fine et al., 1994], and thus, these low values in the eastern Bismarck Sea should not be the result of O2 depleted waters being transported from the south. Additionally, these negative values are not near the output of any river. Thus the only remaining explanation for the negative NCP in these waters is local net heterotrophy, but more work should be done to confirm this hypothesis.
 Some of the highest values of NCP found in this study occur in the open ocean to the north of PNG. There are two maxima of NCP, occurring at the same time as wind bursts, which occur between stations and would have been missed by traditional sampling methods. In part, because of the response to these storm events, NCP is correlated with zonal wind in the PNG region. Westerly wind bursts can puncture the barrier layer [Lukas and Lindstrom, 1991], bringing up nutrients to the surface that may be fueling NCP. This is supported by the negative correlation between NCP and salinity we found in this region (Table 1 and Figure 5b), since the waters below the barrier layer will have higher salinity.
 Spectral density plots of NCP (Figure 6d) as well as of physical properties such as sea surface salinity (Figure 6e), sea surface density (not shown), sea surface temperature (not shown), average wind speed (Figure 6f), and zonal winds (not shown) in the PNG region all show red spectra with no obvious peaks. The linear range of the spectral density plots for NCP, salinity, temperature, and density extend from 1 to 200 km whereas for wind speed extends from 1 to 30 km. The similarity in the spectral densities suggest that unlike in the WEP, it is plausible that the spectrum for NCP here is statistically similar to that of salinity, density, and temperature, suggesting a single mechanism may be controlling the distribution of all these properties. However, the general absence of coherence between NCP and salinity (see below) is an indicator that salinity does not directly control NCP. The linear range and slopes are significantly different for wind speed than for NCP.
 NCP, salinity, temperature, and density all have similar second order statistics. The length scales of autocorrelation are similar for all the properties as well as for wind speed. Specifically, the e-folding correlation length scale is 70 km for NCP, 80 km for salinity, 90 km for density, 96 km for temperature, and 69 km for wind speed. This length scale is on the same order as that for the rain correlation length scale for deep convective systems [Masunaga et al., 2005] and as of eddies. As in the WEP, coherence analysis showed that there was no strong relationship between NCP and salinity (Figure 7d), temperature, or density, whereas there was a strong coherence between density and salinity (Figure 7e). There is significant coherence between NCP and wind speed at wavelengths of 100–150 km, suggesting that as in the WEP, there may be a fundamental link between wind speed and NCP.
 There are no significant correlations between NCP and nutrients or other biogeochemical parameters measured at the stations. Again, this may suggest a real lack of relationship between the parameters or may be a result of difference in temporal time scales of the measurements.