Comparison of Argo profiles observations against numerical model simulations in Ionian Sea (2014)

D.Kassis*,(1) , A. Konstantinidou(1), L.Perivoliotis(1) and G.Korres(1)

(1) :  HCMR Greece

Corresponding author:

Autonomous profiling floats array has been providing important information regarding ocean dynamics during the last decade. Lately, the expansion of the so-called Argo Network into Regional Seas has upgraded monitoring and forecasting activities in enclosed sea basins giving the opportunity of more enhanced studies of the mesoscale and sub-mesoscale dynamics dominating in such areas. For the Ionian and Adriatic Seas a substantial increase in the number of Argo profiling floats has been recorded for the period 2008-2012 due to the combined activities of Euro Argo RI and several multinational initiatives. This gathered information is used in this study in order to validate the outputs of the Southern Adriatic North Ionian (SANI) hydrodynamic model, developed for IONIO INTERREG-III project being integrated for the same time period. This preliminary analysis of these 5 years Argo data profiles presented interesting features and trends for the region while the comparison against model outputs revealed the cases of different representation regarding physical properties variability in the area.

Data analysis

SANI model covers the geographical area 15⁰E - 21.76⁰E & 35⁰N - 42.5⁰N with a spatial resolution of 1/50° x 1/50° in horizontal and 25 sigma-layers along the vertical with a logarithmic distribution near the surface and the bottom. It has been integrated for a 5 years period (2008-2012) forced with the POSEIDON/Eta analysis and nested within the MyOcean MFS model. To benefit from the data assimilation that takes place in latter model, SANI was periodically (weekly) initialized from the MyOcean MFS model using variational initialization techniques to dump gravity waves and achieve a smooth solution. The period of the comparison of SANI model temperatures and salinities against Argo temperature and salinity profiles is January 2008 to December 2012. Data from 21 Argo profiling floats in the Ionian Sea are used to compare with model results at or near the same position for this period. Some of the older floats were programmed to execute 10-day cycles, drift at 1000 m depth and descend to 2000 m in order to acquire the ascending profile according to International Argo specifications, whereas a number of newer floats are following the recommendations of MedArgo (Poulain et al., 2007). In that case, drifting depth is 350 m which is near the depth of the Levantine Intermediate Water (LIW) core. The cycle length was reduced to 5 days in order to obtain useful estimates of currents, as longer cycles are not able to represent the circulation in the vicinity of the intricate coastlines of regional Mediterranean Seas. Moreover, the assimilation of profiler displacements becomes inefficient to correct modelled velocities if the cycle length is longer than the typical Lagrangian integral time scale characteristic of the circulation at 350m (Molcard et al., 2003). The profiling depth for this configuration was set to 1000 m and, depending on the float telecommunication technology a time span from 30 min to 6 h is estimated to be the surface time for the localization and data transmission. In total 946 Argo profiles were analyzed in the area of study (Fig. 1). An apparent increase of the Argo profiles over time (42 profiles during 2008, 69 during 2009, 162 during 2010, 217 during 2011 and 476 during 2012) reflects the increased deployments activity in the area during this period.

Figure 1. Argo floats trajectories for the period 2008-2012 in the Ionian Sea.

Results and discussions

The inter-comparison of the 5-year average profiles shows that the model underestimates the temperature in the surface layer (0.1-0.75℃) and below 125m (0.1℃), which is accentuated at deep layers 900-1000m (0.22℃). In the layer between 20 and 125m the model temperature appears greater than Argo’s with a maximum of 0.4℃ at 80m. Regarding salinity there is an overestimation at subsurface (20-75m) and deeper (>600m) layers between 0.01-0.03 psu while at intermediate depths (100-500m) the LIW is presented saltier in Argo data reaching a maximum average of 38.89 psu which is approximately 0.02 psu saltier than the model’s estimation (Fig.2). The Argo-to-model comparison in a yearly basis from 2008-2012 comes to enhance the above observations (Fig.3 & 4). Model temperature and salinity yearly average profiles follow the Argo datasets trends with similar variation in the pre-mentioned depth layers for the entire period of 2008-2012 (Fig.3 & 4). The Root Mean Square Errors (RMSE) of temperature and salinity between model and Argo data with depth (shown in Fig. 5) ranges, in the case of temperature, from 0.15-1.13℃ with its maximum at 50m depth, while in the case  of salinity  from 0.02 to 0.18 psu, with  maximum values at 20m. In order to exclude surface layers high variability, In Fig. 6 the TS diagram of intermediate waters (100-500m) and deeper waters (500-1000m) is presented corresponding to the two datasets. Between 100 and 500m both datasets present similar distributions, with the model characterized by slightly denser water masses. On the contrary, at the deeper layers the model shows much denser water masses (below the 29.2 kg/m3 isopycnal), mainly due to the temperature underestimation at these layers, as described previously, which is then translated into large differences in the density field.

Figure 2. Average temperature (upper left panel) and salinity (lower left panel) profiles for Argo data (blue line) and model (red line), along with their differences (right panel) for 2008-2012.

Figure 3. Yearly average temperature profiles for 2008-2012.

Figure 4. Yearly average salinity profiles for 2008-2012.

Figure 5. RMSEs of temperature and salinity between Argo data and model (left panels), along with their differences per profile for 5 different depth levels during 2008-2012.

Figure 6. σθ average profiles of Argo data (blue line) and model (red line) (upper left), their differences (upper right), the RMSE of σθ (Argo vs model) (lower left) and their T-S diagram for 2008-2012 (lower right).

Figure 7. T-S diagram of Argo data and model for two different depth layers: 100-500m (pink and cyan dots for model and Argo respectively) and 500-1000m (red and deep blue dots for model and Argo respectively).


SANI model seems to represent adequately the general variability of the Ionian Sea water column physical properties at an inter-annual scale. Nevertheless, the comparison with Argo temperature and salinity profiles for the period 2008–2012 revealed some interesting differences. The intermediate waters (200-600m) as reproduced by the model appear to be less saline and colder in comparison with the Argo data although the signal of LIW appears at the same depth in both datasets. Deeper than 600m, the model shows higher salinity values (~0.02 psu) and lower temperatures (~0.2℃) with respect to observations. This latter difference in temperature seems to determine the model’s misrepresentation of the deep Ionian waters compared to in-situ measurements. At the surface layers the model – Argo disagreement on both temperature and salinity may be due to a variety of attributes such as Argos weakness whilst measuring at these layers, and the high variability of coastal processes. Nevertheless the warmer MAW presented by the model at subsurface layers should be attributed to its systematic errors. These preliminary results indicate the need for more detailed comparative studies in the near future that will include careful filtering of the datasets, subdivision of the study area into smaller regions with similar physical characteristics, annual and seasonal analysis of each sub-region.

  • Molcard A., Piterbarg, L. I., Griffa, A., Ozgokmen, T. M., and Mariano, A. J.: Assimilation of drifter observations for the reconstruction of the Eulerian circulation field, J. Geophys. Res., 108(C3), 3056, doi:10.1029/2001JC001240, 2003.
  • Poulain, P.-M., Barbanti, R., Font, J., Cruzado, A., Millot, C., Gertman, I., Griffa, A., Molcard, A., Rupolo, V., Le Bras, S., and Petit de la Villeon, L.: MedArgo: a drifting profiler program in the Mediterranean Sea, Ocean Sci., 3, 379-395, doi:10.5194/os-3-379-2007, 2007.