4.2 Data and Methods
4.3.3 Depth-Integrated Primary Production over Time
The primary production estimates derived from the glider and integrated to the euphotic depth show a similar pattern to those of the integrated chlorophyll (Figure 4.1, 4.5), with a strong seasonal cycle. Rates of production were relatively low in autumn and winter, decreasing from ∼1 g C m−2 d−1 in September 2012 to less than 0.5 g C m−2 d−1 throughout February and March 2013. Rates then increased in April. A strong peak in primary production developed in May, with production rates reaching 2.5 gC m−2 d−1. Towards the end of May production decreased to less than 1 g C m−2d−1. A later, more dominant peak (the spring bloom) then developed in June reaching up to 3 g C m−2 d−1 at the beginning of July. The rates of production decreased slightly to between 1 and 2 g C m−2 d−1 throughout August and into September when the gliders were recovered. A linear regression between mixed layer depth and depth integrated primary production to the euphotic depth gave a low correlation (R2 between 0.05 and 0.2 depending on the glider, Table 4.1). However for all gliders the p-value was < 0.001 and therefore there was a significant relationship between shallower MLD and increased primary production, although it is important to note that this may not imply causality. However, for a long timeseries and where there is a lot of variability, where each data point is independent, the p-value can still be significant. The glider with the highest R2 was SG579 (0.2), deployed from January to April, as chlorophyll concentrations increased and the mixed layer began to shoal.
Table 4.1: Regressions between glider primary production with mixed layer depth
Glider R2 P-value Regression Equation
SG566 0.15 < 0.0001 PP = -0.002MLD + 0.4 SG533 0.17 < 0.0001 PP = -0.002MLD + 0.5 SG579 0.2 < 0.0001 PP = -0.001MLD + 0.4 SG566 0.05 < 0.0001 PP = -0.004MLD + 1.7 SG510 0.15 < 0.0001 PP = -0.003MLD + 0.9 SG533 0.03 < 0.0001 PP = -0.009MLD + 2.1
The satellite timeseries of primary production is also shown in Figure 4.5 for comparison with the glider data. Cruise data were fairly limited and satellite data provide a full year for comparison. I will use it to assess when and how the two methods (glider and satellite) may differ. The grey area in the figure shows the range of satellite estimates
Sep12 Nov12 Jan13 Mar13 May13 Jul13 Sep13 Primary Produc ti on (gC m -2 d -1 ) 0 0.5 1 1.5 2 2.5 3 3.5 4 Range of MODIS in 100 x 100 km MODIS Aqua Study Site
Glider PP
Mean Daily Glider PP
Figure 4.5: Time series of glider derived primary production integrated over the euphotic depth at the study site (the 15 x 15 km area where the gliders flew Figure 1.3). The range of production values around the study site in the 100 x 100 km area are shown in grey calculated from the satellite data, also integrated to the euphotic
depth. The solid black line is the satellite estimate from pixels within the survey
region. The glider production values are shown as dark grey dots. The mean daily primary production from the glider is a solid red line.
in a 100 x 100 km box around the deployment site and the solid black line indicates the average of the pixels extracted over the 15 x 15 km survey region. There is very little difference in the satellite estimates of primary production over the larger area compared with the smaller survey site. The general trends in the glider and satellite data sets are similar, with low rates of productivity in winter, increasing throughout spring and peaking in July. However, the satellite timeseries does show several differences to the glider timeseries. The satellite estimates of primary production are generally higher than those from the glider, apart from in the summer, when the subsurface maximum is present. The satellite estimates also missed the spike in primary production in May. A closer inspection suggests that this may be due to the statistical relationship used to estimate chlorophyll at depth from the satellite data (Morel & Berthon, 1989). At the beginning of the deployment the satellite method showed an increase in chlorophyll at depth that the glider did not observe, whereas toward the end of the timeseries when the glider observed a subsurface production maximum the satellite underestimated the amount of chlorophyll present at that depth. The root mean square error, which provides a useful measure of the differences between measured and predicted values, was calculated between each glider and satellite chlorophyll profile to determine which
times of year the profile was best modelled (Figure 4.6). The RMS of the glider profiles over each day is also plotted on the same figure, this allows a visual to determine if the RMSE is large. For example if the RMSE is much lower than the RMS the error is of less significance whereas if the values are similar the error is very large. The error was low in January, March and towards the end of May, however there were significant differences between the measured and observed profiles for most of the year, in particular when chlorophyll concentrations are high. A comparison between satellite chlorophyll and glider chlorophyll can be seen in Appendix B.
Figure 4.6: The root mean squared error between the glider chlorophyll profiles and the profiles modelled from the satellite chlorophyll using Morel and Berthon (orange), against the root mean square of the glider profiles for each day (black), the RMS gives a reference for determining if the RMSE is large compared with the concentrations
observed in the timeseries
Chlorophyll normalised production is shown in Figure 4.7. The correlation between primary production and chlorophyll is significant (p-value < 0.001). However there is a large amount of scatter in the data and the relationship changes over the year.
From the glider estimates of primary production an annual rate of production can be estimated for this region. By integrating all the values of primary production from the glider over time and the euphotic depth, primary production in this area was 255 g C m−2 year−1. A yearly estimate from the satellite data over the deployment site was similar to the glider 225 ± 7.5 g C m−2 year−1, the error estimate comes from the standard deviation of the yearly integrated estimates for each of the 16 pixels within the PAP site.
Figure 4.7: The top panel (a) shows the ratio of integrated chlorophyll to integrated primary production. The bottom panel (b) shows one against the other coloured by date, demonstrating how the relationship changes throughout the year. The black line is the relationship between integrated chlorophyll and carbon fixation for bottle data
found by Painter et al. (2010a) at the PAP site in 2006. R2 = 0.25 p-value < 0.001
RMSE 0.6. The blue line is the line of best fit (y = 0.03x -0.37)