1. Introduction
Coupled oceanic and atmospheric models for prediction of hurricane intensity and structure change are used for public advisories by forecasters and government officials, who rely on the most advanced weather forecasting systems to prepare for landfall (Marks and Shay 1998). Over the past decade, it is increasingly clear that ocean models must include realistic initial conditions to simulate not only the oceanic response to hurricane forcing (Price et al. 1994; Sanford et al. 1987, 2007; Shay 2009; D’Asaro 2003, Jacob and Shay 2003; D’Asaro et al. 2007; Lin et al. 2008, 2009; Halliwell et al. 2008; Fan et al. 2009; Wada et al. 2009; Jaimes and Shay 2009, 2010) but also the atmospheric response to oceanic forcing (Bender and Ginis 2000; Bao et al. 2000; Shay et al. 2000; Hong et al. 2000; Walker et al. 2005; Lin et al. 2008, 2009; Wada and Usui 2007; Wu et al. 2007; Sandery et al. 2010; Chen et al. 2010; Shay et al. 2011). This study aims to describe a product and method to accurately diagnose the ocean thermal structure prior to hurricane passage in the Atlantic basin.
Case studies show the importance of OHC and ocean thermal structure in the rapid intensification of TCs. In 1995, Hurricane Opal rapidly intensified as it moved over a warm core ring (WCR) in the Gulf of Mexico (GOM) under favorable atmospheric conditions due to Opal’s juxtaposition to an upper-level trough (Shay et al. 2000; Bosart et al. 2000). Over a 14-h period, Opal’s unforecasted intensification from a category 1 to a category 4 hurricane occurred directly over a warm feature that enhanced air–sea feedback during passage. Coupled modeling studies showed that Opal’s central pressure was 10 mb higher in the absence of the WCR (Hong et al. 2000).
Over the past few decades, considerable attention has focused on the residual cold wake in the upper ocean after hurricane passage, which is primarily induced by entrainment of cooler thermocline waters by shear-induced mixing (Price 1981, 1983; Sanford et al. 1987; Shay et al. 1992; D’Asaro et al. 2007). Strong wind-induced near-inertial currents generate significant shear across the mixed layer base that forces the ocean to mix (e.g., lowers the Richardson number to below criticality), cooling the upper ocean and lowering the OHC. In the absence of deep warm ocean features (e.g., WCR), the resultant SST cooling typically ranges from 3° to 5°C. Under some circumstances, the rate of thermal energy advected into a region can be significant, especially in frontal regimes where upper-ocean currents are known to be energetic (Jacob and Shay 2003). The northwestern Caribbean Sea is the source of the deep warm Loop Current (LC) and its associated WCR field in the GOM. In these waters, the upper-ocean stratification is usually weak and 26°C water extends well over 100 m. During Hurricanes Isidore and Lili (2002), Shay and Uhlhorn (2008) showed that the shear-induced mixing was weak because the LC horizontal advection tendencies arrested the development of a near-inertial wake and was corroborated by a model study (Uhlhorn and Shay 2013). The SST cooling was also less than 1°C during WCR separation from the LC after the passage of Hurricanes Katrina and Rita (2005) (Jaimes and Shay 2009, 2010). Note that these oceanic features were not noticeable from the SST distributions, which were uniform at 30°C over most of the region prior to Katrina (Scharroo et al. 2005; Sun et al. 2006; Mainelli et al. 2008; Shay 2009; Shay et al. 2011; Jaimes and Shay 2009, 2010; Goni et al. 2009). These studies also point to the importance of initializing coupled models with realistic warm and cold ocean features in the global oceans (Ali et al. 2007; Lin et al. 2008, 2009; Halliwell et al. 2008, 2011; Wada and Usui 2007; Chen et al. 2010; Sandery et al. 2010; Zheng et al. 2008). Careful attention must be focused on these regimes where significant cooling does not occur because SSTs are the critical driver for moisture disequilibrium and enthalpy fluxes into the hurricane boundary layer (Emanuel 1986; Cione and Uhlhorn 2003).
The previous operational climatology for the Atlantic basin, developed by Mainelli-Huber (2000), blended older versions of the World Ocean Atlas (WOA) and the Generalized Digital Environmental Model (GDEM) at 0.5° resolution. Mainelli et al. (2008) used OHC as a predictor of intensity in the Statistical Hurricane Intensity Prediction Scheme (SHIPS; DeMaria et al. 2005), which improved intensity forecasts of category 5 storms by an average of 5%–6%, where the largest improvement (~22%) was found for Ivan (2004). The study herein improves upon Mainelli-Huber (2000) and Shay and Brewster (2010) with the development of the Systematically Merged Atlantic Regional Temperature and Salinity (SMARTS) Climatology, which analytically determines a weighting function to optimally blend the GDEM and WOA 2001 (WOA01) climatologies. Weights are dependent upon the individual climatology’s success in calculating oceanographic parameters using the model compared to over 50 000 in situ profiles. The resultant daily SMARTS Climatology has ¼° resolution, which better resolves features such as the LC and Gulf Stream (Meyers 2011). The higher resolution increases the predictability of the location and extent of coherent mesoscale ocean structures from objectively analyzed satellite altimetry data using the approach of Mariano and Brown (1992).
Several websites and studies developed similar products of hurricane (or TC) heat potential using a dynamic sea surface height anomaly (SSHA) field and the OHC concept of Leipper and Volgenau (1972) (e.g., Ali et al. 2007; Lin et al. 2008; Goni et al. 2009; Wada and Usui 2007; Wu et al. 2007; Sandery et al. 2010). Price (2009) introduced the concept of mixing depth, related to the average ocean temperature of the upper 100 m, which provides a measure of the ocean’s influence on hurricanes moving from deep to shallow water. This study calculates OHC using a two-layer model (Shay and Brewster 2010) because of its direct relevance to TC intensity and broader uses to study seasonal oceanic cycles and potential impacts on coral reefs.
In this manuscript, section 2 discusses data resources, including various in situ observing platforms. In section 3, the model and methods for blending the GDEM and WOA01 climatologies are described. Improvements to Mainelli-Huber (2000) and a detailed analysis of the performance of the SMARTS Climatology are in section 4. Section 5 explores the sensitivity of OHC estimations, followed by concluding remarks in section 6.
2. Data resources
To create the SMARTS Climatology, two oceanographic climatologies were analyzed to determine their individual strengths in calculating the depth of the 20°C isotherm (D20), D26, MLD, and OHC using a two-layer model. The GDEM version 3.0 and the WOA01 are used here (Carnes 2009; Stephens et al. 2002). While these two climatologies were similar on the large scale, thermal structural differences of over 30% in regions of interest necessitated a thorough performance analysis of the individual climatologies (Fig. 1). Climatological values of D20, D26, MLD, and reduced gravity were extracted from these datasets for use in the two-layer model. Both of these climatologies were at 1/4° resolution, which was an improvement to the previous OHC climatology (Mainelli-Huber 2000). The increased resolution resolved frontal features near the LC, GOM eddy shedding region, and Gulf Stream. GDEM and WOA01 did not assimilate any of the in situ data used in this study, allowing for unbiased evaluation of the two climatologies.
(left) Satellite-derived OHC prior to Hurricane Katrina (2005) using the GDEM climatology. (right) Differences in coincident OHC calculated from GDEM and WOA01 climatologies. The National Hurricane Center’s best-track position and intensity data for Hurricane Katrina are overlaid in both figures.
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
a. World Ocean Atlas 2001
WOA01 is a product of the National Oceanic and Atmospheric Administration’s (NOAA) National Oceanographic Data Center. The monthly climatology was computed globally from the World Ocean Database, composed from over 7 million profiles primarily from conductivity–temperature–depth (CTD), mechanical and expendable bathythermographs, and moored buoy data (Stephens et al. 2002; Conkright et al. 2002). Profile data were interpolated to 33 standard depths from the surface to 5500 m. Data were grouped into 1° × 1° boxes, from which the local arithmetic mean, standard deviation, and standard error of the mean are computed for all months. These data were objectively analyzed with a Gaussian weighting scheme to ¼° resolution from 60°S to 60°N at all longitudes following Barnes (1964). The climatology was smoothed with a median smoother using data from five adjacent grid boxes from the data point.
b. GDEM version 3.0
GDEM is a monthly ¼° climatology of temperature and salinity developed by the Naval Oceanographic Office. GDEM is a four-dimensional steady-state model of interpolated ocean profiles. The profile data source was the Master Oceanographic Observation Data Set, containing over 5.5 million profiles dating back to 1920 (Carnes 2009). Profiles were calculated globally at 78 depths down to 6600 m in all ocean regions where depths are deeper than 100 m. Observational data were gridded onto each depth surface over the entire domain and were objectively analyzed to a standard ¼° grid. Values of temperature, salinity, and their respective variances were calculated at all grid points. The GDEM domain extends from 40°S to 60°N globally.
c. Altimetry data
Satellite radar altimeters measure sea surface height along repeated tracks. These polar-orbiting satellites repeat their exact paths on a 10-, 17-, or 35-day basis depending on the mission (Fig. 2). At least two active satellites are required for accurate mesoscale applications of altimetry data (Rosmorduc and Hernandez 2003). Typically, two or more altimeters were operational at any given time (Fig. 3). Daily altimeter-derived SSHA data were acquired from the Naval Oceanographic Office, which measure a departure from the Collecte Localisation Satellites’ combined mean dynamic topography mean background SSH field (Jacobs 2006). These fields have been processed to remove orbit errors, ionosphere effects, barometric pressure, and electromagnetic biases on a daily basis (Lillibridge et al. 2011). The operational product does not correct high-frequency barotropic motions to meet latency requirements.
Tracks of all available altimetry data. Altimeters repeat their exact tracks every 10 (blue), 17 (red), or 35 (green) days.
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
Available satellite altimetry from 1998 to 2011. At least two altimeters are always in service, providing global coverage every 10 days.
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
d. Sea surface temperatures
Global SSTs were acquired from Remote Sensing Systems (Gentemann et al. 2009), which optimally interpolates (OI) data from microwave radiometers aboard the Tropical Rainfall Measuring Mission (TRMM) and NASA’s Aqua polar-orbiting satellite. The TRMM Microwave Imager (TMI) radiometer covers a region from 40°S to 40°N, while Aqua’s Advanced Microwave Scanning Radiometer for Earth Observing System (EOS; AMSR-E) produces global SSTs. The TMI–AMSR-E OI data were gridded daily at ¼° resolution. SST retrieval was inhibited in regions of rain, sun glitter, and close to land, and data gaps were filled in by interpolation (Reynolds and Smith 1994).
e. In situ profiles
Over 50 000 profiles collected from 2000 to 2010 provided the in situ oceanographic data to evaluate the empirical calculation of D20, D26, MLD, and OHC (Fig. 4). Profiles covering the Atlantic basin were collected from drifting Argo profilers, shipborne expendable bathythermographs (XBTs), airborne expendable bathythermographs (AXBTs), airborne expendable conductivity–temperature–depths (AXCTDs), and airborne expendable current profilers (AXCPs), as well as long-term Prediction and Research Moored Array in the Tropical Atlantic (PIRATA) moorings.
Distribution of in situ data from Argo floats, XBT transects, airborne profilers (e.g., AXBTs) and PIRATA moorings in the northern Atlantic Ocean.
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
1) XBTs
XBTs are deployed by vessels of opportunity during cross-basin cruises, creating a “snapshot” of the spatial variability of temperature in the water column. The XBT payload contains a resistor, whose data are transmitted by a fine copper wire back to the ship. The voltage is a measure of the resistance of a conductive material, which is directly related to the environmental temperature surrounding the thermistor. Older XBTs are accurate to approximately ±0.5°C, while newer models are accurate to ±0.2°C. The probe is precision weighted and spin stabilized, which allows for a predictable rate of decent, leading to depths accurate to ±2% in the upper 200 m (Singer 1990).
Quality controlled XBT data used in this study were acquired from NOAA’s Atlantic Oceanographic and Meteorological Laboratory. Profiles are quality controlled by testing for duplicate profiles; comparing profiles to climatological means to identify potential outliers; comparing profiles to other observations from the same cruise to test for spatial consistency between drop points; testing vertical temperature sections between profiles from the same cruise to test for internal consistency; and, mapping monthly averages of SST, temperature at 150 m, and average temperature of the upper 400 m to find outliers (Bailey et al. 1994). Data flagged as “inconsistent,” “doubtful,” or “bad” were excluded from this study.
2) PIRATA moorings
The PIRATA array off the African coast provides temperature data in the upper ocean (Servain et al. 1998; Bourlès et al. 2008). Temperature data from the surface to 500 m are measured by a thermistor chain at a vertical resolution of 20 m in the top 140 m, and at depths of 180, 300, and 500 m. The ratio of mooring length to water depth is kept below 0.985, so there is minimal change of depths of the thermistors based on buoy movement. Daily temperature data accurate to ±0.01°C at the surface and ±0.09°C subsurface are sampled for analysis. When temperature changes of over 5°C occur from the previous day with unrealistic vertical temperature gradients, the data are not used in the analysis.
3) Argo profilers
Basinwide profiles of temperature and salinity are available from the Argo buoy array. At any given time, approximately 3000 Argo floats are deployed over the world’s oceans. The typical sampling mode of a float is to descend to a “parking depth” of 1000 m, where it remains for approximately 10 days. At this point, a hydraulically controlled bladder empties and the profiler descends to 2000 m. The bladder inflates, causing the profiler to rise and start recording temperature and salinity data. On its path to the surface, data are stored at approximately 200 depths. Once the profiler reaches the surface, satellites determine the position of the float and profile data are transmitted from the float to the satellite. Salinity measurements, via conductivity ratios, are accurate to ±0.01 psu. Temperature measurements are accurate to ±0.005°C, with depths accurate to ±5 m (Carval et al. 2010). The real-time data product is used because 19 automated quality control checks of the real-time data are performed to ensure a realistic profile. Argo data are valuable in assessing the SMARTS Climatology because data are collected in all regions of the open Atlantic, and they cover temporal and spatial data gaps compared to repeat XBT transects from ships. However, observations are sparse in the Gulf of Mexico because of the short residence time of floats in the LC and geography restricting the profile from completing a full 2000-m profile.
4) Airborne expendables
Airborne oceanography permits sampling of mesoscale and synoptic features at time scales such that the feature does not significantly change over the sampling period (Bane and Sessions 1984). The airborne expendable data fill in a major data gap in the LC and the Gulf of Mexico, as flights typically target regions of high eddy variability and locations of anomalously high or low SSHAs from radar altimeters. AXBTs, AXCTDs, and AXCPs are launched by NOAA WP-3D aircraft during reconnaissance and research flights. When the profiler is launched by the aircraft, a parachute is deployed to minimize surface impact. After a minute on the ocean surface, the radio frequency (RF) transmitter turns on, and within the next 20–40 s, the probe is released from the surface transmitter, which remains connected to the descending unit by a thin copper wire. Data are multiplexed up the wire and sent to the aircraft via the RF transmitter, where the audio signals are received and processed. As with the XBT, the AXBT measures the voltages that are converted to temperatures. Both the AXCTDs and AXCPs measure temperature, as well as salinity via a conductivity ratio (AXCTD) and currents (AXCP). The AXBT fall rate is 1.5 m s−1. The fall rates of the AXCTD and AXCP are 2.2 and 4.5 m s−1, respectively. A large fraction of the aircraft data were collected during NOAA hurricane hunter reconnaissance and research missions prior to, during, or after hurricane passage, such as Isidore and Lili in 2002 (Shay and Uhlhorn 2008), Katrina and Rita in 2005 (Jaimes and Shay 2009), and Gustav and Ike in 2008. In addition, over 800 profilers were deployed in support of the Deepwater Horizon (DWH) oil spill of 2010 (Shay et al. 2011).
For quality control, profiles are passed through a median filter to remove spikes in the data. To check for potential gross bias of the data, profiles are compared to surrounding profiles and noisy ones are removed from the dataset. Temperatures from AXBTs and AXCPs are accurate to ±0.2°C, while AXCTDs are accurate to ±0.05°C. Comparisons of AXCTD data to simultaneous ship-based CTD measurements confirm that AXCTDs are a viable method of acquiring accurate hydrographic data (Shay and Brewster 2010). Collectively, the in situ profiles provided the necessary data to optimally create and analyze the SMARTS Climatology.
3. Approach
a. In situ profiles
Each temperature profile was interpolated to 2-m resolution using a cubic interpolation scheme to maintain a realistically shaped profile where vertical sampling density is low. The profiles were interpolated from 2 m, to avoid artificially high ocean skin temperatures, down to the deepest observation. D20 and D26 were extracted from the profile by linearly interpolating between the two depths surrounding the isotherm of interest of the interpolated profile.
MLD was determined by finding the deepest point of the interpolated profile within 0.5°C of the temperature at 2 m. This threshold temperature was chosen following Monterey and Levitus (1997), and it was outside the precision of the instrumentation. Strictly using a temperature difference from a near-surface value was preferred to using a temperature gradient threshold due to the large spatial variability of the structure of the thermocline (de Boyer Montégut et al. 2004). Temperature inversions were not important because these typically only occur in polar regions, well outside of the domain of this study.
b. Objective analysis
SSHA data were unevenly spaced over the basin, which necessitated an objective analysis (OA) scheme to interpolate data from multiple altimeters to a common grid. The parameter matrix algorithm of Mariano and Brown (1992) grids nonstationary fields using time-dependent correlation functions. SSHA data from 5 days before and 5 days after the date of interest were used, ensuring basinwide coverage by at least the 10-day altimeters. Parameters for spatial and temporal correlation scales followed Mainelli-Huber (2000) and are presented in Table 1. The OA was performed locally point by point, using 20 influential data points determined by the correlation model from Mariano and Brown (1992).
The previous OA technique assumed a uniform drift velocity over the entire Atlantic basin to the west-southwest (0.03° day−1 westward, 0.01° day−1 southward). However, drift velocities vary spatially with latitude and along bathymetric features (Chelton and Schlax 1996). To assess the spatial variability of feature drift velocity, the OA was applied to SSHAs from 2005 to 2007 without any drift velocity correction. The resulting SSHA fields were plotted in Hovmöller diagrams with a north–south and east–west cross section for 10° × 10° grid boxes (Fig. 5). If the SSHA showed clear trends, then the drift velocity for that grid was determined by taking the average speed of three subjectively chosen trend lines. The entire SSHA dataset was reanalyzed to make adjustments using these estimated drift velocities.
Hovmöller of SSHA at 15°N in the Caribbean Sea for 2 years. Drift velocities are on the order of 0.1° day−1. The prior OA technique assumed a basinwide velocity of 0.03° day−1.
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
c. Daily climatology
GDEM and WOA01 climatologies provided the four-dimensional data needed to construct the final blended climatology necessary for the empirical two-layer model used to calculate ocean thermal structure. For both GDEM and WOA01, extracted profiles of temperature and salinity were interpolated to a 1-m resolution using a cubic interpolation scheme. An iterative process identified the depth of the 20° and 26°C isotherms and applied a linear interpolation between the two points surrounding the isotherm depth to approximate the appropriate depth. MLD was defined as the depth where the temperature deviates from the SST by 0.5°C.
A 15-day running mean was used to create the “daily” climatology. For example, to create the 2 Oct climatology, the September climatology was weighted 6/15 and the October climatology was weighted 9/15.
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
d. Two-layer model
Price (2009) argues that the depth-averaged temperature at 100 m may be a better predictor for the tropical cyclone intensity change than OHC. In coastal regions, where altimetry has difficulties in resolving SSHA, the averaged 100-m temperature may be a more valid parameter, although many continental shelves have shallower depths. Vertically averaged temperatures may be well below the 26°C threshold (Palmén 1948) that meteorologists believe is important for the air–sea enthalpy fluxes in TCs. Given the variability in the ocean basins, our view is that OHC is a valid parameter for hurricane intensity prediction, as shown by DeMaria et al. (2005) and more recently in Mainelli et al. (2008) for category 5 hurricanes.
e. Creating the SMARTS Climatology—Statistical parameters
The performance of each climatology in determining ocean thermal structure was expected to vary spatially and temporally due to implicit errors in the climatology. Attempts to divide the Atlantic basin into broad analysis regions resulted in minimal differences between skill of the GDEM and WOA01 climatologies. Instead, the Atlantic basin was divided into 5° × 5° boxes in which blending weights were to be determined for each box. In border areas between two regions, there was a 2° linear transitional region to smooth the weighting map and avoid abrupt borders. The 2° buffer zone was larger than the Rossby radius of deformation (~1° in the GOM) to avoid gross distortion of features, as they translated through transitional zones.
f. GDEM and WOA01 weighting maps
The weighting maps resulting from RMSD analysis depicted the locations where either GDEM or WOA01 were better suited for ocean thermal structure calculation. The D20 weighting map was created first, considering the new D20 calculations were needed for estimations of D26, MLD, and OHC. The D20 weighting map was also used for the blended climatology of densities for the upper and lower layers in the two-layer model.
For the hurricane season, the two-layer model calculated D20 better using the GDEM climatology in the GOM and most of the LC regime (Fig. 7). GDEM also outperformed WOA01 along the U.S. coastline, being slightly better in the colder waters north of where the Gulf Stream separates from the coastline near Cape Hatteras, North Carolina. In general, the subtropical waveguide was dominated by GDEM, particularly along the African coast. The southern extent Caribbean Sea and the center of the subtropical gyre tended to be better predicted using WOA01.
Weighting maps for GDEM used to create the SMARTS Climatology for (left) the hurricane season and (right) outside of the hurricane season determined by RMSD analysis. Maps were calculated for (top) D20, (middle) D26, and (bottom) MLD.
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
Outside of the hurricane season, observations were more sparse and the ventilation of D26 at the surface moves southward during the winter and spring months. Again, GDEM outperformed WOA01 in the equatorial waveguide. WOA01 predicted D20 better near the Caribbean Islands and in the western GOM, whereas GDEM only slightly improved calculations along the mid-Atlantic and western Florida coasts.
Once the weighting maps of D20 were calculated, 
g. OHC calculation
4. Analysis of the SMARTS Climatology
To justify the use of the SMARTS Climatology, the resulting blended fields must depict realistic fields of these spatially varying variables. There was a risk of creating an unrealistic field due to sharp transitions between the GDEM and WOA01 climatologies. Visual inspection of sample climatological fields based on a binary blending scheme showed no evidence of the transitional areas (Fig. 8). The SMARTS Climatology improved calculations of ocean thermal structure based on the crude two-layer model by reducing errors for all retrievals (Table 2; Fig. 9). Mainelli-Huber (2000) typically overestimated D20 throughout the entire basin by over 30 m. This total bias was essentially negligible (<1 m) with the SMARTS Climatology, where RMSD was reduced by 37%. In the GOM and LC, RMSD was due to larger depth variability of observations. Across the entire basin, RMSD values decreased by using SMARTS to estimate D26 by 25%, although the overall bias is not reduced.
The resultant SMARTS Climatology for D20 (m) using the weighting scheme based on Fig. 7 for 15 Sep.
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
RMSD and bias (parentheses) of satellite calculation of upper-ocean thermal structure using Mainelli-Huber (2000) and SMARTS Climatology using in situ data from (top) the entire basin and (bottom) only the GOM/LC complex.
RMSD of observed values during hurricane season for D20, D26, MLD, and OHC using (left) the Mainelli-Huber (2000) climatology and (right) the SMARTS Climatology.
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
Using all in situ data from the entire Atlantic Ocean basin, the Mainelli climatology performed well when calculating OHC, with a bias of less than 1 kJ cm−2 and an RMSD of 18 kJ cm−2. SMARTS decreased the RMSD to 15 kJ cm−2. In the GOM, where the ocean thermal structure can affect hurricane intensity, Mainelli overestimated OHC by an average of 20 kJ cm−2, whereas the bias from SMARTS was nearly negligible and RMSD decreased by 35%. Such large changes in OHC calculations could have significant impacts on hurricane intensity forecasting in the SHIPS model (DeMaria et al. 2005). Scatterplots comparing in situ values to SMARTS-derived data showed strong correlations, with values typically falling near the perfect fit curve of slope one (Fig. 10). Histograms of absolute differences between satellite and in situ data showed errors of SMARTS were evenly distributed around the mean bias, with 69% of satellite calculations of D20 within 20 m of the observed value and 89% within 40 m. The negative bias of D26 was visually noticeable in the histogram, as only 65% of D26 SMARTS calculations were within 15 m of the observed value, while 90% were within 30 m. For OHC, 85% of satellite calculations were within 20 kJ cm−2 of observed values.
(top) Density distributions and (bottom) histogram of differences between SMARTS calculated values of (left) D20, (middle) D26, and (right) OHC and in situ values.
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
The two-layer model methodology for OHC retrievals was supported by comparing to retrievals made using a Bayesian scheme. The in situ database was separated into bins of SST (0.2°C), SSHA (3 cm), and climatological D20 (20 m). A Bayesian OHC retrieval was performed by averaging the OHC of in situ profiles with similar observed SST, SSHA, and climatological D20 values. RMSD using the Bayesian retrieval was 12.3 kJ cm−2 as opposed to 10.7 kJ cm−2 when using the two-layer model. Sufficient in situ profiles are not available to increase the dimensions of the database to take into account spatial variability of reduced gravity or MLD.
a. Justification of MLD adjustment
Shay and Brewster’s (2010) estimation of upper-ocean thermal structure assumed a climatological MLD, whereas the updated two-layer model with the SMARTS Climatology uses an adjusted MLD. SMARTS reduced basinwide RMSD of MLD by almost 50%. Observations show regions of higher variability of MLD, particularly in the LC, GOM, and Gulf Stream, which are all characterized by a strong eddy field where MLDs are maintained by a variety of forcing mechanisms (Fig. 11). Air–sea momentum fluxes act to mix the upper ocean, while latent and sensible heat fluxes force buoyancy-driven mixing, and shear instabilities entrain cooler thermocline waters at the base of the mixed layer. While the climatological approach is stable, using a climatological MLD could not capture deep MLDs in WCRs.
Scatterplots of (left) MLD and (right) OHC comparing in situ observations to satellite estimations using (top) the climatological MLD and (bottom) the adjusted MLD. Blue dashed line represents a 1:1 relationship, and the linear regression line is solid blue.
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
Adjusting the MLD following the same procedures used to calculate D26 led to a more realistic field of satellite-estimated MLD. The adjusted methodology allowed for shallow mixed layers that often mix quickly when subjected to strong surface winds via shear instability that were typically not present in the climatology. Additionally, adjusting MLD resulted in balancing the low bias for deep MLD in situ observations. The slope of the regression line between in situ and calculated MLD increased from 0.57 to 0.68, which was an improvement but still well below the 1:1 line. Overall bias of calculations decreased by a magnitude of about 2 m; however, this was accompanied by an increase of RMSD on the same scale. Overall, MLD bias decreased by adjusting MLD because using only the climatology as an MLD estimate could not capture occurrences of particularly deep MLDs. RMSD increased primarily in areas of shallow observed MLD, where the adjustment using the two-layer model was disproportionate to the magnitude of the in situ MLD.
Similar results were observed when considering OHC calculated with the climatological and adjusted values. The slope of the regression line improved from 0.83 to 0.9, yet RMSD increased due to the increased scatter of calculated MLDs. The regression analysis suggested a potential link between MLD and SSHA, yet further analysis could not confirm a direct link basinwide. In this study, the adjusted MLD was used because the regression analysis demonstrated an improvement toward a 1:1 relationship between observed and estimated MLD.
b. Realistic characteristics of satellite field
1) Spatial characteristics—XBT transects
Repeated transects across the Atlantic provided snapshots of vertical temperature structure, which allowed for comparison to SMARTS-derived values. Three annual September transects between 2004 and 2006 across a region east of Florida through the Gulf Stream provided over 120 XBT profiles for analysis. Along this transect, the satellite algorithm using the SMARTS Climatology captured the horizontal variability of OHC within the error of observations (Fig. 12). Most of the variability of OHC could be attributed to east/west changes in SST, considering there was little variability of D26. Satellite calculations of D20 had a shallow bias, particularly near local minima along the 20°C isotherm. The satellite methodology accurately calculated the locations of relative maxima and minima of D20, only the magnitudes were not accurate. The two-layer model often exaggerated the slope of the D20 surface, although the locations of maxima and minima were accurate.
(top) Average satellite-derived OHC on 15 Sep 2005 during three annual XBT transects along the dashed black line in Fig. 8. (bottom) Temperature structure along that transect and the corresponding OHC compared to the coincident satellite-derived average with the appropriate 95% confidence interval. D20 and D26 contours are depicted by white and black lines, respectively.
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
2) Temporal characteristics—Argo profilers
Argo profilers provided valuable data for analysis of the two-layer model with the SMARTS Climatology. The floats completed one full profile of the water column every 10 days, the same as the time window of SSHA data, which ensured a corresponding time series of independently calculated satellite fields. Data from 2008 from an Argo profiler west of the Caribbean Islands contains oceanic characteristics fairly representative of the full dataset (Fig. 13). Notice that D20 typically did not change greatly between subsequent profiles, although satellite calculations showed higher temporal variability in D20. The two-layer model’s calculation of D20 was very sensitive to SSHA changes as expected, often overestimating observed in situ changes but still capturing variability on a monthly time scale. Satellite-derived values of D26 also showed such overestimated fluctuations, which was expected considering its dependence on D20. In situ values of D26 fluctuated enough such that climatology alone could not accurately depict the ocean subsurface structures. Generally, the two-layer model accurately determined the directional change of the 26°C isotherm, although the magnitude of change was often overestimated. Differences in OHC were a result of disparities between in situ and satellite values of D26.
Comparisons of D20, D26, SST, and OHC between Argo profiler 4900600 (solid black) and SMARTS-derived profiler (dashed red) from 2008.
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
5. Limits of OHC predictability
When first introduced, Mainelli-Huber’s (2000) two-layer model approach for OHC calculations was verified with a small dataset of airborne expendables deployed in the GOM during hurricane aircraft reconnaissance. Now with more than two orders of magnitude more observations, the approach could be evaluated basinwide. Even if a perfect climatology existed such that it results in exact depth calculations by satellite altimetry, inaccuracies would still exist in the OHC calculation from the two-layer model. To evaluate the calculation of OHC by trapezoidal approximation, all in situ upper-ocean profiles were simplified to a homogeneous mixed layer with a temperature equal to the SST and a thermocline with constant stratification (dT/dZ) from the base of the mixed layer to D26 (Fig. 14). The resulting OHC from simplified profiles was then compared to the in situ profiles, effectively setting the limit of accuracy of the fully satellite-derived OHC.
Trapezoidal (black dashed) and satellite estimations are compared to an in situ profile (black solid) in the GOM prior to the passage of Hurricane Gustav in 2008. Trapezoidal approach uses in situ values of SST, MLD, and D26 to give a rough upper-ocean temperature profile. Derived profiles show sources for errors due to satellite SST (green dashed) and estimated depths (blue solid), and their combined error (red dashed).
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
It was expected that the trapezoidal calculation overestimated OHC because of the typical concavity of the temperature structure; profiles were typically most strongly stratified at the top of the thermocline and stratification weakened at greater depth. Basinwide, the trapezoidal approximation overestimated OHC by 2.5 kJ cm−2. In the GOM and LC, most observations occurred during hurricane season such that SSTs, and therefore OHC, were artificially high. The trapezoidal calculation led to a high bias of 5.7 kJ cm−2 in the GOM/LC complex. This bias led to RMSD values basinwide of 4.1 kJ and 5.7 kJ cm−2 in the GOM/LC. In essence, this sets the bar for what can be expected for OHC calculations using the previously described two-layer model. Even if predictability of MLD and D26 were perfect, differences would still exist between in situ observations and satellite estimation. Further insights into the source of errors could be found by replacing satellite-observed parameters with in situ values one by one (Table 3). Basinwide, satellite depth estimates were responsible for about 70% of the remaining error, while satellite SST inaccuracies accounted for the remaining 30%.
Errors arising from the trapezoidal approach to calculating OHC from satellite-derived depths and SST. Satellite depths are used with in situ SST, and satellite SST is used with in situ D26 and MLD. Last column uses both satellite depths and temperatures.
An accurate representation of SSTs was critical for a realistic calculation of OHC by satellite. A complete analysis was necessary to justify the use of AMSR-E for OHC calculation. AMSR-E was compared to two daily Reynolds SST products optimally interpolated to a 1/4° grid (Reynolds et al. 2007). The first used the Advanced Very High Resolution Radiometer (AVHRR) infrared SST data. The second combined the AVHRR data with AMSR microwave data (AVHRR–AMSR). Both Reynolds products incorporated in situ data from vessels and buoys. It should be noted that the in situ XBT and Argo temperatures used in this comparison were taken at 2 m, which varies from the skin temperature measured by the radiometers.
RMS differences and bias between each satellite product and in situ SSTs were of primary importance for application to OHC calculations. The AVHRR products were compared with AMSR-E only where observations were available from both sources and in situ SST was greater than 26°C. The AVHRR-only data had a low bias of 0.14°C and an RMSD of 0.50°C. AMSR-E outperformed AVHRR, with a smaller low bias of 0.03°C and an RMSD of 0.40°C. Not surprising, adding the AMSR data to AVHRR improved the satellite observations. RMSD values were similar between AMSR-E (0.39°C) and AVHRR–AMSR (0.39°C). The postprocessed AMSR-E data had errors that were smoothly distributed, whereas the AVHRR data were unevenly distributed around zero (Fig. 15).
Distribution of differences between in situ near-surface temperatures and satellite SSTs from RSS–AMSR-E and Reynolds–AVHRR–AMSR. Bins are centered around 0° at 0.1°C increments. Curves with outlined markers show the expected distribution if the distributions were Gaussian.
Citation: Journal of Atmospheric and Oceanic Technology 31, 1; 10.1175/JTECH-D-13-00100.1
6. Summary and concluding remarks
An extensive dataset containing XBTs, Argo profilers, moorings, and airborne expendable profilers confirmed that the SMARTS Climatology improved the estimation of the upper-ocean thermal structure in a two-layer framework compared to Mainelli-Huber (2000). The updates to versions of the WOA01 and GDEM climatologies, as well as using a daily climatology, produced most of the improvements to satellite estimations. Further improvements were achieved by systematically blending WOA01 and GDEM depending on regional performance.
Typically, the two-layer framework correctly diagnosed the directional change of D20; however, the magnitude of the adjustment to climatology was overestimated, leading to overestimated maxima and underestimated minima of D20, D26, and OHC. This error could be attributed to small climatological values of reduced gravity, which may be accounted for by updating the algorithm for the density calculation.
The algorithm for satellite estimations of OHC has been adjusted to include a varying MLD. In situ observations in the GOM showed that the LC and WCRs are characterized by deeper MLDs than in CCEs, such that a correlation between SSHAs and MLD anomalies would be expected. This characterization breaks down outside of the GOM, LC, and Gulf Stream because the momentum and air–sea fluxes that maintain the ocean mixed layer are not included. Nonetheless, the adjusted MLD improved OHC calculations basinwide by about 5%.
OHC estimations were quite sensitive to satellite SST values and the approach to estimating OHC. Using in situ values to frame the trapezoidal upper-ocean structure accounted for nearly 33% of the error in OHC estimation. In the GOM, OHC was more sensitive to satellite SSTs than depth estimates due to deep thermal structure, although basinwide errors were attributed to depth estimates more than SST errors.
Not surprisingly, the regions with the greatest variability of ocean thermal structure also had the largest differences between observations and satellite estimates. The LC, GOM, and Gulf Stream were the areas of largest variability due to significant eddy activity, and also the primary regions of interest for accurately estimating OHC for hurricane intensity forecasting.
The two-layer framework assumes atmospheric forcing is negligible, which is not valid in the hurricane environment. Mechanical mixing forced and shear instability at the base of the mixed layer can be primarily attributed to surface wind stress. For example, Hurricane Ike had a large wind field that forced extensive deepening of the ocean mixed layer by over 80 m in the LC to the right of the storm track. This deepening was not reflected in SSHAs due to a lack of coincident altimeter tracks across Ike’s wake. Missing this deepening caused satellite estimates of OHC to be too low in the poststorm surveys, leading to overestimates of cooling in the LC by over 30 kJ cm−2 after Hurricane Ike. When analyzing prestorm and poststorm conditions measured purely by satellite observations, deepening of the mixed layer must be considered because the estimated heat loss will likely be overestimated.
The new estimations of OHC using the SMARTS Climatology require the recalculation of SHIPS coefficients (DeMaria et al. 2005) due to significant changes in OHC relative to using the Mainelli-Huber (2000) climatology, particularly in the LC and GOM, where SMARTS estimations of OHC are typically smaller. Additionally, the current minimum threshold of 60 kJ cm−2 for OHC to contribute to the SHIPS intensity forecast could be reduced to be more consistent with observations.
The framework developed for the creation of the SMARTS Climatology is being expanded to the North Pacific Ocean for typhoon intensity forecasting. The general methodology, however, has more than 267 000 thermal profiles to evaluate the approach or more than a factor of 5 than those used here. Differences in thermocline gradients between basins must be accounted for as described in Shay and Brewster (2010). Long-term moorings, Argo profiles, XBT transects, and airborne expendable profiles are available in these basins for in situ comparisons.
Additionally, increased resolution of SST products will potentially create a more accurate field for OHC estimation considering the sensitivity to SST as found in this study. Other improvements to accuracy of the two-layer framework could be expected by a more thorough calculation of reduced gravity values across the basin using density profiles from Argo floats.
A study on the impacts of OHC and upper-ocean thermal structure on tropical cyclone intensification can be completed with the SMARTS Climatology. A study of surface enthalpy fluxes in different oceanic regimes using collocated AXBTs and GPS sondes in the hurricane environment would be valuable to analyze the subsurface ocean’s effects on hurricane intensification.
Starting in the summer of 2012, NOAA’s National Environmental Satellite, Data, and Information Service (NESDIS) implemented the SMARTS Climatology and two-layer OHC calculation technique in near–real time (Shay et al. 2012). In this application, NESDIS is using a blended daily SST product observed by Geostationary Operational Environmental Satellite and Polar-orbiting Operational Environmental Satellite platforms (Harris and Maturi 2012). OHC data are currently available through NOAA’s Office of Satellite and Product Operations.
The authors are grateful for the support from NOAA NESDIS in supporting this applied research, which has been successfully transitioned to operations. In addition, we also appreciate the support for the basic research used here from NASA (Grant NNX09AC47G) as part of the hurricane science team as part the Genesis and Rapid Intensification Processes experiment and NSF (Grant AGS-04-44525). Microwave OI SST data are produced by Remote Sensing Systems and sponsored by the National Oceanographic Partnership Program (NOPP), the NASA Earth Science Physical Oceanography Program, and the NASA MEaSUREs DISCOVER Project. Data are available online (www.remss.com). Processed SSHA fields from ALPS are provided by NAVOCEANO for these calculations. We appreciate the insights and constructive comments of all three anonymous reviewers, which improved the quality of the manuscript.
REFERENCES
Ali, M. M., , Jagadeesh P. S. V. , , and Jain S. , 2007: Effects of eddies on Bay of Bengal cyclone intensity. Eos, Trans. Amer. Geophys. Union, 88, 93–95, doi:10.1029/2007EO080001.
Bailey, R., , Gronell A. , , Phillips H. , , Tanner E. , , and Meyers G. , 1994: CSIRO cookbook for quality control of expendable bathythermograph (XBT) data. CSIRO Marine Laboratories Rep. 220, 75 pp.
Bane, J. M., Jr., , and Sessions M. H. , 1984: A field performance test of the Sippican deep aircraft-deployed expendable bathythermograph. J. Geophys. Res., 89 (C3), 3615–3621.
Bao, J.-W., , Wilczak J. M. , , Choi J.-K. , , and Kantha L. H. , 2000: Numerical simulations of air–sea interaction under high wind conditions using a coupled model: A study of hurricane development. Mon. Wea. Rev., 128, 2190–2210.
Barnes, S. L., 1964: A technique for maximizing details in numerical weather map analysis. J. Appl. Meteor., 3, 396–409.
Bender, M. A., , and Ginis I. , 2000: Real-case simulations of hurricane–ocean interaction using a high-resolution coupled model: Effects on hurricane intensity. Mon. Wea. Rev., 128, 917–946.
Bosart, L. F., , Velden C. S. , , Bracken W. E. , , Molinari J. , , and Black P. G. , 2000: Environmental influences on the rapid intensification of Hurricane Opal (1995) over the Gulf of Mexico. Mon. Wea. Rev., 128, 322–352.
Bourlès, B., and Coauthors, 2008: The PIRATA program: History, accomplishments, and future directions. Bull. Amer. Meteor. Soc., 89, 1111–1125.
Carnes, M. R., 2009: Description and evaluation of GDEM-V3.0. NRL Memo. Rep. NRL/MR/7330–09-9165, 24 pp.
Carval, T., and Coauthors, 2010: Argo data management user’s manual. Argo Data Management Team, 83 pp.
Chelton, D. B., , and Schlax M. G. , 1996: Global observations of oceanic Rossby waves. Science, 272, 234–238.
Chen, S., , Campbell T. J. , , Jin H. , , Gaberšek S. , , Hodur R. M. , , and Martin P. , 2010: Effect of two-way air–sea coupling in high and low wind speed regimes. Mon. Wea. Rev., 138, 3579–3602.
Cione, J., , and Uhlhorn E. , 2003: Sea surface temperature variability in hurricanes: Implications with respect to intensity change. Mon. Wea. Rev., 131, 1783–1796.
Conkright, M. E., and Coauthors, 2002: Introduction. Vol. 1, World Ocean Database 2001, NOAA Atlas NESDIS 42, 167 pp.
D’Asaro, E. A., 2003: The ocean boundary layer below Hurricane Dennis. J. Phys. Oceanogr., 33, 561–579.
D’Asaro, E. A., , Sanford T. B. , , Niiler P. P. , , and Terrill E. J. , 2007: Cold wake of Hurricane Frances. Geophys. Res. Lett., 34, L15609, doi:10.1029/2007GL030160.
de Boyer Montégut, C., , Madec G. , , Fischer A. S. , , Lazar A. , , and Iudicone D. , 2004: Mixed layer depth over the global ocean: An examination of profile data and a profile-based climatology. J. Geophys. Res., 109, C12003, doi:10.1029/2004JC002378.
DeMaria, M., , Mainelli M. , , Shay L. K. , , Knaff J. A. , , and Kaplan J. , 2005: Further improvements to the Statistical Hurricane Intensity Prediction Scheme (SHIPS). Wea. Forecasting, 20, 531–543.
Emanuel, K. A., 1986: An air–sea interaction theory for tropical cyclones Part I: Steady-state maintenance. J. Atmos. Sci., 43, 585–605.
Fan, Y., , Ginis I. , , and Hara T. , 2009: The effect of wind–wave–current interaction on air–sea momentum fluxes and ocean response in tropical cyclones. J. Phys. Oceanogr., 39, 1019–1034.
Fisher, E. L., 1958: Hurricane and the sea-surface temperature field. J. Meteor., 15, 328–333.
Gentemann, C. L., and Coauthors, 2009: MISST: The Multi-Sensor Improved Sea Surface Temperature project. Oceangraphy, 22 (2), 76–87.
Goni, G. J., and Coauthors, 2009: Applications of satellite-derived ocean measurements to tropical cyclone intensity forecasting. Oceangraphy, 22 (3), 190–197.
Halliwell, G. R., , Shay L. K. , , Jacob S. D. , , Smedstad O. , , and Uhlhorn E. , 2008: Improving ocean model initialization for coupled tropical cyclone forecast models using GODAE nowcasts. Mon. Wea. Rev., 136, 2576–2591.
Halliwell, G. R., , Shay L. K. , , Brewster J. K. , , and Teague W. J. , 2011: Evaluation and sensitivity analysis of an ocean model response to Hurricane Ivan. Mon. Wea. Rev., 139, 921–945.
Harris, A., , and Maturi E. , 2012: Geostationary and polar blended sea surface temperature analysis (geo-polar blended SST) algorithm theoretical basis document. NOAA NESDIS Center for Satellite Applications and Research, Version 2.1, 43 pp.
Hong, X., , Chang S. W. , , Raman S. , , Shay L. K. , , and Hodur R. , 2000: The interaction between Hurricane Opal (1995) and a warm core eddy in the Gulf of Mexico. Mon. Wea. Rev., 128, 1347–1365.
Jacob, S. D., , and Shay L. K. , 2003: The role of oceanic mesoscale features on the tropical cyclone–induced mixed layer response: A case study. J. Phys. Oceanogr., 33, 649–676.
Jacobs, G. A., 2006: Large-scale ocean circulation. Remote Sensing of the Marine Environment, J. F. R. Gower, Ed., Vol. 6, The Manual of Remote Sensing, American Society of Photogrammetry and Remote Sensing, 1–27.
Jaimes, B., , and Shay L. K. , 2009: Mixed layer cooling in mesoscale oceanic eddies during Hurricanes Katrina and Rita. Mon. Wea. Rev., 137, 4188–4207.
Jaimes, B., , and Shay L. K. , 2010: Near-inertial wave wake of Hurricanes Katrina and Rita over mesoscale oceanic eddies. J. Phys. Oceanogr., 40, 1320–1337.
Leipper, D. F., , and Volgenau D. , 1972: Hurricane heat potential of the Gulf of Mexico. J. Phys. Oceanogr., 2, 218–224.
Lillibridge, J., , Scharroo R. , , Jacobs G. , , Russell L. , , and Tabor V. , 2011: Quality assessment of the Jason-2 operational and interim geophysical data records. Mar. Geod., 34, 191–213.
Lin, I.-I., , Wu C.-C. , , Pun I.-F. , , and Ko D.-S. , 2008: Upper-ocean thermal structure and the western North Pacific category 5 typhoons. Part I: Ocean features and the category 5 typhoons’ intensification. Mon. Wea. Rev., 136, 3288–3306.
Lin, I.-I., , Pun I.-F. , , and Wu C.-C. , 2009: Upper-ocean thermal structure and the western North Pacific category 5 typhoons. Part II: Dependence on translation speed. Mon. Wea. Rev., 137, 3744–3757.
Mainelli, M., , DeMaria M. , , Shay L. K. , , and Goni G. , 2008: Application of oceanic heat content estimation to operational forecasting of recent Atlantic category 5 hurricanes. Wea. Forecasting, 23, 3–16.
Mainelli-Huber, M., 2000: On the role of the upper ocean in tropical cyclone intensity change. M.S. thesis, Division of Meteorology and Physical Oceanography, Rosenstiel School of Marine and Atmospheric Science, University of Miami, 73 pp.
Mariano, A. J., , and Brown O. B. , 1992: Efficient objective analysis of dynamically heterogeneous and nonstationary fields via the parameter matrix. Deep-Sea Res., 39, 1255–1271.
Marks, F. D., , and Shay L. K. , 1998: Landfalling tropical cyclones: Forecast problems and associated research opportunities. Bull. Amer. Meteor. Soc., 79, 305–323.
Meyers, P. C., 2011: Development of the systematically merged Atlantic regional temperature and salinity climatology. M.S. thesis, Division of Meteorology and Physical Oceanography, Rosenstiel School of Marine and Atmospheric Science, University of Miami, 98 pp.
Millero, F. J., , and Poisson A. , 1981: International one-atmosphere equation of state of seawater. Deep-Sea Res., 28A, 625–629.
Monterey, G. I., , and Levitus S. , 1997: Climatological cycle of mixed layer depth in the World Ocean. NOAA NESDIS, 5 pp.
Palmén, E., 1948: On the formation and structure of tropical hurricanes. Geophysika, 3, 26–38.
Price, J. F., 1981: Upper ocean response to a hurricane. J. Phys. Oceanogr., 11, 153–175.
Price, J. F., 1983: Internal wave wake of a moving storm. Part I: Scales, energy budget and observations. J. Phys. Oceanogr., 13, 949–965.
Price, J. F., 2009: Metrics of hurricane–ocean interaction: Vertically-integrated or vertically-averaged ocean temperature? Ocean Sci., 5, 351–368.
Price, J. F., , Sanford T. B. , , and Forristall G. Z. , 1994: Forced stage response to a moving hurricane. J. Phys. Oceanogr., 24, 233–260.
Reynolds, R. W., , and Smith T. M. , 1994: Improved global sea surface temperature analyses using optimum interpolation. J. Climate, 7, 929–948.
Reynolds, R. W., , Smith T. M. , , Liu C. , , Chelton D. B. , , Casey K. S. , , and Schlax M. G. , 2007: Daily high-resolution blended analyses for sea surface temperature. J. Climate, 20, 5473–5496.
Rosmorduc, V., , and Hernandez F. , 2003: Two altimetric satellites minimum are needed for ocean observation and forecasting. Aviso Users Newsletter, No. 9, Aviso, Ramonville Saint-Agne, France, 12–14.
Sandery, P. A., , Brassington G. B. , , Craig A. , , and Pugh T. , 2010: Impacts of ocean–atmosphere coupling on tropical cyclone intensity change and ocean prediction in the Australian region. Mon. Wea. Rev., 138, 2074–2091.
Sanford, T. B., , Black P. G. , , Haustein J. R. , , Feeney J. W. , , Forristall G. Z. , , and Price J. F. , 1987: Ocean response to a hurricane. Part I: Observations. J. Phys. Oceanogr., 17, 2065–2083.
Sanford, T. B., , Price J. F. , , Girton J. B. , , and Webb D. C. , 2007: Highly resolved observations and simulations of the ocean response to a hurricane. Geophys. Res. Lett., 34, L13604, doi:10.1029/2007GL029679.
Scharroo, R., , Smith W. H. , , and Lillibridge J. L. , 2005: Satellite altimetry and the intensification of Hurricane Katrina. Eos, Trans. Amer. Geophys. Union, 86, 366–367, doi:10.1029/2005EO400004.
Servain, J., , Busalacchi A. J. , , McPhaden M. J. , , Moura A. D. , , Reverdin G. , , Vianna M. , , and Zebiak S. E. , 1998: A Pilot Research Moored Array in the Tropical Atlantic (PIRATA). Bull. Amer. Meteor. Soc., 79, 2019–2031.
Shay, L. K., 2009: Upper ocean structure: Response to strong forcing events. Encyclopedia of Ocean Sciences, 2nd ed. J. Steele et al., Elsevier Press International, 4619–4637.
Shay, L. K., , and Uhlhorn E. , 2008: Loop current response to Hurricanes Isidore and Lili. Mon. Wea. Rev., 136, 3248–3274.
Shay, L. K., , and Brewster J. K. , 2010: Eastern Pacific oceanic heat content estimation for hurricane intensity forecasting. Mon. Wea. Rev., 138, 2110–2131.
Shay, L. K., , Black P. G. , , Mariano A. J. , , Hawkins J. D. , , and Elsberry R. L. , 1992: Upper ocean response to Hurricane Gilbert. J. Geophys. Res., 97 (C12), 20 227–20 248.
Shay, L. K., , Goni G. J. , , and Black P. G. , 2000: Effects of a warm oceanic feature on Hurricane Opal. Mon. Wea. Rev., 128, 1366–1383.
Shay, L. K., and Coauthors, 2011: Airborne surveys of the loop current complex from NOAA WP-3D during the Deep Water Horizon oil spill. Monitoring and Modeling the Deep Water Horizon Oil Spill: A Record Breaking Enterprise, Geophys. Monogr., Vol. 195, Amer. Geophys. Union, 131–151.
Shay, L. K., , Brewster J. K. , , and Maturi E. , 2012: Algorithm theoretical testbed document: Satellite derived oceanic heat content product suite. Satellite Products and Services Review Board Tech. Rep., NOAA NESDIS, 32 pp.
Singer, J. J., 1990: On the error observed in electronically digitized T-7 XBT data. J. Atmos. Oceanic Technol., 7, 603–611.
Stephens, C., , Antonov J. I. , , Boyer T. P. , , Conkright M. E. , , Locarnini R. , , O’Brien T. D. , , and Garcia H. E. , 2002: Temperature. Vol. 1, World Ocean Atlas 2001, NOAA Atlas NESDIS 49, 168 pp.
Sun, D., , Gautam R. , , Cervone G. , , Boyeyi Z. , , and Kaptos M. , 2006: Comment on “Satellite altimetry and the intensification of Hurricane Katrina.” Eos, Trans. Amer. Geophys. Union,87, 89, doi:10.1029/2006EO080006.
Uhlhorn, E. W., , and Shay L. K. , 2013: Loop current mixed layer energy response to Hurricane Lili (2002). Part II: Idealized numerical simulations. J. Phys. Oceanogr., 43, 1173–1192.
Wada, A., , and Usui N. , 2007: Importance of tropical cyclone heat potential for tropical cyclone intensity and intensification in the western North Pacific. J. Oceanogr., 63, 427–447.
Wada, A., , Niino H. , , and Nakano H. , 2009: Roles of vertical turbulent mixing in the ocean response to Typhoon Rex (1998). J. Oceanogr., 65, 373–396.
Walker, N., , Leben R. R. , , and Balasubramanian S. , 2005: Hurricane-forced upwelling and chlorophyll a enhancement within cold-core cyclones in the Gulf of Mexico. Geophys. Res. Lett., 32, L18610, doi:10.1029/2005GL023716.
Wu, C.-C., , Lee C.-Y. , , and Lin I.-I. , 2007: The effect of the ocean eddy on tropical cyclone intensity. J. Atmos. Sci., 64, 3562–3578.
Zheng, Z.-W., , Ho C.-R. , , and Kuo N.-J. , 2008: Importance of pre-existing oceanic conditions to upper ocean response induced by Super Typhoon Hai-Tang. Geophys. Res. Lett., 35, L20603, doi:10.1029/2008GL035524.
