National Aeronautics and Space Administration

December 2015

Santha Akella, Max Suarez and Ricardo Todling

Summary

Satellite and in-situ observations of the Sea-Surface Temperature (SST) show high variability, including a diurnal cycle and very thin, cool skin layer in contact with the atmosphere. The GEOS-5 Atmospheric Data Assimilation (ADAS) has been modified to incorporate these effects and also to assimilate Advanced Very High Resolution Radiometer (AVHRR) brightness temperatures to obtain realistic estimates of the skin SST. Results with the updated system for one-month experiment indicate an overall improvement in the performance of the ADAS.

Motivation

Skin SST is very important for air-sea interaction and in the versions of the GEOS-5 ADAS used for real-time analyses and reanalyses it is specified from an analyzed bulk SST as part of prescribed lower boundary condition. However, radiometric and in-situ observations1 show skin SST variation of up to 2- 3K in the Tropics. This temperature difference is largely due to diurnal warming and the presence of a cool skin layer. Incorporating a realistic skin SST is also essential for atmosphere-ocean coupled data assimilation as part of an Integrated Earth System Analysis (IESA).

Methodology

The GEOS-5 Atmospheric General Circulation Model (AGCM) and also the observing and analysis systems have been updated with the following additions. AGCM now includes a prognostic model for the evolution of skin SST. The diurnal warming model is based on Takaya et al. (2010)² and cool skin layer formulation follows Fairall et al. (1996)1. Monthly mean skin minus bulk SST at 18 UTC is shown in Figures 1a and 1b for GEOS-5 and ECMWF operational Integrated Forecast System (IFS), respectively. Negative values correspond to contribution from cool skin layer and positive values are from diurnal warming.

Infrared radiometers such as the AVHRR are sensitive to the skin SST; these observations have been added to the observing system used in GEOS-5. They are analyzed using GEOS-5 background fields through the Gridpoint Statistical Interpolation (GSI) atmospheric analysis using the Community Radiative Transfer Model (CRTM); in this implementation, skin SST is analyzed simultaneously with the upper-air fields. Figure 2 shows coverage and observation minus background (OMB) residuals for channel 4 of AVHRR (after quality control and bias correction); with two polar orbiting satellites, NOAA-18 and MetOp-A, there is near global coverage within a six-hour assimilation window. The global mean and standard deviation of OMB are within 0.1K and 0.4K, respectively.

The analysis increment in skin SST is now used by the GEOS-5 AGCM through an incremental analysis update to forecast for the following analysis cycle. Figure 3 shows the monthly mean increment in skin SST at 00, 06, 12, and 18 UTC analyses. Positive (negative) values indicate that the analyzed temperature is warmer (cooler) than the model forecast, this information is used to further tune and improve the biases in the ADAS. For example, where the increment is positive (negative), the surface heat and momentum flux parameterizations in the AGCM, and background error correlations in the GSI analysis, could be retuned to obtain a colder (warmer) skin SST, that would better fit the observations.

Figure 1a. Monthly mean difference between skin SST and bulk SST for Apr 2012 at 18 UTC. Land and sea-ice surface types have been masked out; bulk SST is from the OSTIA analysis. Diurnal cycle of skin minus bulk SST is shown in This animated GIF.

Figure 1b. Same as Fig 1a but using ECMWF IFS, sea-ice regions are not masked [courtesy of Anton Beljaars].

Figure 2. Coverage and departure of background brightness temperature from observations for AVHRR Channel 4, analysis is centered at 12 UTC on Apr 1, 2012. NOAA-18 is plotted with circles, and Metop-A with triangles. Shaded area represents local nighttime.

Figure 3. Monthly mean of skin SST analysis increment for 0, 6, 12 and 18z analyses.

Validation

Overall, the modified GEOS-5 ADAS system produces fields that better fit the observations in the low levels of atmosphere and also in the near-surface ocean; a promising result for the atmosphere-ocean coupled data assimilation through IESA.

The mean and standard deviation of the departure of background and analyses from Advanced Microwave Sounding Unit (AMSU)-A observations (on the AQUA satellite) is shown in Figure 4. Smaller OMB in the modified system as compared to the present system indicates overall improvement due to the skin SST assimilation. We obtained similar results with other microwave observations (not shown), which suggests a positive synergetic contribution from the inclusion of AVHRR observations in the infrared part of the electromagnetic spectrum, that are able to positively influence the assimilation of microwave observations. This also shows that the modified GEOS-5 ADAS is in better agreement with surface peaking observations (IR & MW), and future addition of SST relevant MW observations (for example, on board GPM-GMI) could be complementary to the assimilation of IR observations.

Figure 4. Validation of skin SST via satellite radiance measurements: Statistics of OBS Minus Background (OMB, solid line) and Analysis (OMA, dashed) for AMSU-A on AQUA, only active channels are plotted for April 2012. Panels (a) - (d) show the mean, standard deviation (SDEV), mean bias correction and number of observations, respectively. The modified GEOS-5 ADAS experiment (Assim) has a lower mean OMB, (Ch.6) almost same SDEV and number of observations during quality control (CTL).

Near-surface SST computed using the diurnal warming model is compared to the in-situ (withheld) drifting buoy observations, shown in Figure 5. It is evident that the modified system is closer to the observations than the bulk (Operational Sea Surface Temperature and Sea Ice Analysis- OSTIA) SST. This result also shows that the model produced diurnal warming in the Indian Ocean is about 0.5K cooler than observations and decays faster than the observed diurnal warming. Other ocean basins lead to similar conclusion (not shown) and future work will be directed to decrease this bias.

The developments in this work amount to a statistically significant positive impact on five-day forecast skill scores in the southern hemisphere.

Figure 5. Validation of diurnal warming via in-situ SST observations: Mean of OMB for April 2012 in the Indian Ocean for drifting buoys (DBUOY). Top panel shows the monthly mean OMB and bottom panel plots the hourly mean temperature variation for the same region. The background (BKG) is closer to the observed (OBS) than the OSTIA SST, which by design tries to exclude observations that may have any diurnal warming. DBUOY SST observations are taken at a nominal depth of 20 cm, and were not used in analysis.

Current & Future Work

In terms of updates to the observing system, microwave observations from Global Precipitation Measurement (GPM) Microwave Imager (GPM-GMI) and Advanced Microwave Scanning Radiometer-2 (AMSR-2) onboard Global Change Observation Mission (GCOM-W) provide measurements of sub-skin SST (few millimeters deeper than the air-sea interface). Analysis of these and in-situ observations is currently being developed for use in GSI.

The GEOS-5 ADAS was recently updated to a hybrid analysis system. Accounting for the relevant horizontal correlation length scales for skin SST and vertical correlation with the near-surface fields will be pursued in the context of dynamically derived background-error covariances.

Concurrently, we are working on coupling the ADAS with a surface wave model and the GEOS-5 integrated Ocean Data Assimilation System (iODAS), proceeding toward enhancing the connections between atmosphere and ocean in the GMAO’s Earth System analyses.

References:

1.   1.Fairall CW, et al., 1996. Cool-skin and warm-layer effects on sea surface temperature. J Geophys Res-Oceans 101: 1295–1308, doi:10.1029/95JC03190.

2.   Takaya Y, et al., 2010. Refinements to a prognostic scheme of skin sea surface temperature. J Geophys Res-Oceans 115: C06 009, doi:10.1029/2009JC005985.

« GMAO Research Briefs