Surface Heat Fluxes over Global Oceans Exclusively from Satellite Observations

Randhir Singh Atmospheric Sciences Division, Meteorology and Oceanography Group Space Applications Centre (ISRO), Ahmedabad, India

Search for other papers by Randhir Singh in
Current site
Google Scholar
PubMed
Close
,
C. M. Kishtawal Atmospheric Sciences Division, Meteorology and Oceanography Group Space Applications Centre (ISRO), Ahmedabad, India

Search for other papers by C. M. Kishtawal in
Current site
Google Scholar
PubMed
Close
,
P. K. Pal Atmospheric Sciences Division, Meteorology and Oceanography Group Space Applications Centre (ISRO), Ahmedabad, India

Search for other papers by P. K. Pal in
Current site
Google Scholar
PubMed
Close
, and
P. C. Joshi Atmospheric Sciences Division, Meteorology and Oceanography Group Space Applications Centre (ISRO), Ahmedabad, India

Search for other papers by P. C. Joshi in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

A new approach is introduced for determining surface latent heat flux (LHF) and sensible heat flux (SHF) over the global oceans exclusively from satellite observations. Measurements of wind speed (U), sea surface temperature (SST), near surface specific humidity (Qa), and air–sea temperature difference (ΔT = SST − Ta) are required for computing these fluxes by bulk formulas. To compute the heat fluxes exclusively from satellite data, U is obtained from Special Sensor Microwave Imager (SSM/I), SST is obtained from Advanced Very High Resolution Radiometer (AVHRR), empirical algorithm proposed earlier is used to compute ΔT, and a new one is developed to estimate Qa. The developed empirical equation for Qa estimations is an extension of the authors’ previous method. Compared to the Comprehensive Ocean–Atmosphere Data Set (COADS), the Qa retrieved by the previous approach had a negative bias of the order of more than 2 g kg−1 over the Gulf Stream and Kuroshio during winter but had a positive bias of more than 2 g kg−1 over the Arabian Sea and the Bay of Bengal during summertime. The new empirical equation takes into account these seasonal biases over the Gulf Stream, Kuroshio, and the Arabian Sea. Compared to COADS observations, the Qa retrieved from the developed empirical equation has global mean root mean square error (rmse), bias, and correlation of the order of 0.55, −0.007, and 0.98 g kg−1, respectively.

Compared to COADS, the satellite-derived monthly mean LHF has global mean rmse, bias, and correlation of the order of 20, 6, and 0.97 W m−2, respectively. Likewise, satellite-derived monthly mean SHF has global mean rmse, bias, and correlations of the order of 6, 0.4, and 0.98 W m−2, respectively. The monthly fields show that the spatial patterns and seasonal variability of satellite-derived latent and sensible heat fluxes are generally good in agreement with those of the COADS and earlier satellite-derived fluxes.

Sixteen-year (January 1988–December 2003) datasets of surface heat fluxes and basic input parameters over the global oceans have been constructed using SSM/I and AVHRR data. This dataset has a spatial resolution of 1° × 1° latitude–longitude and temporal resolution of one month. This unique dataset is constructed exclusively from satellite observations, and it can be obtained from the Meteorology and Oceanography Group Space Applications Centre.

Corresponding author address: Dr. Randhir Singh, Atmospheric Sciences Division, Meteorology and Oceanography Group Space Applications Centre (ISRO), Ahmedabad-380015, India. Email: randhir_h@yahoo.com

Abstract

A new approach is introduced for determining surface latent heat flux (LHF) and sensible heat flux (SHF) over the global oceans exclusively from satellite observations. Measurements of wind speed (U), sea surface temperature (SST), near surface specific humidity (Qa), and air–sea temperature difference (ΔT = SST − Ta) are required for computing these fluxes by bulk formulas. To compute the heat fluxes exclusively from satellite data, U is obtained from Special Sensor Microwave Imager (SSM/I), SST is obtained from Advanced Very High Resolution Radiometer (AVHRR), empirical algorithm proposed earlier is used to compute ΔT, and a new one is developed to estimate Qa. The developed empirical equation for Qa estimations is an extension of the authors’ previous method. Compared to the Comprehensive Ocean–Atmosphere Data Set (COADS), the Qa retrieved by the previous approach had a negative bias of the order of more than 2 g kg−1 over the Gulf Stream and Kuroshio during winter but had a positive bias of more than 2 g kg−1 over the Arabian Sea and the Bay of Bengal during summertime. The new empirical equation takes into account these seasonal biases over the Gulf Stream, Kuroshio, and the Arabian Sea. Compared to COADS observations, the Qa retrieved from the developed empirical equation has global mean root mean square error (rmse), bias, and correlation of the order of 0.55, −0.007, and 0.98 g kg−1, respectively.

Compared to COADS, the satellite-derived monthly mean LHF has global mean rmse, bias, and correlation of the order of 20, 6, and 0.97 W m−2, respectively. Likewise, satellite-derived monthly mean SHF has global mean rmse, bias, and correlations of the order of 6, 0.4, and 0.98 W m−2, respectively. The monthly fields show that the spatial patterns and seasonal variability of satellite-derived latent and sensible heat fluxes are generally good in agreement with those of the COADS and earlier satellite-derived fluxes.

Sixteen-year (January 1988–December 2003) datasets of surface heat fluxes and basic input parameters over the global oceans have been constructed using SSM/I and AVHRR data. This dataset has a spatial resolution of 1° × 1° latitude–longitude and temporal resolution of one month. This unique dataset is constructed exclusively from satellite observations, and it can be obtained from the Meteorology and Oceanography Group Space Applications Centre.

Corresponding author address: Dr. Randhir Singh, Atmospheric Sciences Division, Meteorology and Oceanography Group Space Applications Centre (ISRO), Ahmedabad-380015, India. Email: randhir_h@yahoo.com

1. Introduction

Evaporation at the air–sea interface results in a transport of energy and water vapor into the atmosphere. The energy transport partly compensates losses of the energy through radiation processes in the atmosphere. The global mean of this energy transport is equivalent to 26% of the incoming solar energy at the top of the atmosphere. On this account, the exchange of the energy between the sea surface and the atmosphere is a major energy source for the atmospheric circulation. The exchange of water vapor and heat at the surface takes place simultaneously and connects the energy cycle to hydrological cycle. Thus, the latent heat flux causes a cooling of the upper layer of the ocean and through the loss of water, an increase of the salinity in the oceanic mixed layer. Good estimates of latent heat flux at the sea surface (together with fluxes of momentum and sensible heat) with global coverage could be very useful for verifying coupled ocean–atmosphere models, as well as for driving ocean models. They are also very useful for understanding the basic physics of air–sea interaction. Most of the studies carried out to estimate the evaporation fields on a global scale have used only historical ship reports (Bunker 1976; Hastenrath and Lamb 1977, 1979; Hsiung 1986). Since the temporal and spatial resolution of ship measurements is limited, their use in multiyear studies is limited.

On the other hand, the uniform spatial and temporal sampling of satellite measurements provides a unique way to remotely derive these heat fluxes over open oceans and complement in situ observations. The main deficiency of the satellite method however is the difficulty of the estimation of the near-surface specific humidity (Qa) and air temperature (Ta).

Liu (1986, hereafter L86) developed empirical relation between the monthly mean total precipitable water (W), which can be measured by satellite, and monthly mean near surface specific humidity (Qa). Using this empirical relation along with satellite-derived surface wind speeds (U) and sea surface temperature (SST), the determination of surface latent heat flux over remote areas in the open oceans became feasible (Liu 1988). In spite of improvements obtained by recent studies (Schulz et al. 1993; Chou et al. 1995), there still exist significant systematic and random differences between satellite and in situ estimations of surface latent heat fluxes and these arise primarily from uncertainties in the near-surface specific humidity (Qa) field (Jourdan and Gautier 1995).

After pioneer work of L86, several authors have attempted the estimation of Qa from satellite measurements. Schulz et al. (1993), Schlussel et al. (1995, hereafter SSE), Chou et al. (1995), Chou et al. (1997, hereafter CSSA), and Jones et al. (1999) all have presented statistical relationships toward that end. Many successful studies have been completed using these statistically based methods; however, all the methods suffer from either bias or significant scatter, or both. Detailed review of these methods for retrieving Qa is given by Singh et al. (2005b, hereafter S05b).

To further improve the satellite estimates of LHF, S05b developed an empirical equation to retrieve monthly mean Qa from SSM/I total precipitable water vapor (W) and AVHRR-derived SST using genetic algorithm (GA). S05b compared their results with other existing methodologies (L86; L86; CSSA) and concluded that the GA-based empirical equation yields better results in term of rmse and biases. However, it was found that Qa retrieved by S05b had a large negative bias over the Gulf Stream and Kuroshio during wintertime and positive bias over the Arabian Sea and Bay of Bengal during summer season, as compared to COADS. In this paper, to reduce these humidity biases, the methodology of S05b is extended and a new empirical equation has been developed, which takes into account these large seasonal biases present over above-mentioned places. The newly developed empirical equation has negligible biases and provides much better agreement with in situ observations than equation proposed by S05b.

On the other hand, it is not easy to estimate sensible heat flux from satellite data because it is difficult to obtain air temperature (Ta), which is necessary to estimate SHF by bulk formula, by using satellite data. Since there are no means to directly access the air temperature from satellite data, some empirical methods of varying accuracy have been developed, to compute Ta, which is then used to compute ΔT. Simple methods assume a constant ΔT (∼1°C) or assume a value of Ta that would lead to a constant or climatological value of relative humidity (Liu 1988), provided Qa is known. This might be accurate enough to compute exchange coefficients but too rough to determine the sensible heat flux (Schulz 2003). Jourdan and Gautier (1995, hereafter JG95) used radiosonde and SSM/I data to develop a polynomial fit between W and air temperature (Ta). Kubota and Shikauchi (1995) derived ten-day mean Ta from SSM/I-derived W and compared with the air temperature from two ocean data buoys of Japan Meteorological Agency (JMA). The rmse was found to be of the order of 1.0°C. Taking a different approach to the problem, Konda and Imasato (1996) used the bulk formulas to obtain a relationship among air temperature, SST, wind speed, and surface specific humidity. Validation with in situ observation revealed an error of 3.1°C on monthly averaged Ta. Jones et al. (1999) used artificial neural network (ANN) to obtain monthly averages of Ta from SSM/I measurements of W, and SST analysis from National Centers for Environmental Prediction (NCEP). The global mean rmse was stated to be 0.721° ± 0.38°C. Bourras et al. (2002) developed a new technique for the estimation of air temperature that uses a combination of satellite observations and surface analysis from operational NWP models. The reported accuracy of estimated Ta on instantaneous time scales was 0.3°C. Going one step further, recently Singh et al. (2005a, hereafter S05a) developed an empirical equation to estimate ΔT from satellite-based observation of SST, W, and U. The global mean rmse was stated to be 0.40° ± 0.11°C. The uniqueness of this method is that instead of Ta it directly estimates ΔT, which is the basic parameter required for estimations of SHF. The best algorithm provides Ta (exclusively from satellite observations) with an accuracy of ∼1°C, while the accuracy of monthly mean SST is ∼0.6°C (JG95). Because of the additive nature of errors, the accuracy of ΔT that can be achieved by using the difference of above two parameters can be as large as the value of ΔT itself. These values of ΔT lead to considerable errors in the estimation of SHF. On the other hand, the method of S05a that computes ΔT directly from satellite observation can be expected to improve the SHF estimation significantly.

The purpose of this paper is to estimate monthly mean surface latent and sensible heat fluxes exclusively from satellite observations and compare the heat fluxes with that derived from in situ observations (COADS). For this purpose, the new empirical equation for estimation of Qa from satellite-based observations is developed. This equation takes into account the region specific biases present in S05b proposed method. The U from SSM/I, SST from AVHRR, ΔT is computed using method of S05a, and Qa from new developed empirical equation is used to compute the surface heat fluxes using bulk formulas. To the best of our knowledge, this is the first study that computes the sensible heat fluxes over global oceans exclusively from satellite observation with a reasonable accuracy.

2. Data

a. In situ data

In the present study we have used monthly mean near-surface specific humidity (Qa), wind speed (U), sea surface temperature (SST), and air temperature (Ta) observations from the surface marine data (SMD; da Silva et al. 1994). The SMD are available at 1° latitude × 1° longitude resolution for the period 1988 to 1993. The SMD are based on ship reports compiled in the COADS. The SMD however has stricter quality-control criteria than in COADS and applies bias corrections for several surface observations. A large number of studies (Chou et al. 1997; Gautier et al. 1998; Jones et al. 1999; S05a) have been carried out using this data. Some grid points, particularly over higher latitudes of Southern Hemisphere, did not have any COADS observations (from which the SMD data were derived), during entire 6-yr period. Such grid points were excluded from the analysis, because SMD analysis may not be a true representative of the processes at these points.

b. Satellite data

The satellite data used in the present study include observations from National Oceanic and Atmospheric Administration (NOAA)–AVHRR, and the SSM/I onboard the Defense Meteorological Satellite Program (DMSP) satellite. AVHRR pathfinder (Walton 1988; Brown et al. 1993) data consist of daily fields of gridded SST with a spatial resolution of 54 km2 with data gaps over cloudy regions. We computed monthly averages of SST fields at 1° latitude × 1° longitude resolution for the period 1988–2003, using the daily values from pathfinder data. Vertically integrated water vapor (W) and surface wind speed (U) fields are part of the global SSM/I ocean products (Wentz 1997). Monthly values were obtained by averaging daily at 1° × 1° grids for the period 1988–2003.

To compare GA-retrieved Qa with other existing Qa products, we have used Qa product of CSSA and L86. These two products are available through the Goddard Satellite Based Surface Turbulent Fluxes version 2 (GSSTF2; Chou et al. 2003) and Hamburg Ocean–Atmosphere Parameters and fluxes from Satellite Data (HOAPS; Grabl et al. 2000).

3. Methodology of Qaestimation

a. Genetic algorithm

A genetic algorithm is programmed to approximate the equation, in symbolic form, that best describes the relationship between independent and dependent parameters. The genetic algorithm considers an initial population of potential solutions, which is subjected to an evolutionary process by selecting those equations that best fit the data. The strongest strings choose a mate for reproduction whereas the weaker strings become extinct. The newly generated population is subjected to mutations that change fractions of information. The evolutionary steps are repeated with the new generation. The process ends after a number of generations a priori determined by the user. The procedural details of genetic algorithm are given by Szpiro (1997) and Alvarez et al. (2000). A brief description of genetic algorithm is as follows. Let us assume that there exists a smooth mapping function P(.) that explains the relationship between a desired variable x and a set of independent variables [a, b, c, d, e, . . .], so that
i1520-0493-134-3-965-e1
First, for an amplitude function x, a set of candidate equations for P(.) is randomly generated. An equation is stored as a set of characters that define the independent variables, a, b, c, d, e . . . etc. in Eq. (1), and four arithmetic operators (+, −, ×, and /). A criterion that measures how well the equation strings perform on a training set of the data is its fitness to the data, defined as sum of the squared differences between data and the parameter derived from the equation string. The strongest individuals (equations with best fits) are then selected to exchange parts of the character strings between them (reproduction and crossover) while individuals less fitted to the data are discarded. Finally, a small percentage of the equation strings’ most basic elements, single operators and variables, are mutated at random. The process is repeated a large number of times to improve the fitness of the evolving population of equations. The fitness strength of the best scoring equation is defined as
i1520-0493-134-3-965-e2
where Δ2 = Σ (xc − xo)2, xc is parameter value estimated by the best scoring equation, xo is the corresponding true value, 〈xo〉 is the mean of the true values of x. Szpiro (1997) has shown the robustness of genetic algorithm to forecast the behavior of one-dimensional chaotic dynamical system. Later, Alvarez et al. (2000) used GA to predict space–time variability of the SST in the Alboran Sea. Kishtawal et al. (2003) used GA for the prediction of seasonal monsoon rainfall over Indian region. Recently S05b and S05a used GA for finding optimum relationship between satellite observations and air–sea interaction parameters, like near-surface specific humidity (Qa), and air–sea temperature difference (ΔT = SST − Ta). Kishtawal et al. (2005) used GA to develop an automatic technique for intensity estimation of the tropical cyclone over global basins.

b. GA training

The genetic algorithm is used to develop the new empirical equation for estimation of Qa. The development of the retrieval algorithm involves a number of steps. First is the selection of satellite-observed parameters that can be used for the estimation of Qa in most optimum manner. Besides SST, W, several other parameters like wind speed (U), outgoing longwave radiation (OLR), cloud liquid water, and boundary layer water vapor were considered in a large set of possible physical parameters. However, the final and most optimized GA solution retained only SST and W as the best parameters for the estimation of Qa. Jones et al. (1999) also used SST and W for the estimation of Qa. The next step is to divide the dataset into training and validation samples. All the available data (1988–93) sets of monthly mean W from SSM/I and SST from AVHRR, and Qa from SMD were divided into two subsamples. The portioning of the data into training and validation was done in the same manner as Jones et al. (1999) who chose two years belonging to contrasting climatic regimes (El Niño and La Niña) for validation of their algorithm. The 47-month period including January 1988–June 1988, July 1989–June 1992, and July 1993–November 1993 was designated as sample I and is used to train the GA. The 24-month period of July 1988–June 1989 and July 1992–June 1993 was taken as sample II and was used for evaluation of the algorithm.

Because the number of surface marine observations varies significantly in space and time, it is necessary to ensure that high quality observations of Qa are used to find the empirical equation. Adequate estimation of monthly mean Qa requires more than 20 observations (N) month−1 (e.g., Luther and Harrison 1984). We have taken Qa observations from those bins in the analysis domain in which more than 20 samples were used to produce monthly averages of Qa in SMD. Based on the data quality control described above, for training, we randomly selected 20 000 points out of ∼300 000 points of sample I data. A GA has the capability to learn the process from relatively smaller datasets compared to other optimization technique (Szpiro 1997).

The first phase (phase I) of the GA training provides empirical equation relating Qa to W and SST. The equation is as follows:
i1520-0493-134-3-965-e3
where SST and W represent the sea surface temperature (°C), and vertically integrated water vapor (gm cm−2), respectively. The values of constants (a, b, c, d, e, f ) are given in Table 1. The Eq. (3) is used to get monthly mean Qa using all the data points from sample II. The assessment of GA retrieved with respect to SMD is made by computing the bias (SMD − GA retrieved) and rmse of monthly averages Qa for sample II. The GA equation from phase I yielded a global mean rmse of 0.80 g kg−1 for sample II data. Interestingly, at ∼75% of the points, the differences between observed and GA-retrieved Qa were within ±0.8 g kg−1. The higher values of negatives differences were observed over western boundary currents, like the Gulf Stream and Kuroshio. These regions are characterized by a continental cold air outbreak and seasonal currents system, particularly during wintertime. Higher positive differences are also found over the Arabian Sea and the Bay of Bengal during summertime. Isolated pockets of higher differences are also observed over the tropical Pacific Ocean and the Southern Hemisphere (south of 30°S). The first phase of the Qa retrieval is also given in the paper by S05b. The Qa retrieved by Eq. (3) is compared with the observed Qa to produce mean bias (SMD–GA retrieved) for each month from sample I data. Bias (B) at a grid point is defined as Bj = where j is the index for month and the overbar represents the time averaging (separate averages for 12 months) over sample I period. It is to be noted that each month has a different bias that is based on the average of four years from the sample I period. Figure 1 shows the variation of Bj over the north Arabian Sea, Kuroshio, and Gulf Stream for the entire length of the available data (1988–93). Figure 1 shows that during different years, the bias does not deviate significantly from the mean annual patterns. This aspect allows us to treat Bj as independent parameters for possible modification of our earlier algorithm. Following the procedure adopted by Jones et al. (1999) we also took into account the mean bias as the third independent parameter in addition to SST, and W, in second phase (phase II) of GA training in order to take into account the systematic regional biases. The procedure of random data selection and training in phase II was similar to those in phase I except that three (W, SST, and B) parameters are now used to train the GA.

c. Final algorithm and validation

The genetic algorithm–based empirical equation developed in the phase II has the following form:
i1520-0493-134-3-965-e4
where SST, W, and B represent the sea surface temperature (°C), vertically integrated water vapor (gm cm−2), and mean bias (g kg−1), respectively. The values of constants (a1, b1, c1, d1, e1) are given in Table 2. To validate the phase II GA-retrieved Qa, we applied Eq. (4) on satellite-derived SST, W, and mean bias (B) on testing data (sample II) and compared against SMD as well as with phase I GA-retrieved Qa. In addition, we compared our estimates with other empirical (L86; L86; CSSA) estimates of Qa so that improvements upon previous methodologies can be assessed. Figure 2 shows the bias between SMD and GA-retrieved Qa for (a) phase I and (b) phase II. Over most of the places the bias between SMD and phase I GA-retrieved Qa (Fig. 2a) are within ±0.6 g kg−1. The higher negative differences (>2 g kg−1) are observed over western boundary currents, like the Gulf Stream and Kuroshio. These regions are characterized by continental cold air outbreaks and seasonal currents system, particularly during wintertime. Under these conditions, the weak coupling between SST and Qa may be one of the reasons for poor retrieval accuracy by phase I GA. In a recent study, Dong and Kelly (2004) also attributed local advection such as over the Gulf Stream region for driving the fluxes. Under these local effects, SST may be poor indicator (Dong and Kelly 2004) of fluxes. Over the Arabian Sea and Bay of Bengal, also, the positive differences between SMD and phase I GA-retrieved Qa are large (>2 g kg−1). The isolated pockets of higher differences are also observed over Southern Hemisphere (south of 30°S) and over tropical Pacific Ocean, which probably indicate the effect of very low number of observations in the monthly averages of SMD data. Over most of the places the biases between SMD and phase II GA (Fig. 2b) are within ±0.2 g kg−1. As expected, substantial improvements over the regions of the Gulf Stream, Kuroshio, and Arabian Sea have resulted from inclusion of systematic bias in phase II GA training. Figure 3 shows the mean rmse between SMD and GA-retrieved Qa for (a) phase 1 GA, and (b) phase II GA for sample II data. The rmse between SMD and Qa from phase II GA (Fig. 3b) show significant improvement (particularly over high sampled regions, e.g., the North Pacific, Atlantic, Arabian Sea, and Bay of Bengal) relative to rmse between SMD and Qa from phase I GA (Fig. 3a). The global mean rmse for Qa from phase II GA using Eq. (4) is 0.55 g kg−1, which is smaller than the natural variance (0.80 g kg−1) of observed Qa. The rmse over some isolated pockets (south Indian and southeast Pacific Oceans) are slightly higher (1.2 to 1.5 g kg−1). It is very difficult to judge the accuracy of new methodology over these places because SMD is undersampled over these regions. To assess the improvement due to new algorithm [Eq. (4)] over the other schemes, we have computed an improvement parameter η defined as η = (rmse between other algorithm and SMD) − (rmse between SMD and phase II GA). The distribution of η is given in Fig. 4. The positive values of η indicate real improvement. Over all the oceanic regions, except some isolated pockets, the value of η is positive, which indicate the strength of GA method after including region-specific biases in phase II. With respect to L86 (Fig. 4a), phase II GA shows large improvement (more than 0.5 g kg−1) over the Southern Hemisphere, equatorial region, western Pacific Ocean, and Arabian Sea. The phase II GA shows very large improvement (3 to 4 g kg−1), when compared to CSSA (Fig. 4b) and L86 (Fig. 4c).

4. Temporal variability of GA-retrieved Qa

A good way to validate the performance of an algorithm is a comparison of retrieved and observed time series in different climatic domains. Figure 5 shows the annual variation (averaged over sample II) of Qa averaged over the north Arabian Sea (20°–23°N, 58°–61°E), Kuroshio (28°–31°N, 127°–130°E), and Gulf Stream (34°–37°N, 60°–63°W). Annual variation of Qa derived by other methods (L86; CSSA; L86) is also presented in Fig. 5 for comparison. To see the difference between phases I and II, Qa retrieved from GA-I as well as GA-II is shown. These regions are chosen because they represent diverse climatic domains and also GA retrieval shows large improvement after inclusion of the seasonal biases over these region. Good agreement over the whole period is observed between phase II of GA- (GA-II) retrieved and SMD-observed Qa over all the regions. Over the north Arabian Sea, the GA-II time series also compares well with SMD, whereas all other algorithms (GA-I; L86; CSSA; L86) strongly underestimates Qa throughout the year, except in winter when GA-I, L86, and CSSA slightly over estimates the Qa. From Fig. 5 it is clear that GA-II shows large improvements over the north Arabian Sea, Kuroshio, and the Gulf Stream when compared with GA-I retrieval. To see the performance of the GA methodology over areas like subtropical highs that are characterized by large-scale downward motion. The northern subtropics (25°–35°N, 0°–360°E) and southern subtropics (25°–35°S, 0°–360°E) are also used for comparison (Table 3) among different Qa products. In summary (Table 3), GA-II provides superior accuracy of Qa retrieval over the Arabian Sea and subtropics, while over other regions selected for comparison, the performance of GA-II is better than L86 and L86, and it is comparable to CSSA.

5. Retrieval of heat fluxes

Usually, the bulk formulas are used to estimate LHF and SHF from satellite measurements (Liu 1988; Jourdan and Gautier 1995; Chou et al. 1995; Zhang 1995; Schulz et al. 1997; Chou et al. 1997; Chou et al. 2003; Bentamy et al. 2003; Singh 2004; S05b). The bulk formulas parameterize these fluxes mainly as a function of surface wind speed, near surface specific humidity, air temperature, and sea surface temperature. Hence the heat fluxes are given by
i1520-0493-134-3-965-e5
where ρ is the air density, Cp is the specific heat at constant pressure, U is the surface wind speed, L is the latent heat of evaporation, CE and CH are bulk coefficients for latent and sensible heat fluxes, Qs is the saturation specific humidity at the surface, and Qa is the near-surface specific humidity (at atmospheric measurement level). SST is sea surface temperature, and Ta is air temperature. The saturation specific humidity, Qs, is calculated from SST assuming saturation at the surface. The LHF depends on the skin SST. However, the AVHRR SST is derived from empirical relation developed with regressing the AVHRR radiance with the bulk SST (Walton 1988; Wick et al. 1992; Chou et al. 2003). To compensate for the cool skin effects on LHF, Qs is estimated using the approximated (Chou et al. 2003) formula. Further Qs is reduced by 2% to account for salinity effects (Bentamy et al. 2003). Based on the behavior of CE as a function of wind speed and air–sea temperature difference and assuming slightly unstable stratification (Ta = SST − 1.25 K) over the global oceans, along with the Hasse and Smith (1997) results, the CE is computed using the empirical relation given by Bentamy et al. (2003). The possible error due to the assumption of unstable stratification is generally less than 2.5% (Bentamy et al. 2003). The bulk coefficient for sensible heat fluxes is same as that of latent heat fluxes (Zhang 1995). When using the monthly mean meteorological variables to compute the surface heat fluxes, one question arises: how close are these estimates to the actual fluxes? Synoptic-scale weather systems have time scales of a week or so, and organized convective systems have time scales from a few hours to a few days. These systems can produce significant fluctuations compared to monthly mean, especially over the warm pool where mean surface winds are weak. Are the effects of these systems on the surface heat fluxes properly accounted for when monthly mean data are used in flux computation using Eq. (5)? These concerns are frequently raised by critics who are skeptical of using monthly mean data to estimate surface heat fluxes. Esbensen and Reynolds (1981), Liu (1988), Chou et al. (1995), and Zhang (1995) compared the monthly averaged heat fluxes using both daily and monthly mean meteorological variables as input to bulk formulas. Let us denote these two fluxes by symbols F and , respectively. After a complete error analysis and a significance test of error, they concluded that two fluxes (F and ) computed from daily and monthly mean data do not differ significantly over the global oceans (Chou et al. 1995) and the monthly averaged wind speeds, temperature, and humidity could be used to estimate the monthly averaged latent and sensible heat fluxes. Further, to verify the above statement we have also computed (not shown) F and for January and July 2000, using bulk formulas. The flux differences (F) are generally within ±10 W m−2 over the global oceans. The negative differences of the order of 30 to 35 W m−2 are found over the Gulf Stream and Kuroshio during January. During July, the negative differences of the order of 20–30 W m−2 are found over the warm pool region. The geographical distribution of F is similar to that of Chou et al. (1995). Hence we used monthly mean basic parameters to compute the monthly mean heat fluxes.

6. SMD and satellite heat flux intercomparison

a. Latent heat flux

In this section, we compare the satellite-derived monthly mean latent heat flux with observed flux, computed using SMD, through bulk formulas [Eq. (5)]. Satellite latent heat flux (LHFGA-II) is computed from monthly mean values of SSM/I surface winds, AVHRR-derived SST, and GA phase II–retrieved Qa.

To assess the improvement resulting from phase II GA-retrieved Qa on satellite derived latent heat flux, we calculated the latent heat flux from satellite observations again by replacing GA-II-retrieved Qa with other (L86; CSSA; L86) satellite estimates of Qa in Eq. (5). The rest of the parameters in Eq. (5) are same as they are in case of LHFGA-II computation. We used Qa of L86 in Eq. (5) to compute the LHF. This is termed as LHFL86. The LHF computed using L86 and CSSA; Qa are termed as LHFSSE and LHFCSSA, respectively.

To evaluate accuracy of each satellite derived latent heat flux (LHFGA-II, LHFSSE, LHFCSSA, LHFL86) dataset, we compare the annual mean (averaged over sample II) SMD latent flux (LHFSMD) with annual mean of each satellite-derived LHF. Analysis indicates (not shown) that LHFGA-II agrees well with those of LHFSMD, with bias of 9 W m2, rmse of 12 W m−2, and a correlation coefficient of 0.97. The LHFSSE has bias, rmse, and correlation of the order of 5, 22, and 0.87 W m−2, respectively. The LHFCSSA overestimates with a bias, rmse, and correlation of the order of −14, 25, and 0.87 W m−2, respectively. The LHFL86 slightly overestimates with bias, rmse, and correlation of the order of −1, 22, and 0.88 W m−2, respectively. So it is quite evident that when phase II GA-retrieved Qa is used in computation of satellite-derived LHF the rmse (correlation) are less (high), compared to other satellite LHF estimates. The zonal averages of the monthly mean rmse and bias for the sample II data (Figs. 6a,b) show that the LHFGA-II is closer to those of LHFSMD. For LHFGA-II the rmse are below 25 W m−2 over most of the latitudes, except higher rmse between 30° and 40°S latitudes, which may be due to undersampling in SMD over these regions. Moreover observations from these regions were not included in the training of GA for Qa retrieval. The global mean rmse for monthly averaged LHFGA-II, LHFSSE, LHFCSSA, and LHFL86, are of the order 20, 33, 34, and 41 W m−2, respectively. Likewise, the global mean biases for monthly averaged LHFGA-II, LHFSSE, LHFCSSA, and LHFL86, are of the order of 9, 10, −7, and 5 W m−2, respectively. Figures 7 –9 show annual cycle (from sample II period) of the satellite-derived as well as observed latent heat fluxes and related parameters over the north Arabian Sea, Kuroshio, and Gulf Stream. The latent heat flux (LHFGA-I) computed using Qa of phase I GA is also presented in these figures. Over the north Arabian Sea, LHFGA-II shows a consistent positive bias. This may be attributed to a positive bias of SSM/I winds over this region that becomes pronounced during summer months (Fig. 7a). In case of other satellite-derived products the large positive bias in the Qa (Fig. 5a), resulted in overestimation of LHF. The positive biases in the Qa (GA-I; L86; CSSA; L86) and SSM/I-derived wind speed offset each other and, as a result, the complete error of Qa does not reflected in LHFGA-I, LHFSSE, LHFCSSA, and LHFL86. The large improvement has been resulted in the LHFGA-II as compared to LHFGA-I. Over Kuroshio and Gulf Stream, during summertime, the LHFL86 shows large positive bias (∼90 W m−2), which is due to the overestimation of Qa by L86 (Figs. 5b,c). LHFGA-I also shows large positive bias (∼100 W m−2) during wintertime over both the Kuroshio and Gulf Stream, as compared to LHFSMD. This underestimation of LHFGA-I is due to the overestimation of the Qa by phase I GA (Figs. 5b,c). In summary (Table 4), over the north Arabian Sea, Kuroshio, and Gulf Stream LHFGA-II shows significant improvement as compared to LHFGA-I. Over the Arabian Sea, except LHFGA-II, all other satellite estimates shows large errors. Over the Kuroshio and Gulf Stream, LHFSSE and LHFCSSA also compares well but not better than LHFGA-II. The LHFGA-II still shows some positive biases over the Arabian Sea and Gulf Stream and a small negative bias over Kuroshio. The positive biases in LHFGA-II over the Arabian Sea and Gulf Stream are due to the positive biases in SSM/I wind speed while negative biases over Kuroshio is due to the negative bias in AVHRR-derived Qs.

The above comparisons show that GA-II-retrieved Qa shows improvement in the satellite-derived LHF. As mentioned above, significant differences between SMD and GA phase II–based fluxes emanate from the regional biases in SSM/I wind fields. To see the real improvement in satellite-derived LHF due to GA-II-retrieved Qa only, we constructed the following alternate LHF dataset using SMD-observed SST and wind speed, along with satellite-derived Qa products. The alternate dataset is made in the following way: satellite-derived Qa products (GA-II; L86; CSSA; L86) along with SMD wind speed and sea surface temperature is used in computation of flux. For example, the ALT-1 (Table 5) is obtained by using GA-II-derived Qa, along with SMD-derived U and SST. Then we compared SMD latent heat flux with each alternate dataset. These comparisons will show the impact of only satellite-retrieved Qa on the latent heat flux. Table 6 shows the statistics of comparison of each alternate data and SMD. When we use GA-II-retrieved Qa and all other parameters from SMD the global mean rmse, bias, and correlation for annually averaged latent heat flux (ALT-1) are of the order of 2.5, 0.28, and 0.99 W m−2, respectively. While other alternate (ALT-2, ALT-3, ALT-4) data show large rmse and biases. We have also plotted the time series (Fig. 10) of these alternate datasets over three regions mentioned above. The annual cycle of the ALT-1 is now closely following the LHFSMD over all the regions and throughout the year. Large disagreement is observed between other alternate data and LHFSMD, particularly over north Arabian Sea. The statistics of the comparison of monthly mean alternate fluxes over three regions is given in Table 7.

The global mean rmse for monthly averaged LHF (ALT-1, ALT-2, ALT-3, ALT-4) are of the order 10, 37, 30, and 44 W m−2, respectively. Likewise, the global mean biases for monthly mean averaged LHF (ALT-1, ALT-2, ALT-3, ALT-4) are of the order of 0.07, −22, −3.4, and −10 W m−2, respectively. It is very clear that improvement in LHF only due to improved GA-II-retrieved Qa is very large. This shows that if satellite retrieval of wind speed is improved, then GA-retrieved Qa along with satellite-retrieved wind speed can be used to compute more accurate fluxes.

b. Sensible heat flux

In this section, we compare the satellite-derived monthly and annual mean sensible heat flux with observed flux, computed using SMD, through bulk formulas [Eq. (5)]. The SSM/I derived wind speed (U) along with air–sea temperature difference (ΔT = SST − Ta) computed using method of S05a is used to compute sensible heat flux from satellite observations (SHFSAT). While SMD observed U, SST, and Ta are used to obtain the SMD sensible heat flux (SHFSMD). To the best of our knowledge there is no study of global sensible heat flux computation exclusively from satellite observations. This is because of difficulty in air temperature–retrieval from satellite. Konda and Imasato (1996) used a physical model to compute air temperature from satellite-based observation. The accuracy of retrieved air temperature was so poor (∼3°C) that it could not be used to compute sensible heat fluxes. Though they mentioned that retrieved air temperature could be used to study the interannual variability in sensible heat flux. Chou et al. (1997, 2003) computed sensible heat fluxes from SSM/I-derived wind speed and model-analyzed air–sea (ΔT = SST − Ta) temperature difference. They showed that over most of regions the sensible heat flux estimated by combing satellite and model analyzed data are within ±10 W m−2 of COADS-observed sensible heat fluxes except over northwestern Pacific and Atlantic Oceans, where the differences are slightly higher.

In this paper, the S05a empirical relation is used to compute the sea–air temperature difference from satellite-observed W, SST, U, and seasonal mean biases. We have compared the SMD sensible heat flux and satellite-derived sensible heat fluxes over sample II data. The rmse, bias, and correlation for annual mean SHF (not shown) are 1.5, −0.23, and 0.98 W m−2, respectively. Figure 11 shows the zonal average biases and rmses between monthly mean SHFSMD and SHFSAT during sample II data. Very low rms and bias are seen over tropical oceans, while slightly higher rmse and biases are seen over the higher latitudes. The global mean rmse, bias, and correlation for the monthly mean SHF are of the order of 6, 0.4, and 0.99 W m−2, respectively. Overall the accuracy achieved by satellite-derived SHFSAT is excellent. This shows the potential of S05a methodology for estimation of sea–air temperature difference and hence sensible heat flux. Furthermore, to see the temporal variability produced by satellite-derived SHF, we have shown the annual cycle of satellite derived as well SMD-observed SHF (Fig. 12), over three regions mentioned in the previous section. The annual cycle shown by satellite estimates compares very well with the SMD observations over all the regions. The upwelling during July over the Arabian Sea is also very well picked up by satellite-measured SHF.

7. Large-scale features in satellite-derived heat fluxes

a. Seasonal means

Figures 13 and 14 compare the geographical distribution of monthly mean latent heat fluxes between SMD (LHFSMD) and satellite (LHFGA-II) for January (Fig. 13) and July (Fig. 14) for sample II data. Figures 13a, 14a are obtained with SMD data, Figs. 13b, 14b are obtained from satellite data (LHFGA-II), and Figs. 13c, 14c are LHFSMD − LHFGA-II. The maximum latent heat fluxes (>160 W m−2) are generally found in the trade zones of both hemispheres (with large fluxes in the winter), due to a large Qs− Qa coupled with stronger wind. The higher fluxes (>200 W m−2) are also found over northwestern Pacific and Atlantic Oceans during January, where strong offshore winds carry cold, dry continental air over the warm Kuroshio Current and Gulf Stream. The higher fluxes (about 140–160 W m−2) are found over the Arabian Sea and Bay of Bengal during January because of higher humidity gradient. The flux decreases poleward because of the decrease in sea–air humidity difference. The qualitative agreements between the two estimates are excellent and the location of characteristics features is well produced by satellite-derived latent heat fluxes. Over most of the places the differences between these two estimates are within ±20 W m−2, except some individual pocket of higher difference (20 to 30 W m−2). These pockets are over Southern Hemisphere (south of 30°S) and the Gulf Stream during January, while the differences are more (∼20 to 30 W m−2) over Arabian Sea during July. As we have seen in Fig. 7, the differences over the Arabian Sea are due to positive bias in SSM/I wind speed.

Figures 15 and 16 compare the geographical distribution of monthly mean sensible heat fluxes between SMD (SHFSMD) and satellite (SHFSAT) for January (Fig. 15) and July (Fig. 16) for sample II data. The patterns of the sensible heat fluxes essentially follow that of sea–air temperature differences. The sensible heat flux is very small over tropical oceans and summertime extratropical oceans. It is less than 10 W m−2 over much of the tropical oceans and summertime extratropical oceans. The maximum sensible heat flux (>45 W m−2) occurs over the northwestern Pacific and Atlantic Oceans during the winter (Fig. 15). This is because strong offshore winds couple with large sea–air temperature differences, resulting from the cold continental air flowing over the warm Kuroshio Current and Gulf Stream. The Arabian Sea undergoes dramatic change after onset of summer monsoon. The negative values (Fig. 16) are found just after the onset of summer monsoon (during July). This is because of cooling of sea surface due to high surface wind, precipitation, and cloud cover.

Over large parts of the oceans, the differences between satellite-estimated and SMD-observed SHF are within ±3 W m−2. Exceptions are western boundary current and higher latitude in the Southern Hemisphere where differences are slightly large. Over western boundary currents during January, the satellite-derived SHF has positive bias of the order of 10 to 12 W m−2, which is due to the underestimation of wind speed by SSM/I. We have analyzed (not shown, partially shown in Figs. 7 to 9) the difference between SMD and SSM/I wind speed over global oceans. We found that SSM/I wind speeds have positive bias (>1 m s−1) over the Arabian Sea, North Atlantic and Pacific Oceans, while it has negative bias of the order of ∼1 m s−1 over the higher latitude of Southern Hemisphere. Overall the satellite-derived SHF shows excellent comparisons with SMD-based estimates.

b. Annual means

Figure 17 displays the geographical distribution of the satellite-derived annual mean of latent (Fig. 17a) and sensible (Fig. 17b) heat fluxes averaged over 1988–2003. We can see a large latent heat loss from the ocean, more than 150 W m−2 in the subtropical regions, while latent heat loss is small in the eastern equatorial Pacific and Atlantic Oceans, due to upwelling-induced cold SST associated with weak wind speed, and in high latitudes due to poleward decrease of QsQa. The large latent heat loss is also found over western boundary current regions, such as, the Kuroshio and Gulf Stream (>150 W m−2). Sensible heat flux is generally quite small and negligible (5 to 10 W m−2) compared with latent heat flux; however, sensible heat flux over western boundary current and at high latitudes is not negligible. The geographical distribution of the GA-based annual mean LHF and SHF averaged over 1988–2003 is similar to those of GSSTF2, HOAPS, the Japanese Ocean Flux Data Sets with Use of Remote Sensing Observations (J-OFURO), and NCEP (Kubota et al. 2002; Chou et al. 2003). However, there are quantitative differences among these global flux datasets.

8. Conclusions

The global monthly mean latent and sensible heat fluxes are derived exclusively from satellite observations. For this purpose new empirical equation for estimation of Qa from satellite-based observations is developed. This new equation takes into account the region specific biases present in the S05b proposed method. The evaluation of the method shows good agreement with the surface marine observation not used in the development of algorithm. The rmse in monthly mean Qa is 0.55 g kg−1, which is smaller than the natural variance in the SMD-observed Qa. The developed empirical equation gives better estimates of Qa over all the oceans when compared with L86, CSSA, and L86. Over some of the oceans, particularly over the Arabian Sea, the improvement is very large.

The wind speed from SSM/I, sea surface temperature from AVHRR, sea–air temperature difference using method of S05a and Qa from the new empirical equation is used to compute the surface heat fluxes using bulk formulas. To the best of our knowledge, this is the first study that computes the sensible heat fluxes over global oceans, exclusively from satellite observations. The heat fluxes determined from satellite data are temporally and spatially coherent. The rms difference between satellite- and SMD-derived latent heat fluxes shows that the two computations differ by about 20 W m−2 when phase II GA-retrieved Qa is used in satellite computations. Biases and rmse of LHFGA-II are significantly lower compared to latent heat fluxes computed from other methods (e.g., L86; CSSA; L86). The rms difference between satellite and SMD-derived sensible heat fluxes shows that the two computations differ by about 6 W m−2. This is the first time that sensible heat flux is exclusively estimated from satellite observation with reasonable accuracy.

Using this methodology, the 16 yr of SSM/I and AVHHR data are used to generate the satellite-based heat flux climatology. The climatological as well as monthly fields are freely available to the interested user on request. We feel that these monthly heat fluxes and climatology are the best among all the other satellite-derived fluxes and climatologies because other satellite-derived flux climatologies used Qa estimated from L86, CSSA, or L86, which shows large disagreement compared with SMD observations.

Acknowledgments

The ISRO-Geosphere Biosphere Programme supported this study. The authors would like to acknowledge the Physical Oceanography Data Archival Centre (PO.DAAC) at the Jet Propulsion Laboratory (JPL), the SMD active archive at http://ingrid.ldeo.columbia.edu, and the SSM/I ocean products at ftp.ssmi.com. The HOAPS data were obtained from http://www.hoaps.zmaw.de/. The GSSTF2 data were obtained from ftp://lake.nascom.nasa.gov/data/TRMM/Ancillary/gsstf/gsstf2.01/.

REFERENCES

  • Alvarez, A., C. Lopez, M. Riera, E. Hernandez-Garcia, and J. Tintore, 2000: Forecasting the SST space-time variability of the Alboran Sea with genetic algorithms. Geophys. Res. Lett., 27 , 27092712.

    • Search Google Scholar
    • Export Citation
  • Bentamy, A., K. B. Katsaros, A. M. Mestas-Nuñez, W. M. Drennan, B. E. Forde, and H. Roquet, 2003: Satellite estimates of wind speed and latent heat flux over the global oceans. J. Climate, 16 , 637655.

    • Search Google Scholar
    • Export Citation
  • Bourras, D., L. Eymard, W. T. Liu, and H. Dupuis, 2002: An integrated approach to estimate instantaneous near-surface air temperature and sensible heat flux fields during the SEMAPHORE experiment. J. Appl. Meteor., 41 , 241252.

    • Search Google Scholar
    • Export Citation
  • Brown, J. W., O. B. Brown, and R. H. Evans, 1993: Calibration of AVHRR infrared channels: A new approach to non-linear corrections. J. Geophys. Res., 98 , 1825718268.

    • Search Google Scholar
    • Export Citation
  • Bunker, A. F., 1976: Computation of surface energy flux and annual air–sea interactions cycle of the North Atlantic Ocean. Mon. Wea. Rev., 104 , 11221140.

    • Search Google Scholar
    • Export Citation
  • Chou, S. H., R. M. Atlas, C. L. Shie, and J. Ardizzone, 1995: Estimates of surface humidity and latent heat fluxes over oceans from SSM/I data. Mon. Wea. Rev., 123 , 24052425.

    • Search Google Scholar
    • Export Citation
  • Chou, S. H., C. L. Shie, R. M. Atlas, and J. Ardizzone, 1997: Air–sea fluxes retrieved from Special Sensor Microwave Imager data. J. Geophys. Res., 102 , 1270512726.

    • Search Google Scholar
    • Export Citation
  • Chou, S. H., E. Nelkin, J. Ardizzone, R. M. Atlas, J. Ardizzone, and C. L. Shie, 2003: Surface turbulent heat and momentum fluxes over global oceans based on the Goddard Satellite Retrievals, Version 2 (GSSTF2). J. Climate, 16 , 32563273.

    • Search Google Scholar
    • Export Citation
  • da Silva, A. M., C. C. Young, and S. Levitus, 1994: Algorithms and Procedures. Vol. 1, Atlas of Surface Marine Data 1994, NOAA Atlas NESDIS 6, 83 pp.

  • Dong, S., and K. A. Kelly, 2004: Heat budget in the Gulf Stream region: The importance of heat storage and advection. J. Phys. Oceanogr., 34 , 12141231.

    • Search Google Scholar
    • Export Citation
  • Esbensen, S. K., and R. W. Reynolds, 1981: Estimating monthly averaged air–sea transfers of heat and momentum using the bulk aerodynamic method. J. Phys. Oceanogr., 11 , 457465.

    • Search Google Scholar
    • Export Citation
  • Gautier, C., P. Peterson, and C. Jones, 1998: Ocean surface air temperature derived from multiple data sets and artificial neural network. Geophys. Res. Lett., 25 , 42174220.

    • Search Google Scholar
    • Export Citation
  • Grabl, H., V. Jost, R. Kumar, J. Schulz, P. Bauer, and P. Schlussel, 2000: A climatological atlas of satellite-derived air–sea interaction parameters over the world ocean. Rep. 312, Max Planck Institute for Meteorology, Hamburg, Germany, 120 pp.

  • Hasse, L., and S. D. Smith, 1997: Local sea surface wind, wind stress, and sensible and latent heat fluxes. J. Climate, 10 , 27112724.

    • Search Google Scholar
    • Export Citation
  • Hastenrath, S., and P. J. Lamb, 1977: Climate Atlas of the Tropical Atlantic and Eastern Pacific Oceans. University of Wisconsin Press, 340 pp.

    • Search Google Scholar
    • Export Citation
  • Hastenrath, S., and P. J. Lamb, 1979: Heat Budget Atlas of the Tropical Atlantic and Eastern Pacific Oceans. University of Wisconsin Press, 104 pp.

    • Search Google Scholar
    • Export Citation
  • Hsiung, J., 1986: Mean surface energy fluxes over the global ocean. J. Geophys. Res., 91 , 1058510606.

  • Jones, C., P. Peterson, and C. Gautier, 1999: A new method for deriving ocean surface specific humidity and air temperature: An artificial neural network approach. J. Appl. Meteor., 38 , 12291245.

    • Search Google Scholar
    • Export Citation
  • Jourdan, D., and C. Gautier, 1995: Comparison between global latent heat flux computed from multisensor (SSM/I, AVHRR) and from in situ data. J. Atmos. Oceanic Technol., 12 , 4672.

    • Search Google Scholar
    • Export Citation
  • Kishtawal, C. M., S. Basu, F. Patadia, and P. K. Thapliyal, 2003: Forecasting summer rainfall over India using genetic algorithm. Geophys. Res. Lett., 30 , 15.

    • Search Google Scholar
    • Export Citation
  • Kishtawal, C. M., F. Patadia, R. Singh, S. Basu, M. S. Narayanan, and P. C. Joshi, 2005: Automatic estimation of tropical cyclone intensity using multi-channel TMI data: A genetic algorithm approach. Geophys. Res. Lett., 32 .L11804, doi:10.1029/2004GL022045.

    • Search Google Scholar
    • Export Citation
  • Konda, M., and N. Imasato, 1996: A new method to determine near sea surface air temperature by using satellite data. J. Geophys. Res., 101 , 1434914360.

    • Search Google Scholar
    • Export Citation
  • Kubota, M., and A. Shikauchi, 1995: Air temperature at ocean surface derived from surface level humidity. J. Oceanogr., 51 , 619634.

  • Kubota, M., N. Iwasaka, S. Kizu, M. Konda, and K. Kutsuwada, 2002: Japanese Ocean Flux Data Sets with Use of Remote Sensing Observations (J-OFURO). J. Oceanogr., 58 , 213225.

    • Search Google Scholar
    • Export Citation
  • Liu, W. T., 1986: Statistical relation between monthly mean precipitable water and surface-level humidity over global oceans. Mon. Wea. Rev., 114 , 15911602.

    • Search Google Scholar
    • Export Citation
  • Liu, W. T., 1988: Moisture and latent heat flux variabilities in the tropical Pacific derived from satellite data. J. Geophys. Res., 93 , 67496760.

    • Search Google Scholar
    • Export Citation
  • Luther, D. S., and D. E. Harrison, 1984: Observing long-period fluctuations of surface winds in the tropical Pacific: Initial results from island data. Mon. Wea. Rev., 112 , 285302.

    • Search Google Scholar
    • Export Citation
  • Schlussel, P., L. Schanz, and G. English, 1995: Retrieval of latent heat flux and longwave irradiance at sea surface differences at the sea surface from SSM/I and AVHRR measurements. Adv. Space Res., 16 , 107116.

    • Search Google Scholar
    • Export Citation
  • Schulz, J., 2003: Satellite remote sensing of turbulent heat fluxes. Mausam, 54 , 1. 7588.

  • Schulz, J., P. Schlussel, and H. Grassl, 1993: Water vapour in atmospheric boundary layer over oceans from SSM/I measurements. Int. J. Remote Sens., 14 , 27732789.

    • Search Google Scholar
    • Export Citation
  • Schulz, J., M. Jeans, E. Stefan, and P. Schlussel, 1997: Evaluation of satellite-derived latent heat fluxes. J. Climate, 10 , 27822795.

    • Search Google Scholar
    • Export Citation
  • Singh, R., 2004: Satellite observation of boundary layer heat fluxes: Retrieval and variability analysis. Ph.D. thesis, Gujarat University, Ahmedabad, India, 170 pp.

  • Singh, R., C. M. Kishtawal, and P. C. Joshi, 2005a: Estimation of monthly mean air–sea temperature difference from satellite observations using genetic algorithm. Geophys. Res. Lett., 32 .L02807, doi:10.1029/2004GL021531.

    • Search Google Scholar
    • Export Citation
  • Singh, R., P. C. Joshi, and C. M. Kishtawal, 2005b: A new technique for estimation of surface latent heat fluxes using satellite-based observations. Mon. Wea. Rev., 133 , 26922710.

    • Search Google Scholar
    • Export Citation
  • Szpiro, G. G., 1997: Forecasting chaotic time series with genetic algorithms. Phys. Rev., E55 , 25572568.

  • Walton, C. C., 1988: Nonlinear multichannel algorithm for estimating sea surface temperature with AVHRR satellite data. J. Appl. Meteor., 27 , 115124.

    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., 1997: A well calibrated ocean algorithm for SSM/I. J. Geophys. Res., 102 , 87038718.

  • Wick, G. A., W. J. Emery, and P. Schlussel, 1992: A comprehensive comparison between satellite measured skin and multichannel sea surface temperature. J. Geophys. Res., 97 , 55695595.

    • Search Google Scholar
    • Export Citation
  • Zhang, G. J., 1995: Use of monthly mean data to compute surface turbulent fluxes in the tropical Pacific. J. Climate, 8 , 30843090.

Fig. 1.
Fig. 1.

Temporal variation of the bias B (g kg−1) over the north Arabian Sea, Kuroshio, and the Gulf Stream, during 1988–93.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 2.
Fig. 2.

Bias between SMD- and GA-retrieved Qa (g kg−1) for (a) phase I GA, and (b) phase II GA, during sample II data.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 3.
Fig. 3.

Rmse between SMD- and GA-retrieved Qa (g kg−1) for (a) phase I GA, and (b) phase II GA, during sample II data.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 4.
Fig. 4.

Improvement parameter η (g kg−1), w.r.t. (a) L86, (b) CSSA, and (c) L86 during sample II data.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 5.
Fig. 5.

Annual cycle of Qa (g kg−1), (a) the north Arabian Sea, (b) Kuroshio, and (c) the Gulf Stream over sample II data.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 6.
Fig. 6.

Zonal averages, (a) biases (W m−2), and (b) rmse (W m−2) between SMD-observed and satellite-derived monthly mean LHF (W m−2) for sample II data.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 7.
Fig. 7.

Annual cycle of (a) U (m s−1), (b) Qs (g kg−1), and (c) LHF (W m−2) over the north Arabian Sea, over sample II data.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 8.
Fig. 8.

Annual cycle of (a) U (m s−1), (b) Qs (g kg−1), and (c) LHF (W m−2), over Kuroshio, over sample II data.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 9.
Fig. 9.

Annual cycle of (a) U (m s−1), (b) Qs (g kg−1), and (c) LHF (W m−2), over the Gulf Stream, over sample II data.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 10.
Fig. 10.

Annual cycle of alternate LHF (W m−2) over (a) the north Arabian Sea, (b) Kuroshio, and (c) the Gulf Stream, over sample II data.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 11.
Fig. 11.

Zonal averages rmse (W m−2) and biases (W m−2) between SMD-observed and satellite-derived monthly mean SHF (W m−2) for sample II data.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 12.
Fig. 12.

Annual cycle of SHF (W m−2) over (a) the north Arabian Sea, (b) Kuroshio, and (c) the Gulf Stream over sample II data.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 13.
Fig. 13.

Monthly averaged LHF (W m−2), (a) LHFSMD, (b) LHFGA-II, and (c) LHFSMD − LHFGA-II, during January of sample II data.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 14.
Fig. 14.

Same as in Fig. 13, but during July.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 15.
Fig. 15.

Monthly averaged SHF (W m−2), (a) SHFSMD, (b) SHFSAT, and (c) SHFSMD − SHFSAT, during January of sample II data.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 16.
Fig. 16.

Same as in Fig. 15, but during July.

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Fig. 17.
Fig. 17.

Annually averaged (1988–2003) GA-based heat fluxes: (a) LHF (W m−2) and (b) SHF (W m−2).

Citation: Monthly Weather Review 134, 3; 10.1175/MWR3119.1

Table 1.

Value of constants used in Eq (3).

Table 1.
Table 2.

Value of constants used in Eq. (4).

Table 2.
Table 3.

Statistics of satellite-derived Qa.

Table 3.
Table 4.

Statistics of satellite-derived LHF.

Table 4.
Table 5.

The alternate datasets of LHF.

Table 5.
Table 6.

Statistics of annual mean alternate LHF.

Table 6.
Table 7.

Statistics of monthly mean alternate LHF.

Table 7.
Save
  • Alvarez, A., C. Lopez, M. Riera, E. Hernandez-Garcia, and J. Tintore, 2000: Forecasting the SST space-time variability of the Alboran Sea with genetic algorithms. Geophys. Res. Lett., 27 , 27092712.

    • Search Google Scholar
    • Export Citation
  • Bentamy, A., K. B. Katsaros, A. M. Mestas-Nuñez, W. M. Drennan, B. E. Forde, and H. Roquet, 2003: Satellite estimates of wind speed and latent heat flux over the global oceans. J. Climate, 16 , 637655.

    • Search Google Scholar
    • Export Citation
  • Bourras, D., L. Eymard, W. T. Liu, and H. Dupuis, 2002: An integrated approach to estimate instantaneous near-surface air temperature and sensible heat flux fields during the SEMAPHORE experiment. J. Appl. Meteor., 41 , 241252.

    • Search Google Scholar
    • Export Citation
  • Brown, J. W., O. B. Brown, and R. H. Evans, 1993: Calibration of AVHRR infrared channels: A new approach to non-linear corrections. J. Geophys. Res., 98 , 1825718268.

    • Search Google Scholar
    • Export Citation
  • Bunker, A. F., 1976: Computation of surface energy flux and annual air–sea interactions cycle of the North Atlantic Ocean. Mon. Wea. Rev., 104 , 11221140.

    • Search Google Scholar
    • Export Citation
  • Chou, S. H., R. M. Atlas, C. L. Shie, and J. Ardizzone, 1995: Estimates of surface humidity and latent heat fluxes over oceans from SSM/I data. Mon. Wea. Rev., 123 , 24052425.

    • Search Google Scholar
    • Export Citation
  • Chou, S. H., C. L. Shie, R. M. Atlas, and J. Ardizzone, 1997: Air–sea fluxes retrieved from Special Sensor Microwave Imager data. J. Geophys. Res., 102 , 1270512726.

    • Search Google Scholar
    • Export Citation
  • Chou, S. H., E. Nelkin, J. Ardizzone, R. M. Atlas, J. Ardizzone, and C. L. Shie, 2003: Surface turbulent heat and momentum fluxes over global oceans based on the Goddard Satellite Retrievals, Version 2 (GSSTF2). J. Climate, 16 , 32563273.

    • Search Google Scholar
    • Export Citation
  • da Silva, A. M., C. C. Young, and S. Levitus, 1994: Algorithms and Procedures. Vol. 1, Atlas of Surface Marine Data 1994, NOAA Atlas NESDIS 6, 83 pp.

  • Dong, S., and K. A. Kelly, 2004: Heat budget in the Gulf Stream region: The importance of heat storage and advection. J. Phys. Oceanogr., 34 , 12141231.

    • Search Google Scholar
    • Export Citation
  • Esbensen, S. K., and R. W. Reynolds, 1981: Estimating monthly averaged air–sea transfers of heat and momentum using the bulk aerodynamic method. J. Phys. Oceanogr., 11 , 457465.

    • Search Google Scholar
    • Export Citation
  • Gautier, C., P. Peterson, and C. Jones, 1998: Ocean surface air temperature derived from multiple data sets and artificial neural network. Geophys. Res. Lett., 25 , 42174220.

    • Search Google Scholar
    • Export Citation
  • Grabl, H., V. Jost, R. Kumar, J. Schulz, P. Bauer, and P. Schlussel, 2000: A climatological atlas of satellite-derived air–sea interaction parameters over the world ocean. Rep. 312, Max Planck Institute for Meteorology, Hamburg, Germany, 120 pp.

  • Hasse, L., and S. D. Smith, 1997: Local sea surface wind, wind stress, and sensible and latent heat fluxes. J. Climate, 10 , 27112724.

    • Search Google Scholar
    • Export Citation
  • Hastenrath, S., and P. J. Lamb, 1977: Climate Atlas of the Tropical Atlantic and Eastern Pacific Oceans. University of Wisconsin Press, 340 pp.

    • Search Google Scholar
    • Export Citation
  • Hastenrath, S., and P. J. Lamb, 1979: Heat Budget Atlas of the Tropical Atlantic and Eastern Pacific Oceans. University of Wisconsin Press, 104 pp.

    • Search Google Scholar
    • Export Citation
  • Hsiung, J., 1986: Mean surface energy fluxes over the global ocean. J. Geophys. Res., 91 , 1058510606.

  • Jones, C., P. Peterson, and C. Gautier, 1999: A new method for deriving ocean surface specific humidity and air temperature: An artificial neural network approach. J. Appl. Meteor., 38 , 12291245.

    • Search Google Scholar
    • Export Citation
  • Jourdan, D., and C. Gautier, 1995: Comparison between global latent heat flux computed from multisensor (SSM/I, AVHRR) and from in situ data. J. Atmos. Oceanic Technol., 12 , 4672.

    • Search Google Scholar
    • Export Citation
  • Kishtawal, C. M., S. Basu, F. Patadia, and P. K. Thapliyal, 2003: Forecasting summer rainfall over India using genetic algorithm. Geophys. Res. Lett., 30 , 15.

    • Search Google Scholar
    • Export Citation
  • Kishtawal, C. M., F. Patadia, R. Singh, S. Basu, M. S. Narayanan, and P. C. Joshi, 2005: Automatic estimation of tropical cyclone intensity using multi-channel TMI data: A genetic algorithm approach. Geophys. Res. Lett., 32 .L11804, doi:10.1029/2004GL022045.

    • Search Google Scholar
    • Export Citation
  • Konda, M., and N. Imasato, 1996: A new method to determine near sea surface air temperature by using satellite data. J. Geophys. Res., 101 , 1434914360.

    • Search Google Scholar
    • Export Citation
  • Kubota, M., and A. Shikauchi, 1995: Air temperature at ocean surface derived from surface level humidity. J. Oceanogr., 51 , 619634.

  • Kubota, M., N. Iwasaka, S. Kizu, M. Konda, and K. Kutsuwada, 2002: Japanese Ocean Flux Data Sets with Use of Remote Sensing Observations (J-OFURO). J. Oceanogr., 58 , 213225.

    • Search Google Scholar
    • Export Citation
  • Liu, W. T., 1986: Statistical relation between monthly mean precipitable water and surface-level humidity over global oceans. Mon. Wea. Rev., 114 , 15911602.

    • Search Google Scholar
    • Export Citation
  • Liu, W. T., 1988: Moisture and latent heat flux variabilities in the tropical Pacific derived from satellite data. J. Geophys. Res., 93 , 67496760.

    • Search Google Scholar
    • Export Citation
  • Luther, D. S., and D. E. Harrison, 1984: Observing long-period fluctuations of surface winds in the tropical Pacific: Initial results from island data. Mon. Wea. Rev., 112 , 285302.

    • Search Google Scholar
    • Export Citation
  • Schlussel, P., L. Schanz, and G. English, 1995: Retrieval of latent heat flux and longwave irradiance at sea surface differences at the sea surface from SSM/I and AVHRR measurements. Adv. Space Res., 16 , 107116.

    • Search Google Scholar
    • Export Citation
  • Schulz, J., 2003: Satellite remote sensing of turbulent heat fluxes. Mausam, 54 , 1. 7588.

  • Schulz, J., P. Schlussel, and H. Grassl, 1993: Water vapour in atmospheric boundary layer over oceans from SSM/I measurements. Int. J. Remote Sens., 14 , 27732789.

    • Search Google Scholar
    • Export Citation
  • Schulz, J., M. Jeans, E. Stefan, and P. Schlussel, 1997: Evaluation of satellite-derived latent heat fluxes. J. Climate, 10 , 27822795.

    • Search Google Scholar
    • Export Citation
  • Singh, R., 2004: Satellite observation of boundary layer heat fluxes: Retrieval and variability analysis. Ph.D. thesis, Gujarat University, Ahmedabad, India, 170 pp.

  • Singh, R., C. M. Kishtawal, and P. C. Joshi, 2005a: Estimation of monthly mean air–sea temperature difference from satellite observations using genetic algorithm. Geophys. Res. Lett., 32 .L02807, doi:10.1029/2004GL021531.

    • Search Google Scholar
    • Export Citation
  • Singh, R., P. C. Joshi, and C. M. Kishtawal, 2005b: A new technique for estimation of surface latent heat fluxes using satellite-based observations. Mon. Wea. Rev., 133 , 26922710.

    • Search Google Scholar
    • Export Citation
  • Szpiro, G. G., 1997: Forecasting chaotic time series with genetic algorithms. Phys. Rev., E55 , 25572568.

  • Walton, C. C., 1988: Nonlinear multichannel algorithm for estimating sea surface temperature with AVHRR satellite data. J. Appl. Meteor., 27 , 115124.

    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., 1997: A well calibrated ocean algorithm for SSM/I. J. Geophys. Res., 102 , 87038718.

  • Wick, G. A., W. J. Emery, and P. Schlussel, 1992: A comprehensive comparison between satellite measured skin and multichannel sea surface temperature. J. Geophys. Res., 97 , 55695595.

    • Search Google Scholar
    • Export Citation
  • Zhang, G. J., 1995: Use of monthly mean data to compute surface turbulent fluxes in the tropical Pacific. J. Climate, 8 , 30843090.

  • Fig. 1.

    Temporal variation of the bias B (g kg−1) over the north Arabian Sea, Kuroshio, and the Gulf Stream, during 1988–93.

  • Fig. 2.

    Bias between SMD- and GA-retrieved Qa (g kg−1) for (a) phase I GA, and (b) phase II GA, during sample II data.

  • Fig. 3.

    Rmse between SMD- and GA-retrieved Qa (g kg−1) for (a) phase I GA, and (b) phase II GA, during sample II data.

  • Fig. 4.

    Improvement parameter η (g kg−1), w.r.t. (a) L86, (b) CSSA, and (c) L86 during sample II data.

  • Fig. 5.

    Annual cycle of Qa (g kg−1), (a) the north Arabian Sea, (b) Kuroshio, and (c) the Gulf Stream over sample II data.

  • Fig. 6.

    Zonal averages, (a) biases (W m−2), and (b) rmse (W m−2) between SMD-observed and satellite-derived monthly mean LHF (W m−2) for sample II data.

  • Fig. 7.

    Annual cycle of (a) U (m s−1), (b) Qs (g kg−1), and (c) LHF (W m−2) over the north Arabian Sea, over sample II data.

  • Fig. 8.

    Annual cycle of (a) U (m s−1), (b) Qs (g kg−1), and (c) LHF (W m−2), over Kuroshio, over sample II data.

  • Fig. 9.

    Annual cycle of (a) U (m s−1), (b) Qs (g kg−1), and (c) LHF (W m−2), over the Gulf Stream, over sample II data.

  • Fig. 10.

    Annual cycle of alternate LHF (W m−2) over (a) the north Arabian Sea, (b) Kuroshio, and (c) the Gulf Stream, over sample II data.

  • Fig. 11.

    Zonal averages rmse (W m−2) and biases (W m−2) between SMD-observed and satellite-derived monthly mean SHF (W m−2) for sample II data.

  • Fig. 12.

    Annual cycle of SHF (W m−2) over (a) the north Arabian Sea, (b) Kuroshio, and (c) the Gulf Stream over sample II data.

  • Fig. 13.

    Monthly averaged LHF (W m−2), (a) LHFSMD, (b) LHFGA-II, and (c) LHFSMD − LHFGA-II, during January of sample II data.

  • Fig. 14.

    Same as in Fig. 13, but during July.

  • Fig. 15.

    Monthly averaged SHF (W m−2), (a) SHFSMD, (b) SHFSAT, and (c) SHFSMD − SHFSAT, during January of sample II data.

  • Fig. 16.

    Same as in Fig. 15, but during July.

  • Fig. 17.

    Annually averaged (1988–2003) GA-based heat fluxes: (a) LHF (W m−2) and (b) SHF (W m−2).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 269 69 17
PDF Downloads 115 27 4