Satellite SST

Fig. 18.6 The profiles received in real-time on 13th September 2009 that represent the best profile available from previous retrievals and the source of the profile

are several known limitations to the use of observed SST for ocean forecasting with diurnal warming and skin effects. Specific algorithms are required to perform quality control relevant to the foundation temperature (refer to the online definitions maintained by the Global High Resolution Sea Surface Temperature (GHRSST) science team, Foundation temperature specifically refers to the near surface temperature of the ocean excluding diurnal skin effects. In practice, observations are withheld from the analysis as being impacted by diurnal skin effects based on the time of day and the magnitude of the 10 m winds as a proxy for near surface mixing (Donlon et al. 2002). The algorithms do not attempt to correct the temperature values for any diurnal effects, therefore the day time temperatures will include a small residual bias.

Night-time observations of SST are also affected by a cool skin effect however this is a relatively small perturbation compared with daytime biases. Algorithms use a smaller constraint on atmospheric winds resulting in greater coverage. Therefore the night-time foundation SST's represent a more robust estimate and offer greater coverage compared with day-time products. The majority of ocean forecasting systems at present do not explicitly represent the diurnal skin layer which requires a vertical resolution <1 m. Therefore the temperature in the top model grid cell as well as the statistical covariance over the surface layer are compatible with foundation temperature products. It should however be noted that some ocean models are implementing finer surface resolution in order to represent a portion of the diurnal variability throughout the forecast. Such models require a more sophisticated strategy to constrain the diurnal variability with observations.

Microwave sensors observe SST through clouds but not precipitation and therefore offer improved coverage over infrared away from the inter-tropical convergence zones. Microwave bands reduce the resolution of SST observations to ~25 km/pixel for AMSR-E. This resolution is comparable to (approximately half) the resolution of present forecast system grids. However, the interference from land boundaries reduces performance within two pixels (or ~50 km) of the coastline. AMSR-E does not observe that part of the continental shelf with the highest temperature variability and offers limited coverage for Straits and Gulfs. The orbit of AMSR-E on TERRA is sun-synchronous providing a day-time (ascending) and night-time (descending) equator crossing for swath observations. The percentage of observations from AMSR-E that provide valid foundation temperatures is shown in Fig. 18.7. The night-time observations Fig. 18.7a, c provide greater coverage compared with day-time observations Fig. 18.7b, d as expected for both Austral summer and winter. Note that there is a specific swath line that appears to offer lower coverage but is an artefact of the difference between the period of the satellite orbit >24 h and the period of one earth orbit 24 h. Both descending and ascending swaths show reduced coverage over the inter-tropical convergence zones and monsoons, though the position of these changes with season. In the high latitudes, SST coverage is near 100% up to the ice edge where the atmospheric conditions are of high winds and dry air. Foundation SST from AMSR-E must remove all pixels

Fig. 18.7 Percentage of days observed by AMSR-E for Austral seasons and ascending(asc)/ desending(desc) orbits a summer, desc, b summer, asc, c winter (desc) and d winter (desc)

contaminated by precipitation up to a chosen threshold. In some applications, where maximum coverage is essential a higher threshold can be used. However, for ocean forecasting (foundation temperature) a more conservative approach is important. The so-called Level-2 Pre-processed (L2P) product (refer to L2P-Observations.html) provides all of the necessary fields to diagnose and select the threshold for ocean forecast applications.

The NOAA AVHRR series has been sustained as an operational platform with wide-swath infrared sensors and multiple satellites in sun-synchronous orbits. NAV-OCEANO provide a merged, foundation temperature, swath L2P product available in near real-time. The resolution ~1 km is greater than that of current and near future ocean forecast systems. This permits the construction of super-observations (e.g., Lorenc 1981; Purser et al. 2000) that have reduced representation error increasing the weighting in the analysis. The higher resolution also provides observations over the continental shelf and Gulf regions compared with microwave sensors. An observation error for the foundation temperature can be constructed to account for residual diurnal signals based on the time from nearest local night-time as well as an age penalty for time from the analysis time (Andreu-Burillo et al. 2009).

18.5.3 Satellite Altimetry

Remotely sensed satellite altimetry observes a broad spectrum of dynamical processes including: tides, wind-waves and swell and steric anomalies. Steric anomalies relate to the changes in height from the vertical integral of specific volume anomalies from the background. Vertically coherent specific volume anomalies are prominent in ocean eddies where they can have relatively warm and/or fresh cores relative to the surrounding ocean state leading to positive height anomalies or relatively cool and/or salty cores leading to negative height anomalies. Analyses of merged altimetry have revealed that 50% of the variability of the world ocean is accounted for by eddies with height anomalies of 5-25 cm and diameters 100200 km (Chelton et al. 2007). The speed of propagation for the majority of eddies is found to range from 2.5 to 12.5 cm/s with a westward propagation ±10° (Chelton et al. 2007). In regions where the geostrophic turbulence is more active such as near western boundary currents the eddy propagation speeds can transiently exceed 40 cm/s (Brassington 2010) and can develop height anomalies in excess of 25 cm and diameters in excess of 200 km (see Fig. 18.8).

Recovering sea surface height anomalies from satellite altimetry requires precise estimation of a large number of corrections (Chelton 2001). For example the ssha is recovered from Jason1 by the following equation (Desai et al. 2003).

ssha = (orbit - (range_ku + iono + dry + wet + ssb))

- (mss + setide + otide + pole + invbar) + bias (18.1)

where range_ku refers to the range delay for the Ku-band and iono, dry, ssb refer to range corrections for the ionosphere, dry/wet troposphere and sea state bias.

Fig. 18.8 a An example of on d 98.55 ay of altimetry passes from Envisat, Jasonl and Jason2 for the 1st January 2010, in the Australian region. b ±2 days of altimetry passes about the 1st January 2010 overlaying the corresponding background sea level anomaly in the Tasman Sea from Ocean-MAPS. c same as b but for ±5 days

Fig. 18.8 a An example of on d 98.55 ay of altimetry passes from Envisat, Jasonl and Jason2 for the 1st January 2010, in the Australian region. b ±2 days of altimetry passes about the 1st January 2010 overlaying the corresponding background sea level anomaly in the Tasman Sea from Ocean-MAPS. c same as b but for ±5 days

The terms mss, setide, otide, pole and invbar refer to geophysical effects of mean sea surface, solid Earth tide, ocean and load tide, pole tide, and inverse barometer response. Bias is a correction term resulting from calibration of the orbits. The mean sea surface or geoid is estimated by the time mean of orbit tracks repeated for several years to a precision of 1 km. It is for this reason that the repeat missions Jason1 and Jason2 to TOPEX-Poseidon are put into the same orbits (Robinson 2006). Ocean tidal harmonics are known and can be estmated to high precision with inverse methods (Le Provost 2001). The errors attributed to the TOPEX-Poseidon, Jason class missions is 3 cm, ERS, Envisat and Sentinel missions is 6 cm and GFO is 10 cm (Robinson 2006). The precision that can be achieved by the merger of the Jason class mission and ERS missions is 5 cm (Ducet et al. 2000). Future altimetry missions from the HY-2 series from China, SARAL for the Ka-band altimeter (Altika) from an Indian consortia and Cryosat have as yet unknown errors but are able to obtain improved errors through calibration against the Jason series.

All altimeters launched to date have been nadir-viewing instruments. The spatial and temporal scales that are resolved by these missions are then determined by the spatial and temporal coverage offered by the satellite orbit. It is essential for many of the corrections that a non-sunsynchronous, repeat orbit pattern be used. The repeating polar orbits used are a trade-off between the period between repeat orbits, the equator separation between adjacent passes and the latitudinal range (inclination). The Jason series have a repeat orbit of 9.92 days and a pass separation of

156.6 km (254 passes/cycle) and a latitudinal range of ±66.15°. The ERS/Envisat/ Sentinel series use a retrograde sunsynchronous repeat orbit period of 35 days, a pass separation of 79.9 km (501 passes/cycle) and a latitudinal range of ±81.45° (inclination 98.55°).

The combination of multiple satellite missions is critical to providing improved temporal and spatial coverage to support SLA analyses and operational ocean forecasting (Ducet et al. 2000; Pascual et al. 2009). At present we have Jason1 and Ja-son2 in a tandem orbit and Envisat delivering near real-time products. An example of the passes obtained for a single day (1st January 2010) in the Australian region from these three missions is shown in Fig. 18.8a. The coverage per day is sparse compared with the spatial scales of the error covariances used in ocean forecasting which scales with the order of eddies, 100 km (Oke et al. 2005, 2008; Martin et al. 2007; Brasseur et al. 2005; Cummings 2005). A larger observation window is employed in all operational systems in order to increase the spatial coverage and improve the quality of the least squares analysis. Examples of the coverage or a 5 day window and 11 day window overlayed on a background of SLA from the OceanMAPS system for the Tasman Sea is shown in Fig. 18.8b, c. A 5 day window shows gaps in coverage that are comparable or larger than the spatial scale of the ocean eddies. An 11 day window provides full coverage from Jason1 and Jason2 and partial coverage from Envisat and offers spatial coverage that is comparable to the scales of the ocean eddies (see Fig. 18.8c).

The average altimetry coverage in the Australian region has been estimated for 1° x 1° bins for single and multiple missions available in near real-time (see Fig. 18.9). The along-track observations have been normalised by thinning to a sampling rate of ~1 observation per 50 km which corresponds to a skip of 8 for Jason1 and Jason2 (i.e., 8 x 5.78 km ~ 46 km) and a skip of 6 for Envisat, (i.e., 6 x 7.53 km ~ 45 km). The thinning can be interpreted as the scale that might be used to construct so-called super-observations (e.g., Lorenc 1981; Purser et al. 2000). This is a formal method for compacting observations to reduce the redundancy of the raw observations relative to the target scales which in this case is chosen to be 1° x 1° bins. Super-obs have a number of beneficial properties in practice including: increasing the homogeneity of coverage, reducing the observation space (i.e., computational cost) improve the condition of the matrix inversion in an analysis (see Daley 1991, p. 111).

The average coverage obtained by the multi-satellite missions is a function of the orbit properties described earlier. In practice the coverage is also impacted by periods of communication failures and satellite manoeuvres or equipment failovers. This is evident in the coverage for Envisat (see Fig. 18.9b) which is impacted by the loss of satellite passes during maintenance between the 12th and 27th November 2009 (approximately half a repeat orbit period). The average coverage obtained from Jason1, Jason2 and Envisat (see Fig. 18.9d) over the open ocean ranges between 0.2 and 0.7 observations per 1° x 1° bin per day with the mean coverage ~0.44. The coverage in the coastal regions is reduced in all cases and is effected by the quality control of observations and the proportion of the 1 ° x 1° bin that is sea-water. The average coverage of Jason1 (see Fig. 18.9a) does not exceed ~0.45. The

Fig. 18.9 The average SLA observations per 1 ° x 1 ° bin per day over the period 1 January 2009 to 1 March 2010 obtained from a Jasonl. b ENVISAT. c Jasonl and Jason2 and d Jasonl and Jason2 and ENVISAT. The along-track observations have been normalised to ~1 observation/50 km

tandem mission of Jasoni and Jason2 (see Fig. 18.9c) show the overall improvement in the spatial distribution of coverage compared with Jason1.

The normalized frequency distribution of SLA observation coverage corresponding to each Fig. 18.9a-d is plotted in Fig. 18.10. Due to the relatively coarse orbit sampling of Jason1, the 23% of 1° * 1° bins are not sampled at all. With the introduction of the tandem missions Jason1 and Jason2 the number of 1° * 1° bins that are not sampled drops to ~8%. The Envisat mission samples virtually all of the bins. The mode of each distribution curve is (0.15; 0.2; 0.35; 0.5) obs. per bin per day for Envisat; Jason1 (ignoring the zero peak); Jason1 and Jason2; Jason1, Jason2 and Envisat respectively. The number of obs. per day for all bins never exceed 0.73 for the three altimeters. The distribution shows that 50% of bins in the Australian region have a coverage of better than (0.15; 0.15; 0.3; 0.45) obs. per bin per day for Envisat; Jason1; Jason1 and Jason2; Jason1, Jason2 and Envisat respectively.

Sea level anomaly products are processed in two to three modes dependent on the satellite which vary in quality and timeliness. The quality is determined by the

Frequency distribution of altimetry observations

Frequency distribution of altimetry observations

Fig. 18.10 Normalized frequency distribution of altimetry observations per 1° x 1° bin per day for the Australian region and satellite combinations shown in Fig. 18.9

quality to which the Geophysical Data Record (GDR) is estimated as well as the precision of other correction terms. Precise orbit positions are determined some time after real-time (e.g., 60 days) and are only relevant to hindcasting. Interim GDR (IGDR) target a faster orbit determination that is less accurate but can be delivered within 2-3 days (Jason series) and 3-5 days (Envisat). For the Jason series additional on-board satellite instrumentation allow an Operational GDR (OGDR) product to be delivered within 24 h of real-time. Due to instrument failure on Ja-son-1 the OGDR was unavailable but has been restored on the AVISO server. Following the launch of Jason-2 this product is now also available. A summary of events related to operational satellite altimetry can be found online from AVISO ( In summary, the complete orbit of the Jason1 and Jason2 IGDR product is available between 3 and 12 days behind real-time, the complete orbit of Envisat IGDR product is available between 5 and 40 days behind real-time and Jason2 OGDR product is available 1-10 days behind real-time. Due to the reduced quality of the IGDR and OGDR products as well as the timeliness of the products it has been determined that the analysis performance from four real-time altimeters is equivalent to two delayed mode altimeters (Pascual et al. 2009).

18.6 Real-Time Forcing System

The ocean is a forced dissipative system, where the forcing is largely through the flux of mass, heat and momentum through the air-sea interface. Atmospheric fluxes are available from operational numerical weather prediction systems that are mature and provide robust and consistent performance. However, the performance of atmospheric fluxes is relatively low compared with the state variables due to the limited direct observations of fluxes and errors in boundary conditions. The properties that influence the selection of atmospheric flux products and flux parameterizations for ocean prediction is summarized in Table 18.5.

The oceans relatively large inertia, thermal inertia compared with the atmosphere mean that on short timescales air-sea fluxes are a relatively small perturbation to the ocean state at the surface and decays with depth. Even under extreme conditions, such as tropical cyclones, the surface temperature in the cold wake has been observed to be between 1°C and 6°C (Price 1981) and that the majority of the temperature change is due to entrainment and mixing of the ocean water masses in response to the momentum fluxes rather than changes due to surface heat flux. The momentum flux local to the atmospheric winds is largely transferred into the gravity waves which radiate from the source region. Local momentum transfer from high winds occurs through Lang-muir circulation (McWilliams et al. 1997), wave breaking (Melville 1996) and wave dissipation which persist during the wind event and is a function of wave age (Drennan et al. 2003). A large fraction of the energy radiates away and dissipates through small scale turbulence and topographic interactions in locations remote from the winds.

In the coastal region, the reduced volume of seawater is more sensitive to atmospheric fluxes. Storm-surge and coastal trapped waves (e.g., coastal Kelvin waves) are a result of horizontal mass flux into the coast as an Ekman response to the applied wind stress and lower atmospheric pressure (see Fig. 18.3h, i). Coastal up-welling of more dense, often cool and nutrient water masses are a response to mass flux away from the coast from an applied wind stress in the opposing direction (see

Table 18.5 Properties of the atmospheric flux products that impact the ocean forecasting system Real-time forcing system

Real-time surface fluxes

Robust, well-defined and consistent performance

Period, resolution and region of forecast systems

Global, regional, sub-regional

Forecast skill curve

Boundary conditions, persisted SST, surface roughness

Land-sea-ice masks

Atmospheric boundary layer, cloud and radiation physics

Observational constraints (e.g., scatterometry)

Hindcast fluxes

Performance during data assimilation

Flux parameterisation

Fixed boundary condition flux products

Forecast atmospheric state with dynamic ocean boundary conditions

Coupled air-sea or air-wave-sea

Ocean dynamics

Sensitivity of the ocean state to surface fluxes

Sensitivity of ocean forecast error to surface flux errors

Fig. 18.3a, b). The coastal region also has less heat capacity due to its reduced depth and is more sensitive to diurnal warming. The coast is also a region where atmospheric precipitation collects over land basins and can discharge from river mouth as a less dense freshwater plume. All of these processes have timescales comparable to those of the atmospheric weather and can produce observable changes to the ocean state and circulation of the coastal region. The skill of coastal ocean state forecasts is therefore sensitive to the skill of the atmospheric fluxes.

The atmospheric fluxes for ocean forecasting systems are obtained from operational numerical weather prediction systems (e.g., GASP; Seaman et al. 1995). A typical configuration for NWP is to perform an analysis every 6 h and a forecast every 12 h. During ocean hindcasting, 24 h of analysis fluxes can be composed of four 6 h analyses. Common averaging periods for surface fluxes are 3 h and 6 h. Atmospheric forecasting is typically composed of a suite of global and multiply nested regional prediction systems. In general, the horizontal resolution of the atmospheric models are coarser than the comparable ocean model and require regridding. One of the key discrepancies between models of differing resolutions is the mismatch in land-sea mask. A comparison of the land-sea masks from GASP (0.75°) and Ocean Forecast Australia Model (OFAM; Schiller et al. 2008) (0.1°) is shown in Fig. 18.11. There are specific areas that show where some area correspond to New Land (Sea mask in the source and land in the target) or New Sea.

OFAM mask to GASP mask differences

OFAM mask to GASP mask differences

flOE 120E 130E 140E 150E 160E 110E 120E 130E 140E 150E Longitude

Fig. 18.11 A comparison of the land-sea masks of GASP (Seaman et al.) and OFAM (Schiller et al.). The four combinations both land (brown), both sea (blue), GASP land/OFAM sea (yellow) and GASP sea/OFAM land (red) ignoring ice masks in the Australian region flOE 120E 130E 140E 150E 160E 110E 120E 130E 140E 150E Longitude

Fig. 18.11 A comparison of the land-sea masks of GASP (Seaman et al.) and OFAM (Schiller et al.). The four combinations both land (brown), both sea (blue), GASP land/OFAM sea (yellow) and GASP sea/OFAM land (red) ignoring ice masks in the Australian region

In general, the magnitude of atmospheric fluxes across the land-sea boundaries is discontinuous largely due to the change in surface roughness, albedo and heat capacity. The magnitude of discontinuities varies with each variable and with the time of day. As the coastal ocean state is sensitive to atmospheric fluxes the fluxes over land need to be explicitly removed. Replacing the fluxes into New Sea locations is commonly performed by a Laplacian operation with the boundary conditions of the fluxes over sea points as this is computationally inexpensive. This method does not apriori preserve the alignment of winds or other properties with the coastline. There are many software packages that perform regridding including many of the earth system couplers (e.g., OASIS, Redler et al. 2010) however it is important to test these schemes and not assume that they will satisfy the requirements. An important property for regridding is to conserve the total integral of the field from the source grid to the target grid. The OASIS coupler has implemented the Spherical Coordinate Remapping and Interpolation Package (SCRIP; Jones 1999) as a regridding option. Another simple approach is to use an integral variable where the control volume integrals Eq. 18.1a are summed to form the discrete integral variable Eq. 18.1b, (Leonard 1995) which is exact at each cell interface and implicitly conserves the fluxes on the source grid.

Regridding the original cell volume to a finer grid resolution, AX with an index k e [1,K] with the constraint that IAx=KAX, can be performed by constructing an equivalent integral variable through interpolation of the integral variable as,

Was this article helpful?

0 0
Guide to Alternative Fuels

Guide to Alternative Fuels

Your Alternative Fuel Solution for Saving Money, Reducing Oil Dependency, and Helping the Planet. Ethanol is an alternative to gasoline. The use of ethanol has been demonstrated to reduce greenhouse emissions slightly as compared to gasoline. Through this ebook, you are going to learn what you will need to know why choosing an alternative fuel may benefit you and your future.

Get My Free Ebook

Post a comment