01.03.11: Comparison of anomaly in nitrate and σ between ARGO and WHOTS data

Here, I want to check my statement that the distribution of events in nitrate and potential density σ as measured by the ARGO data are different than that at the WHOTS data. Figs. 1 and 2 show the anomaly in nitrate and σ, respectively, measured by the ARGO float. Fig. 3 shows the anomaly in σ in the two WHOTS deployment. All horizontal axis start from a January first and have the same length.

Figure 1: Anomaly in nitrate measured by ARGO float Hawaii5145. The anomaly is defined as the difference from the time-mean.

Figure 2: Anomaly in σ measured by ARGO float Hawaii5145. The anomaly is defined as the difference from the monthly-mean σ (APDRC-ARGO product).

Figure 3: Anomaly in σ measured during WHOTS. The anomaly is defined as periods shorter than half a year –in this note, it was shown that this type of anomaly should be equivalent to the anomaly defined from the monthly mean as done in Fig. 2.

As already observed, the nitrate and σ anomalies are correlated below the mixed layer. For completeness, Figs. 4 and 5 show the anomalies in the ARGO data down to 1000 m.

Figure 4: As in Fig. 1.

Figure 5: As in Fig. 2.

In Fig. 6, I plot the vertical averaged anomalies between 125 and 155 m and between 150 and 200 m (only for the float data) to ease the comparison. The nitrate and σ anomalies in the ARGO data are correlated over these depth ranges. However, the σ anomalies are larger and have a lower frequency than in the ARGO data. Both differences may be due to 1) the different sampling scheme, 2) the bias of ARGO float data toward lower SSH (ARGO floats samples more in-between and around eddies than within eddies; see below) and 3) the different region sampled (the WHOTS data sample a domain just north of Hawaii that has on average a larger eddy kinetic energy than the domain to the east sampled by the ARGO float). My guess is: the difference in amplitude is mostly due to the different regime sampled and the difference in frequency is due to the different sampling scheme.

Figure 6: Vertical averaged anomalies in nitrate and σ in the ARGO data of Hawaii5145 and in WHOTS data.

To check that the main reason for the difference in amplitude is due to the fact that a different regime is being sampled, I show the anomalies in σ from the two other ARGO floats Hawaii6401 and Hawaii6891 that have sampled the region around WHOTS so far (Fig. 7). It is, unfortunately, inconclusive as only one of float data, that of Hawaii6891, shows effectively larger anomalies. A more definitive answer may be obtained if I include the data from the many more ARGO floats (for instance, compute the standard deviation of the anomalies in density at the nutricline depth over the region; does this reflect the SSH?).

Figure 7: Vertical averaged anomalies in σ in the ARGO data of Hawaii6401 and Hawaii6891.

In the meantime, in the 1/10th OFES simulation, the standard deviation of the σ anomalies at the depth of the nutricline are larger north of Hawaii than to the east, which might explain the lower amplitude seen in most of the ARGO time series of the float Hawaii5145 (Fig. 8).

Figure 8: Standard deviation of σ anomalies between 120 and 150 m in OFES over years 2002 to 2008.

On the possible bias in ARGO float sampling

According to my study last summer on shallow/deep events of isopycnal layer (see Pres_072410_shallow_deep_events_ARGO.pdf in /RESEARCH/PAPERS/Pres_072410_shallow_deep_events_ARGO), I found that ARGO data seem to undersample cyclonic eddies and oversample the region between eddies, particularly the region with convergent flow (positive Okubo-Weiss parameter). Based on Davis (1982), Middleton and Garrett (1986) argue that drifters will tend to oversample cyclonic over anticyclonic eddies, because cyclonic eddies are region of convergence (why?). Provenzale (1999; see also Provenzale et al. 1998) studies the behavior of heavy particles (like ARGO floats) inside an eddy regime in a non-rotating and rotating system. In a rotating system, heavy particles are ejected out of coherent vorticies due to the centrifugal force. In a rotating system, the Coriolis force tends to push heavy particles out of cyclonic eddies and into anti-cyclonic eddies. Thus, when the Rossby number is small, the Coriolis force prevails on the centrifugal force and there is a bias toward sampling anticylonic eddies.

Finally, I studied the change in the spectrum between a time series at a fixed point at station ALOHA, a time series at a fixed point to the east and a time series that is similar to the ARGO time series. Indeed, the spectra for the two fixed time series are similar, suggesting that there is as much energy in the east than in the west but the ARGO-like time series has more energy at high frequency suggesting that short events in the ARGO time series might be due to sampling (Fig. 9).

Figure 9: Averaged spectra for WHOTS-like time series, time series from a fixed position in the East and ARGO-like time series. The average is obtained using spectra obtained from 1000-day long time series shifted in time by 9 days. See sampling_sig_as_ARGO_float.m in `RESEARCH/PROJECTS/MARINE_BIOLOGY/SUBMESOSCALE_PROCESSES/OFES/sampling_sig_as_ARGO_float on ipu1.