Academia.eduAcademia.edu

Figure 4 – uploaded by Mati Kahru

See full PDFdownloadDownload figure
Fig. 5. Linear slope of the e-ratio versus log10 of NPP (see Fig. 3) estimated for preceding periods of variable length (corresponding to the lag) and for the in situ data (large open circle). Fig. 4. e-ratio (EF/NPP) versus log10 of NPP. (A), using in situ EF and in situ NPP; (B) using in situ EF and satellite estimated average 6-day NPP prior to | measurement. The dotted lines are the OLS linear regressions: (A) e-ratio = 0.93-0.223 x log10NPP and (B) e-ratio = 0.98-0.220 x log10ONPP.

Figure 5 Linear slope of the e-ratio versus log10 of NPP (see Fig. 3) estimated for preceding periods of variable length (corresponding to the lag) and for the in situ data (large open circle). Fig. 4. e-ratio (EF/NPP) versus log10 of NPP. (A), using in situ EF and in situ NPP; (B) using in situ EF and satellite estimated average 6-day NPP prior to | measurement. The dotted lines are the OLS linear regressions: (A) e-ratio = 0.93-0.223 x log10NPP and (B) e-ratio = 0.98-0.220 x log10ONPP.

Related Figures (6)

When combining all in situ data (Fig. 2, N = 35), including the “frontal” cycles, the relationship between EF and NPP becomes noisy and it is difficult to approximate the scatter with a good relationship. We compared satellite match-ups with three simple models: The Kelly et al., 2018 model (EF = 0.08 x NPP + 72mg Cm 7d‘), the OLS model (Fig. 2, EF = 0.07 x NPP +147 mg Cm~?.d~') and the Type II model (Fig. 2, EF = 0.28 x NPP +10 mg C m? d~). The higher intercept in the OLS model, relative to the Kelly et al., 2018 model is
When combining all in situ data (Fig. 2, N = 35), including the “frontal” cycles, the relationship between EF and NPP becomes noisy and it is difficult to approximate the scatter with a good relationship. We compared satellite match-ups with three simple models: The Kelly et al., 2018 model (EF = 0.08 x NPP + 72mg Cm 7d‘), the OLS model (Fig. 2, EF = 0.07 x NPP +147 mg Cm~?.d~') and the Type II model (Fig. 2, EF = 0.28 x NPP +10 mg C m? d~). The higher intercept in the OLS model, relative to the Kelly et al., 2018 model is
Fig. 1. Same day match-ups of satellite-derived water column productivity versus in situ measured productivity in Log10 scales. N = 15, R2 = 0.78. The crossed circles are from the “front” cruises of 2009, 2011, 2012, 2017. The dashed line is the OLS regression and the solid line is the Type II regression.
Fig. 1. Same day match-ups of satellite-derived water column productivity versus in situ measured productivity in Log10 scales. N = 15, R2 = 0.78. The crossed circles are from the “front” cruises of 2009, 2011, 2012, 2017. The dashed line is the OLS regression and the solid line is the Type II regression.
Comparison of different statistics of satellite EF retrievals using the Kelly et al., (2018) model, OLS and York Type II regressions in Fig. 1B. Bold values show improved estimates. The log10 range of in situ data was 1.00.
Comparison of different statistics of satellite EF retrievals using the Kelly et al., (2018) model, OLS and York Type II regressions in Fig. 1B. Bold values show improved estimates. The log10 range of in situ data was 1.00.
Fig. 3. Satellite estimates of EF with same-day match-ups using (A) Kelly et al. (2018); (B) OLS regression from Fig. 2; and (C) the York Type II model in Fig. 2. The dotted line is the one to one line.
Fig. 3. Satellite estimates of EF with same-day match-ups using (A) Kelly et al. (2018); (B) OLS regression from Fig. 2; and (C) the York Type II model in Fig. 2. The dotted line is the one to one line.
Fig. 6. Coefficient of determination (R?) between in situ measurements and satellite estimates averaged during a preceding period of a number of days as a function the respective preceding period of averaging for NPP (gray filled cir- cles, the dashed line has been smoothed with a 3-point smoothing boxcar filter) and EF (open circles connected with a solid line).
Fig. 6. Coefficient of determination (R?) between in situ measurements and satellite estimates averaged during a preceding period of a number of days as a function the respective preceding period of averaging for NPP (gray filled cir- cles, the dashed line has been smoothed with a 3-point smoothing boxcar filter) and EF (open circles connected with a solid line).
To evaluate the differences resulting from using different EF models on satellite-derived time series we chose an area around Station M (34°50’N, 123°0’ W) in a transition region between coastal upwelling areas of the CCE and the oligotrophic offshore domain. Observations at Station M have been carried out since 1989 (Smith et al., 2013, 2018). We used a time series of NPP calculated at 5-day intervals using 4-km satellite data (http://www.wimsoft.com/CC4km.htm) and applied both the Kelly et al., 2018 model and the Type II model. The valid EF pixel values were averaged in a circle with 100-km radius. The para- meterization of the EF model has a dramatic effect on estimated EF time series (Fig. 7), particularly on estimates of EF variability. For example, standard deviation (SD) of the time series of EF for the period of 2007-2018 more than doubled from 24.2mgC m7? day (0.078 in logi9 units) when using the Kelly et al., 2018 model to 65.8 (0.185 in logio units) mg C m~? day” ' when using the Type II model. As shown by Smith et al. (2018), episodic fluxes play an increasing role in the total flux reaching the bottom. It is therefore important to investigate the processes driving variability in EF for the in situ data (see discus- sion), as well as the differences between two algorithms derived from a Use of a Type II regression only explains a portion of the variability in abyssal carbon flux noted by Smith et al. (2018). This is not sur- prising, because NPP alone only explains a portion of the variability in export flux seen in Fig. 2. Indeed, the highest flux events recorded in this study occurred at intermediate values of NPP. These spikes of EF are likely related to a suite of physical, biogeochemical, and ecological dynamics that affect pelagic ecosystems, only some of which are likely Fig. 7. Time series of satellite estimates of EF within a circle (radius = 100 km) around Station M (see Smith et al., 2018]) determined with the Kelly2018 model (r line) and the York Type II model (black line).
To evaluate the differences resulting from using different EF models on satellite-derived time series we chose an area around Station M (34°50’N, 123°0’ W) in a transition region between coastal upwelling areas of the CCE and the oligotrophic offshore domain. Observations at Station M have been carried out since 1989 (Smith et al., 2013, 2018). We used a time series of NPP calculated at 5-day intervals using 4-km satellite data (http://www.wimsoft.com/CC4km.htm) and applied both the Kelly et al., 2018 model and the Type II model. The valid EF pixel values were averaged in a circle with 100-km radius. The para- meterization of the EF model has a dramatic effect on estimated EF time series (Fig. 7), particularly on estimates of EF variability. For example, standard deviation (SD) of the time series of EF for the period of 2007-2018 more than doubled from 24.2mgC m7? day (0.078 in logi9 units) when using the Kelly et al., 2018 model to 65.8 (0.185 in logio units) mg C m~? day” ' when using the Type II model. As shown by Smith et al. (2018), episodic fluxes play an increasing role in the total flux reaching the bottom. It is therefore important to investigate the processes driving variability in EF for the in situ data (see discus- sion), as well as the differences between two algorithms derived from a Use of a Type II regression only explains a portion of the variability in abyssal carbon flux noted by Smith et al. (2018). This is not sur- prising, because NPP alone only explains a portion of the variability in export flux seen in Fig. 2. Indeed, the highest flux events recorded in this study occurred at intermediate values of NPP. These spikes of EF are likely related to a suite of physical, biogeochemical, and ecological dynamics that affect pelagic ecosystems, only some of which are likely Fig. 7. Time series of satellite estimates of EF within a circle (radius = 100 km) around Station M (see Smith et al., 2018]) determined with the Kelly2018 model (r line) and the York Type II model (black line).
Related topics:
Environmental ScienceGeochemistryOceanographyEcology
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved