IASNFS CTD comparison


Intra-Americas Sea Ocean Nowcast/Forecast System

Evaluation against CTD

Dong S. Ko/NRL

      The temperature and salinity profiles measured by the CTD in the Gulf of Mexico (GOM) are used for the evaluation. The CTD data are provided by Tom Leming of the NOAA Marine Fisheries Service. The measurement covered a 2-month period from 2003/06/14 to 2003/08/15. There are 44 CTD profiles available. For the evaluation, 44 temperature and salinity profiles were sampled from the IASNFS 12-hourly nowcasts at the location of CTD stations (nearest model grid) and at the measurement time (linear interpolation between nowcasts).
      During the 2-month period of the CTD measurement the GOM Loop Current (LC) underwent a rigorous evolution as shown by IASNFS prediction (see IASNFS ssh movie). At middle of June 2003, the beginning of measurement period, a Loop Current Eddy (LCE) was separated from the LC (plot). It re-attached to the LC a few weeks later (plot). Over the next month, at the end of July, the LCE detached again from the LC (plot). Toward the end of measurement period at middle of August the LCE re-attached back to the LC (plot) but a smaller eddy was shedded a few weeks later at the end of August (plot).





      The circulation in the region is complex. The water masses, however, are simple and they are clearly distinguishable as shown on the T-S diagram of the CTD profiles. There are basically two types of water mass in the GOM. The LC and LCE water, shown in red, is warmer and saltier. On the other hand, the GOM ambient water, shown in blue, is cooler and fresher. The profiles located on the shelf shown in green are the freshest.

Fig. 1 T-S Diagram from 44 CTD profiles (The graph at the upper-right shows the location of the CTD stations)
      Examining the IASNFS profiles, we found that their water types in general agree with those from CTD ( Fig. 2). A few IASNFS profiles ( 011, 018, 020, 023, 043) are of different water type than the CTD measurement. All those profiles are located near the edge of the LC or the LCE. This suggests that the general location of the LC and the LCE is well predicted by the IASNFS but their frontal location may not be predicted precisely at all times.

Fig. 2 Comparison of types of the water mass between CTD and IASNFS
      The evolution of the LC and LCE shedding is complex and the mechanism involved highly nonlinear. They can only be predicted with data assimilated. The altimeter data which are the main source of data IASNFS assimilates are of relative coarse resolution, particularly, in the off-track direction. The altimeter data are also sparse in time (30-day/17-day/13-day repeat) relative to the evolution of the LC/LCE. The precision of frontal location of LC and LCE in the IASNFS prediction therefore is limited. Lost of any altimeter satellite will further degrade the prediction. Nevertheless, the fair agreement in the water types between IASNFS prediction and the CTD measurement indicates that the data assimilation applied in the IASNFS performs well to predict the general location of the LC and LCE.
      The IASNFS predicted temperature profiles in the LC and LCE compare well with the CTD measurement. The predicted salinity, however, does not compare well with the observations. The LC/LCE water is originated in part from the North Equatorial Current (NEC) of North Atlantic Subtropic Gyre and in part from the South Equatorial Current (SEC) of South Atlantic Subtropic Gyre. The water of NEC/SEC is warm and salty. The upper layer water from SEC, however, becomes fresh due to the influx of large precipitation as it flows through the Equatorial region and due to the infusion of a large fresh water flux from Amazon, Orinoco and other rivers along the northern South America coast. The NEC and SEC water are combined in the Caribbean Sea and form the LC in the GOM. The LC near surface water is fresher and it has a maximum salinity at about 120 m depth (007 and 008). Once the LCE separates from LC it undergoes geostrophic adjustment and its isopycnals are depressed. The maximum salinity is downward to 200 m in LCE (024, 025 and 026). This geostrophic adjustment process is somehow tainted in the IASNFS due to dominant effect of assimilating the altimetry/ MCSST derived MODAS temperature/salinity analyses. In this case the MODAS temperature analysis seems to be reasonable but not the salinity. The inconsistence in the MODAS salinity analysis to the temperature, both should be deepen at the same rate, results in a large difference between the IASNFS salinity profiles and the CTD. The inconsistency can also be clearly seen from the T-S diagram of IASNFS profiles. The T-S relation from IASNFS is much loose at greater depths than the CTD data indicated.

      Also the IASNFS predicted salinity in the LCE is lower then the CTD measurement near the surface (024, 025 and 026). It is likely due to the inaccurate surface salinity flux applied by the IASNFS. The IASNFS uses the MODAS salinity analysis to estimate the surface salinity flux. The MODAS surface salinity shows no trace of the LCE or LC at the time. The analysis basically shows the seasonal (summer) climatology with fresh surface water covering entire northern part of the GOM. As a result, an unrealistic large surface fresh water flux was applied over the LCE and the IASNFS near surface salinity in the LCE became lower then the CTD measurement.

      The GOM ambient water is originated from LC/LCE but has undergone modifications from surface fluxes, river runoff and mixing/diffusion processes. In the measurement area, mainly in the northern half of the GOM, the upper layer temperature/salinity of the ambient water is in general lower then the LC/LCE water (TS diagram). (The water in the SW GOM is saltier and warmer due to larger evaporation and heating). The water on the shelf, particularly for those near the coastal river runoff, is much fresher near the surface (001, 014, 016, 031, 033, 044).

      The IASNFS temperatures in the GOM ambient water, including those on the shelf, compare well to the CTD (for example, 029, 035, 036, 037, 039 and 041). The IASNFS salinities, however, are in general fresher then the CTD data near the surface. It may be the result of applying an unrealistic surface salinity flux from MODAS as discussed. But it may in part due to the seasonal climatological river runoff used in IASNFS. If the Year 2003 is a dry season then the average river runoff will be unrealistically large at the summer and the near surface salinity during the time becomes too fresh.

      There are a few profiles in the GOM ambient water from IASNFS that are not consistent with the CTD data. For example, profiles 028, 032, 034 and 042 show that the IASNFS predictions have a lower temperature then the CTD data below 50-100 m. This is due to that the profiles are located in a cyclonic (cold) eddy predicted by the IASNFS. This eddy, however, may not exist or locate at the right place. The IASNFS salinity in these profiles seems to be consistent with temperature. Both temperature and salinity were upwelled due to the cyclonic motion. Unlike the temperature below 50-100 m, the near surface temperatures compare very well with CTD in those profiles. It may be due to the application of a vertical weighting function that is very weak at the upper layer in assimilating the MODAS profiles. The predicted near surface temperature, therefore, is more of a reflection of the surface heat flux and the MCSST assimilation. The weighting function works well in this case. On the other hand, the profiles 019 and 030 show that the IASNFS sometimes does not predict the surface mixed layer deep enough compared to the CTD data. It is very likely caused by the unrealistic fresh surface water which stabilize the water column and prevent mixing under the summertime light wind condition. Another group of profiles, 036, 037, 038, 039 and 040, indicate an anticyclonic warm eddy near the Texas coast. The IASNFS temperatures match well with the CTD data except at the profile 040. It suggests that this anticyclonic eddy was predicted by the IASNFS but its frontal location might not be predicted precisely.

      Forty four CTD temperature and salinity profiles measured at the GOM from 2003/06/14 to 2003/08/15 are used to evaluate the IASNFS. Comparing the water types of the IASNFS profiles to the CTD data we found that IASNFS predicted well the general location of LC/LCE during a period when the GOM LC/LCE underwent a rigorous evolution. The precise location of LC/LCE front and some smaller cyclonic/anticyclonic eddies, however, are sometimes not predicted correctly by IASNFS. (The same conclusion can also be drawn from the comparison of IASNFS ssh to SeaWIFS and from the comparison of IASNFS temperature to AXBT). Within the same water type, the IASNFS temperature profiles compare very well with the CTD data including those located on the shelf. The IASNFS salinity profiles, however, do not compare well with the CTD data, particularly in the LCE. Over all, the IASNFS predicted salinity is too fresh near the surface.

      From the evaluation, it suggests that the altimeter data which IASNFS assimilated are sufficient to resolve the larger scale features such as the LC/LCE but may not have enough spatial/temporal resolution to precisely resolve the LC/LCE frontal location and the smaller cyclonic/anticyclonic eddies which evolve rapidly. The lost of some altimeter data, caused by GFO turnoff for example, will further degrade the prediction.

      The MODAS temperature analysis is fairly good in the region. The surface forcings from the NOGAPS and the MCSST are accurate and the model near suface temperature responds reasonably well to the forcings. The MODAS salinity analysis, however, is not consistent with temperature in the LC/LCE. At the surface, MODAS salinity analysis fails. The surface salinity flux applied to IASNFS is badly estimated from the MODAS analysis. The estimation of surface salinity flux may be improved by using the evaporation/precipitation if the reliable data are available. Also the MODAS analyses may be improved if a vertical correlation is applied. The seasonal river runoff is not sufficient for the synoptic prediction. The real-time river runoff should be used.

Comparison of T/S profiles between CTD and IASNFS

001 | 002 | 003 | 004 | 005 | 006 | 007 | 008 | 009 | 010
011 | 012 | 013 | 014 | 015 | 016 | 017 | 018 | 019 | 020
021 | 022 | 023 | 024 | 025 | 026 | 027 | 028 | 029 | 030
031 | 032 | 033 | 034 | 035 | 036 | 037 | 038 | 039 | 040
041 | 042 | 043 | 044

Comparison of Temperature between AXBT and IASNFS

Comparison of the temperature at 200 m (300 m) between NAVO AXBTs and IASNFS prediction sampled at AXBT locations. It shows that the IASNFS well predicted the Loop Current frontal location based on the location of 16 (14) deg C isotherm. The graph at the upper-right shows the IASNFS sea surface height prediction and the AXBT locations.

Main| Movie| Archive