Schmidt et al. [1]

Comparison between satellite wind products and in situ Wave Glider

wind observations in the Southern Ocean

Kevin M. Schmidta, Sebastiaan Swartb,c, Chris Reasonc, Sarah Nicholsonb,c

a University of Cape Town, Marine Research Institute

b Southern Ocean Carbon & Climate Observatory, Council for Scientific

and Industrial Research (CSIR), Rosebank, Cape Town

c University of Cape Town, Department of Oceanography

Author’s emails:

Kevin Schmidt: kmschmdt@gmail.com

Sebastiaan Swart: seb.swart@gmail.com

Chris Reason: chris.reason@uct.ac.za

Sarah-Anne Nicholson: sarahanne.n@gmail.com

Keywords: Sea surface winds, reanalysis, intercomparison, wave gliders, wind speed,

scatterometer, wind products, Southern Ocean

Abstract:

Surface wind datasets are required to be of high spatial and temporal resolution

and high precision in order to accurately force coupled atmosphere-ocean numerical

models and understand ocean-atmospheric processes. In situ observed sea surface

winds from the Southern Ocean are scarce and consequently the validity of simulation

models is often in question. Multiple wind data products were compared to high

resolution in situ measurements of wind speed from Wave Glider deployments in the

Subantarctic Southern Ocean. The intent was to determine which blended satellite or

reanalysis product best represents the magnitude and variability of the observed wind

field. Results show that the ECMWF reanalysis wind product is more accurate in

representing the temporal variability of winds, exhibiting consistently higher correlation

coefficients with in situ data across all wind speed categories. However, the

NCEP/NCAR Reanalysis II product matches in situ trends of deviation from the mean

and performs best in depicting the mean wind state, especially during high mean wind

states. The high-resolution ECMWF product is best used during periods of lower mean

wind state, and exhibits lower biases in wind speed across all speed categories. It can

be concluded that the ECMWF product leads to smaller differences in wind speeds from

the in situ data due to a combination of higher temporal and spatial sampling.

Schmidt et al. [2]

1. Introduction:

Mid- to high-latitude regions in the Southern Ocean are host to the strongest

wind fields at the ocean surface. These strong winds (e.g. speeds >20 m.s-1; Yuan

(2004)) significantly impact upper ocean properties and processes, such as mixed layer

dynamics, Ekman processes and air-sea exchange. Exchanges in heat, moisture, and

momentum at the air-sea interface are facilitated by sea surface winds. In addition to

driving physical processes at the sea surface, these winds also have implications for

biological processes which extend below the surface. The flux in carbon dioxide (CO2)

between the atmosphere and ocean is closely related to wind speed (Wanninkhof,

2009), and as such sea surface winds are of interest to many scientists studying the

biogeochemical cycle within the ocean.

The Southern Ocean, owing to its limited access to sources of iron (e.g.

landmasses), is known as an iron limited High Nitrate Low Chlorophyll (HNLC) region. It

remains an integral player in the regulation of the global climate by sequestering

atmospheric CO2 (Takahashi et al., 2012). Many recent studies (Thomalla et al., 2011;

Fauchereau et al., 2011; Swart et al., 2014; Domingues et al., 2014; Carranza and Gille,

2015; and Nicholson et al., submitted 2016) have been conducted in this region to

investigate the physical mechanisms which drive the abundance of primary production.

During times of iron limitation, transient deepening events of the surface mixed-layers is

hypothesized to be driven by variability in sea surface winds which consequently

creates a supply of iron via entrainment. However, generally these studies base their

assumptions on coarse satellite derived data, which may represent this process poorly.

A better understanding of Southern Ocean wind field characteristics, such as variability

and strength, may help to explain its impacts on mixing and the response of physical-

biological processes in the Southern Ocean.

Surface winds are measured using in situ techniques or remote sensing

instruments. It has been observed by several scientists (Stoffelen and Anderson, 1997;

Andrews and Bell, 1998; Atlas, 1999) that numerical weather prediction (NWP) models

benefit from scatterometer wind observations. However due to the varying methods of

collecting and synthesizing wind measurements, the uniformity of sea surface wind data

is often questioned. Satellite instruments use wave scatter to infer wind vectors

Schmidt et al. [3]

generally at a height of 10 meters above sea level, while in situ measurements are

generally taken at varying heights above sea level dependent on the platform (e.g. ship,

mooring, or glider). As a result, many studies (Yuan et al., 2009), involving wind data

require a validation prior to the start of their analysis. In order to determine the ability of

satellite datasets to accurately represent wind fields at the air-sea interface, and to see

if post processing is required to reduce errors once compared to in situ measurements.

There is a great need to reduce or eliminate this time costly step in wind data analysis in

the Southern Ocean.

A commonly used global wind product is the QuikSCAT dataset. QuikSCAT was

a remote sensing satellite, which provided global ocean surface wind vector

measurements from 1999 to 2009. In an evaluation of extreme wind events in the

Southern Ocean, Yuan (2004) validated QuikSCAT wind data with reanalysis wind data

from the National Center for Environmental Prediction (NCEP), National Center for

Atmospheric Research (NCAR) and European Centre of Medium Weather Forecasts

(ECMWF). It was found that both weather station and QuikSCAT winds were stronger

than reanalysis data, albeit only during high wind events (Yuan, 2004).

The approach of Yuan (2004) required the input of many different wind products

to investigate the spatial and seasonal variability of high wind characteristics. The

occurrence and intensity of extreme events was examined along with a cross

examination of the QuikSCAT scatterometer data against other wind products,

specifically reanalysis wind products. Reanalysis wind products are created by

assimilating in-situ observations into numerical weather prediction models. This

produces datasets with greater spatial and temporal resolutions. The reanalysis

products that Yuan (2004) used were limited with regard to in-situ data input. The

majority of the input came from weather station observations from Macquarie Island,

due to the remote area of the study. Yuan (2004) showed that monthly mean wind

observations from satellite averages over the Southern Ocean were in agreement with

model simulations except for the period from May to October (winter months), when

scatterometer winds were stronger than simulated winds, thereby suggesting a strong

winter bias in data. Similar to Hilburn et al. (2003), Yuan (2004) attributed this

Schmidt et al. [4]

discrepancy to the NCEP/NCAR reanalysis missing some storms entirely, along with a

lack of observational data assimilated into the reanalyses.

Patoux et al. (2009) modified scatterometer derived pressure fields with

reanalysis wind data. QuikSCAT data was blended with ECMWF forecast data to

produce modified pressure swaths. A wavelet-based method was used to blend the

swaths from the scatterometer and surface pressure model. The resulting blends were

successful in extending the tracks of more than 1100 cyclones by at least one synoptic

period. The modified cyclones, however, were observed to be deeper than the ECMWF

cyclones with a maximum mean difference of 0.3 hPa observed in the Southern Ocean

between 45⁰S - 55⁰S (Patoux et al., 2009). This difference was initially attributed to the

partial blending methodology, but was further investigated by Yuan et al. (2009) who

assessed the impacts of the differences in momentum and heat fluxes at the air-sea

interface in the Southern Ocean. Yuan et al. (2009) claim that even small mesoscale

cyclones contribute significantly to heat fluxes at the air-sea interface due to the high

frequency of their occurrence. As such, a correct representation of mesoscale cyclones

in NWP models of the Southern Ocean is therefore critical (Yuan et al. 2009). Although

these studies have aided in improving estimates of wind stress and understanding time

and space variability of Southern Ocean wind events, significant gaps in our

understanding of wind stress accuracy and uncertainty exist.

The aim of this paper is to determine which wind data product (satellite blended,

reanalysis and modelled) best represents the wind field variability of the Southern

Ocean. This is done by comparing wind products with in situ sea surface wind data

gathered by Wave Glider technology deployed in a series of experiments in the

Subantarctic Ocean. The paper is organized as follows: an introduction to the surface

wind datasets, their sources, and the transformations performed to collocate them in

space and time. In this work, both in-situ real time winds as well as satellite blended

near surface derived wind fields are considered. In the following section, a detailed

description of the statistical analysis performed in this study is provided. In Section 3,

remotely measured winds are compared with in-situ winds at a fixed point in the

Southern Ocean. The final section presents a discussion of the results and their

implications for biological studies.

Schmidt et al. [5]

2. Data and Methodology

2.1 Satellite Wind Data

The QuikSCAT satellite launched by the U.S. National Aeronautics and Space

Administration (NASA) in 1999 was fitted with a SeaWinds scatterometer. This active

microwave radar sampled 25 kilometer measurements across a 1600 kilometer swath

every 24 hours (Chelton et al. 2004). The SeaWind (SW) gridded and blended sea

surface vector winds were generated from the multiple satellite observations of DOD,

NOAA and NASA along with the input of satellite wind retrievals from Remote Sensing

Systems, Inc. Surface wind fields are available for download at https://www.ncdc.noaa.

gov/data-access/marineocean-data/blended-global/blended-sea-winds. The methods

used to generate such data include objective analysis and simple spatiotemporally

weighted interpolation. This product provides global ocean coverage with a spatial

resolution of 0.25⁰ x 0.25⁰ and a 6 hourly temporal resolution. The data are provided at

a height of 10m above sea level.

NOAA uses a global NWP model called the Climate Forecast System which

represents the interaction between the air-sea interfaces throughout the world’s oceans.

The CFS version II (CFSv2) operational near real time wind vector dataset provides

time series at a period of record from 01 January 2011 – 01 April 2016 at a 6 hourly

temporal resolution. CFS time series products are available at 0.2, 0.5, 1.0, and 2.5

degree horizontal resolutions and for this study the highest resolution was selected for

use; a spatial scale of 0.205⁰ x 0.204⁰. This product is available for download at

http://rda.ucar.edu/datasets/ds094.1/. The data are provided at a height of 10m above

sea level.

European Centre of Medium Weather Forecasts (ECMWF) ERA-Interim Global

Reanalysis Sea Surface Winds are available for download at http://apps.ecmwf.int/

datasets/data/ interim-full-daily/. The spatial resolution of the data set is T255 spectral,

specifically 0.125⁰ x 0.125⁰. The project is in near real-time production and the data are

calculated at a height of 10m above sea level and are available at a 6 hourly temporal

resolution.

Schmidt et al. [6]

The NCEP/NCAR Reanalysis project has created a sea surface wind dataset

which covers the period from 01 January 1979 to the present. The development of the

NCEP–DOE Atmospheric Model Intercomparison Project (AMIP-II) Reanalysis R-2

(NCEPII) was initiated in the late 1990’s and the data are available for download at

http://rda.ucar.edu/datasets/ds091.0. It is important to note that NCEPII is of the same

spatial (1.9047⁰ x 1.8750⁰) and temporal resolution as NCEPI and incorporates similar

raw observational data. NCEPII is a global Gaussian Latitude/Longitude T62 resolution.

The data are calculated at a height of 10m above sea level and are available at a 6

hourly temporal resolution.

Table 1: Characteristics of the different wind products used in this study

2.2 In situ Wind Observations

2.2.1 Experimental Setup and Region

The in-situ data used in this study were collected via a Liquid Robotics Wave

Glider (WG) autonomous surface vehicle. The primary components of the WG include a

surface float, an underwater glider, and a tether cable which connects the two. Solar

panels and batteries power sensors and communication systems which enabled it to

remain at sea for an extended period of time (multiple months). At the ocean surface,

this vehicle derived its propulsion from wave energy and was set to navigate an octagon

track around a fixed location in the Southern Ocean (Figure 1). Two different gliders

were deployed at different locations during research cruises in 2013 and twice in 2015.

Using real-time Iridium burst communication, these gliders navigated to predetermined

Wind

Products

Time (h)

Resolution

Spatial

Resolution

Paper

Reference

Time Coverage

SW 6 0.25⁰ x 0.25⁰ Zhang 2006 09 July 1987 -

Present

ECMWF 6 0.125⁰ x 0.125⁰ Dee et al. 2011 1979 – 31 Jan 2016

CFSv2 6 0.205⁰ x 0.204⁰ Saha et al. 2014 01 April 2011 -

Present

NCEPII 6 1.875⁰ x 1.904⁰ Kanamitsu et al. 2002 1979 - Present

Schmidt et al. [7]

coordinates. Glider CSIR1, deployed in November of 2013 and 2015, was set to

navigate to 43°S, 8.5°E. Glider CSIR2 was deployed in July of 2015 and navigated a

large area, as shown in Figure 1. Shown in Figure 2, upon arrival at the study site a

pseudomooring circular sampling pattern with a diameter of 16 km was maintained

(Monteiro et al., 2015). This technology is designed with a cloud-based infrastructure

that manages the transfer of real-time data from the WG to the pilots and scientists.

Figure 1: Scope of the study area in the Subantarctic Southern

Ocean, with the majority of data points collected at 43°00'S

8°30'0E. A color map overlay shows the monthly averaged wind

speeds in meters per second (ECMWF) of the study area for

December 2015. A photo insert shows the surface setup of the

Wave Glider with Airmar meteorological station.

The WG is designed to augment existing marine technology by providing

autonomous real-time data collection of parameters at the air-sea interface. The

Schmidt et al. [8]

Southern Ocean Carbon & Climate Observatory (SOCCO) at CSIR uses this technology

in their integrated multi-platform approach for a series of the Southern Ocean Seasonal

Cycle Experiments (SOSCEx), which use research vessels, WGs, profiling gliders, bio-

optic floats, satellite data and numerical models to explore the climate sensitivity of

carbon and ecosystem dynamics. The WGs deployed during the research expeditions

were equipped with an Airmar WX-200 meteorological sensor mounted on a mast 0.70

m above sea level. Via ultrasonic transducers, this sensor detected instantaneous

changes in weather and measured wind speed and wind direction. The sensor

averaged wind speed and direction every 10 minutes.

Figure 2: Map of sampling region in the Subantarctic Southern Ocean, centered at 43°00'S

8°30'0E. The WG sampling locations are indicated with respect to spatial distribution of gridded

wind products.

2.2.2 Direct comparison approaches between in situ and satellite product winds

Satellite derived winds are remotely retrieved by microwave sensors which

measure wind in an equivalent neutrally stable state. This process makes them

sensitive to ocean surface roughness due to the stratification of the atmosphere above

Schmidt et al. [9]

sea level. These sensors are thus calibrated to an equivalent neutral wind at a

reference height of 10m above the sea surface (Liu and Tang, 1996; Bourassa et al.

1999a; Bourassa et al. 2003; Chelton et al. 2004). These equivalent neutral winds are

the winds that would exist if the atmospheric boundary layer was neutrally stratified

(Chelton et al. 2004; Carvalho et al. 2013). In order to compare real-time stability

dependent winds to equivalent neutral winds, a transformation must be performed to

shift the WG winds to the same reference level as satellite winds. Any extrapolation

method to shift measured wind speed to a different height is a function of atmospheric

turbulence. Wind shear and the buoyant forcing of the atmosphere must be taken into

account; ultimately the vertical density stratification of the atmosphere must be

accurately represented (Ruti et al., 2008).

Various methods can be used depending on the amount of input data available.

One commonly used method, proposed by Liu and Tang (1996) is based on the bulk

aerodynamic relation. This method takes into account differences in atmospheric

stability during extrapolation. A prior analysis by Liu (1990) on moisture induced

atmospheric instability indicated that along with wind speed, air and sea surface

temperatures, pressure, and relative humidity should be included in the computation of

equivalent neutral wind. The algorithm developed to do this transformation is based on

the following relation

𝑈𝑛 − 𝑈𝑠 = (𝑢∗

𝑘)[ln(

𝑍

𝑍0)] + (

𝑢∗

𝑘)𝜑𝑗(𝑍, 𝑍0, 𝐿), (2)

where the term (Un – Us) denotes wind speed relative to the surface speed (Us) as a

function of height Z, while ln(Z/Z0) is an adjustment term for height (Singh et al., 2013).

This approach requires additional observational inputs of air and sea surface

temperature, relative humidity, and atmospheric pressure which are not available for

this study period and site. As in Satheesan et al. (2007), Singh et al. (2013), Sudha and

Rao (2013), and Yang et al. (2014), the present study required a method which does

not include inputs associated with atmospheric stability.

Bentamy et al. (2008) and Qing and Chen (2015) use the wind profile power law

in their validation tasks of OSCAT and ASCAT wind data in the Atlantic Ocean. This law

Schmidt et al. [10]

shows the relationship between wind speeds at one height and those at another and is

expressed as

𝑈

𝑈𝑟= (

𝑍

𝑍𝑟)

∝ , (3)

where U denotes wind speed at height Z (Qing and Chen, 2015). Ur is the known wind

speed at reference height Zr. With this method, neutrally equivalent winds are calculated

with the use of the coefficient (α) which varies with atmospheric stability. Qing and Chen

(2015) found that over open water, this coefficient should be assigned a value of 0.11.

However, no intercomparison has been performed to determine the difference between

a power expression method and the method proposed by Liu and Tang (1996). As such

the present study uses a mixing-length approach used by Herrara et al. (2005), Ruti et

al. (2008), Carvalho et al. (2013), Singh et al. (2013), Alvarez et al. (2014) and Yang et

al. (2014) in the vertical transformation of in situ surface wind data to a reference height

of 10 meters. With a lack of atmospheric stability input, a logarithmic method proposed

by Peixoto and Oort (1992) is used. This is expressed as

𝑈𝑍 = 𝑈𝑍𝑚

ln𝑍

𝑍0

ln𝑍𝑚𝑍0

, (4)

where Uz is wind speed at height Z, Zm denotes measurement height, and Z0 represents

the roughness length. A typical oceanic value of 1.52 x 10-4 m was assumed for Z0

(Peixoto and Oort, 1992). This approach generates a logarithmically varying vertical

wind profile while assuming neutral stability conditions. An inter-comparison of the

correction methods proposed by Liu and Tang (1996) and by Peixoto and Oort (1992)

was performed by Mears et al. (2001). They conducted a comprehensive analysis to

determine if these laws can account for effects due to differences in atmospheric

stability. It was concluded that the Liu and Tang (1996) correction is typically on

average 0.12 m.s-1 stronger than a logarithmic correction (Mears et al., 2001; Pickett et

al., 2003; Ruti et al., 2008). Under stable conditions, Liu and Tang (1996) corrected

winds are greater than logarithmic corrected winds, and for stable conditions the

opposite is seen (Carvalho et al., 2013). Ruti et al. (2008) also compared these

correction methods and concluded that generally a difference in collocated wind speed

Schmidt et al. [11]

is only observed during extreme wind events, where wind speeds exceed 15 m.s-1.

Singh et al. (2013) conclude that during these extreme wind events, wind speed

transformation carried out by a logarithmic profile may lead to errors of 1–1.5 m.s-1.

However, it can be assumed that the use of a logarithmic extrapolation method will not

cause discernable error as there are very few instances in the WG data where the wind

speed is greater than 15 m.s-1. As such, the logarithmic method proposed by Peixoto

and Oort (1992) was chosen for this study.

Each WG dataset consisted of more than ten thousand wind measurements

averaged over 10 minute intervals. Initially, the WG data was difficult to interpret due to

the level of noise which accompanies high frequency data. In order to reduce the noise

of the data, a simple running mean filter was applied over the time series. Because we

are interested in a 6 hourly resolution, the running mean was set to average each data

point over a 6 hourly period to match the temporal resolution of the wind products used

in comparison. Once the noise was reduced, it was necessary to temporally sample the

WG data in an attempt to scale down the sample size to match the varying wind

products. A comparison of possible methods was performed at the start of this study. An

instantaneous sampling method at hourly intervals was applied, as well as 10 minute

bins were temporally averaged to estimate hourly WG parameters. Figure 2 shows the

results of the different binning methods at a 6 hourly resolution.

Figure 3: A comparison of binning methods applied to in situ data. WG data are gridded

temporally via averaging and instantaneous sampling on hourly and 6 hourly time scales.

An instantaneous sampling approach may be appropriate in cases where the in-

situ time series matches exactly with the satellite time series. On the other hand,

according to Ruti et al. (2008) an averaging method may require consideration of the

phase velocities of cyclones in the region of interest. By determining the typical phase

Schmidt et al. [12]

velocity for Mediterranean cyclones, Ruti et al. (2008) were able to determine a general

time frame of 30-60 minutes over which in-situ data from the Mediterranean should be

averaged when compared with scatterometer winds. Other studies, such as one by

Bourassa et al. (2003), state that the optimal averaging period varies with respect to

wind speed at the moment of observation. During high wind periods, a shorter

averaging period may result in a lower RMSE, whereas during low wind speed periods,

longer averaging periods result in lower RMSE (Bourassa et al., 2003). This variability in

optimum averaging time with respect to wind speed is anticipated from Taylor’s

hypothesis (Taylor, 1938). This study took an averaged sampling approach to binning

the time series. The data was binned twice, first to create an hourly product and then a

6 hourly product. Instead of sampling over the duration of each hour, data points before

and after each desired interval were averaged.

In order to compare blended wind fields from satellite data with the WG

observations, a collocation procedure was necessary to match the datasets spatially

and temporally. For all wind products, a linear interpolation procedure was performed to

produce wind pairs from both datasets. This method was possible due to the

aforementioned temporal gridding of the WG data. Because the WGs were set to circle

a fixed point, the majority of the collocations were performed on a coordinate points

close to 43°00’S 8°30'E (Figure 1 and 2).

2.3 Statistical Comparison

WG meteorological records are provided every 10 minutes, while the satellite

products and numerical weather products provide four data points per day

(corresponding to the hours 00H00, 06H00, 12H00, and 18H00 UTC). Initial statistical

analysis is intended to assess the quality of the simultaneous records with respect to

temporal and spatial accuracy. In order to determine if the error associated with the

satellite and reanalysis derived wind data is related to a particular wind speed, the wind

data were analysed as a whole as well as in low, medium, and high wind speed

categories. The thresholds for these wind speed categories were determined by the

lower and upper quartiles of the WG dataset. This varied for each time series; for

example, the first deployment of the CSIR1 WG in 2013 had a low speed thresholds of

Schmidt et al. [13]

3.6 m.s-1 and a high wind speed threshold of 6.1 m.s-1. However, the same WG in 2015

had a low speed threshold of 9.0 m.s-1, and a high speed threshold of 16 m.s-1. In all

analyses, the WG wind speeds fall within these categories. However, all valid data

collocated records of each wind product were considered. As a result, there will be

times during which the satellite and reanalysis derived wind speeds fall outside these

wind speed categories. The purpose of this approach was to show the error

dependence on wind speed as well as to determine which product best represented

each wind category.

The mean as well as the standard deviation of all wind products per wind speed

category are initially calculated with the intent to show the standard variance from the

mean. Statistical parameters such as the Root Mean Square Error (RMSE)

𝑅𝑀𝑆𝐸 = [1

𝑁 ∑ (𝑈𝑖 − 𝑈𝑗)

2𝑁𝑖=1 ]1/2

(5)

and correlation coefficient (R2) are calculated to assess the ability of the wind products

to represent the observed winds. Ui and Uj are variables which represent the satellite

product wind speed and the observed wind speed by the WG, respectively. More

specifically, these parameters will determine the accuracy of the varying wind product’s

ability to represent the temporal variability of the wind. Bias is calculated

𝐵𝐼𝐴𝑆 =1

𝑁 ∑ (𝑈𝑖 − 𝑈𝑗)𝑁

𝑖=1 (6)

in order to evaluate the tendency of the data and is intended to estimate differences in

the mean state of the wind field. The intent of observing this parameter is to determine if

wind products have a tendency to over/underestimate the WG measured wind speed.

Lastly, Weibull Probability Density Functions (PDF’s) were used to evaluate the

various wind products ability to describe the WG measured wind regime. This method

was similarly used by Liu et al. (2008) and Carvalho et al. (2013) in their assessment of

QuikSCAT and CCMP’s ability to characterize buoy measured wind regimes closest to

reality.

Schmidt et al. [14]

3. Results

3.1 In-Situ Comparison

Wind speeds measured by the WG are compared against the corresponding

gridded 10m surface winds available from various products. Initially, comparison was

made between the entireties of all datasets. After hourly and 6 hourly binning methods

were applied, the raw data are reduced to 264 collocated data points of each product

with the WG series (Figure 4).

In order to understand the degree of variance from the mean of each product, the

mean and standard deviation of all wind products per wind speed category provided in

Table 1. The mean for the varying wind speed categories increases with respect to wind

speed. Additionally, the magnitude of increase in the mean is fairly uniform between low

and medium wind speeds (on average an increase of 2-3 m.s-1). However this

magnitude varies greatly between medium and high wind speed categories for all wind

products. For instance, between the medium and high wind speed categories the mean

of the SW product increased on the order of 1-2 m.s-1, while the other wind products

increased on the order of 5-7 m.s-1. During the the CSIR2 WG time series in 2015 the

SW, ECMWF, and CFSv2 products underestimated the mean state while the NCEPII

product overestimated the mean state. All wind products, including NCEPII

underestimated the mean state of the winds recorded by the CSIR1 WG in 2015-2016.

The variance of the mean of the WG data, as depicted by the standard deviation, is

consistent at low and medium wind speeds, and increases slightly for high wind speeds.

The only wind product which depicted this trend was NCEPII. For all three time series,

the NCEPII low wind speed category had the lowest variance of all wind speed

categories, and the high wind speed category had the highest variance from the mean.

The SW product exhibited low variance with low and high speeds, while the greatest

variance from the mean was at medium wind speeds. The ECMWF product consistently

exhibited a trend of decreasing variance with an increase in wind speed. The CFSv2

product for two time series depicted low wind speeds exhibiting the highest variance,

however the most recent time series (CSIR1 2015-2016) depicted the opposite; low

wind speeds having the least variance from the mean.

Schmidt et al. [15]

Figure 4: Comparison of wind products with in situ WG data in black and upper and lower limits

of each product dotted lines of their respective colours. (a) WG CSIR1 [2013-2014]. (b) WG

CSIR2 [2015]. (c) WG CSIR1 [2015-2016].

Schmidt et al. [16]

Table 1: The mean and standard deviation of all wind products and WG data per wind speed

category, as defined in Section 2.3. All bold products indicate the closest match to the WG

data.

Dec2013

Jan2014

ALL

(m.s-1)

LOW

(WG ≤ 3.6 m.s-1)

MEDIUM

(3.6 ≤ WG ≤ 6.0 m.s-1)

HIGH

(WG ≥ 6.0 m.s-1)

CSIR1 4.77 ± 1.76 2.46 ± 0.68 4.83 ± 0.68 6.98 ± 0.87

SW 8.41 ± 2.79 6.12 ± 3.52 9.01 ± 2.01 9.49 ± 1.92

CFSv2 9.19 ± 3.78 5.54 ± 3.52 9.22 ± 2.43 12.78 ± 2.55

ECMWF 8.99 ± 3.41 6.77 ± 3.88 9.62 ± 2.92 9.93 ± 2.84

NCEPII 10.25 ± 4.43 6.35 ± 3.57 10.68 ± 3.51 13.29 ± 4.07

July2015

Nov2015

ALL

(m.s-1)

LOW

(WG ≤ 7.3 m.s-1)

MEDIUM

(7.3 ≤ WG ≤ 12.4 m.s-1)

HIGH

(WG ≥ 12.4 m.s-1)

CSIR2 9.98 ± 3.76 5.21 ± 1.32 9.93 ± 1.46 14.83 ± 2.03

SW 9.63 ± 3.24 8.42 ± 3.06 9.51 ± 3.37 11.10 ± 2.49

CFSv2 9.80 ± 3.97 6.30 ± 3.04 9.51 ± 2.86 13.86 ± 2.93

ECMWF 9.58 ± 3.64 6.25 ± 2.92 9.39 ± 2.57 13.26 ± 2.57

NCEPII 11.11 ± 5.03 7.39 ± 3.81 10.52 ± 3.63 16.00 ± 4.70

Dec2015

Feb2016

ALL

(m.s-1)

LOW

(WG ≤ 9.0 m.s-1)

MEDIUM

(9.0 ≤ WG ≤ 16 m.s-1)

HIGH

(WG ≥ 16 m.s-1)

CSIR1 12.50 ± 4.79 6.76 ± 1.84 12.20 ± 2.06 18.82 ± 2.53

SW 8.57 ± 3.14 7.36 ± 2.93 8.67 ± 3.14 9.59 ± 2.99

CFSv2 9.39 ± 4.05 5.17 ± 2.16 8.96 ± 2.39 14.50 ± 2.32

ECMWF 9.21 ± 3.90 4.95 ± 2.28 8.88 ± 2.16 14.12 ± 2.03

NCEPII 11.15 ± 4.97 6.72 ± 2.70 10.67 ± 2.80 17.28 ± 3.41

Further statistical comparison was conducted using parameters such as root

mean square error, mean bias, and the correlation coefficient (Table 2). Globally, there

is a tendency for satellite product wind speed errors to be greater in the presence of low

and high winds (Carvalho et al., 2013). As shown in Table 2, when comparing

observations by WG CSIR1 with the other wind products, there is a clear increasing

trend with respect to increasing wind speed. Only the SW product exhibited a decrease

Schmidt et al. [17]

in error with respect to increasing wind speed. This would indicate that the ECMWF,

CFSv2, and NCEPII wind products are more effective at representing low – medium

wind speeds of the wind fields during the months of December – February. The

observations made by WG CSIR2 during the months of July – November indicate the

opposite trend. The ECMWF, CFSv2, and NCEPII products all exhibit a decreasing

error with increasing wind speed. Overall, the lowest error depicted varies, but is most

commonly exhibited by the NCEPII and ECMWF wind products.

The bias exhibited by all wind products with respect to the data collected by the

WG ranged from as low as 0.18 m.s-1 (CFSv2 vs CSIR2 2015 all wind speeds) to as

high as -9.23 m.s-1 (SW vs CSIR1 2015-2016 high wind speeds). The trends of the bias

are also depicted in Figure 5, which compares the wind speed residuals among the

wind products. For each time series (ranging seasonally), varying products best

represent the temporal variability of different wind speed categories. For the all wind

speed and medium wind speed categories, the ECMWF product had the highest

correlation coefficients with respect to the WG in situ data. ECMWF exhibited the

following correlation coefficients for the all wind speed category: 0.75, 0.76 and 0.93,

and the following coefficients for the medium wind speed category: 0.40, 0.40, and 0.81.

Low wind speeds are best represented by the CFSv2 product, with correlation

coefficients of 0.56, 0.21, and 0.58 respectively.

Table 2: Statistics of the comparison between in-situ WG and products wind speed error per

wind speed category. This includes the root mean square error, bias, and correlation

coefficients of the differences observed in the comparison of the in- situ measurements and the

wind products. The bold values highlight the product providing the closest match with the WG

data (i.e. the smallest value of the error or bias and the highest value of the correlation

coefficient). Correlation coefficients highlighted in green exhibit a confidence interval >99%,

yellow exhibit a confidence interval between 80% and 99%, and red exhibit a confidence interval

less than 80%.

Schmidt et al. [18]

Nov2013

Feb2014

CSIR1

vs.

ALL

(m.s-1)

LOW

(WG ≤ 3.6 m.s-1)

MEDIUM

(3.6 ≤ WG ≤ 6 m.s-1)

HIGH

(WG ≥ 6 m.s-1) N

RMSE

SW 4.40 4.95 4.64 3.11 264

CFSv2 5.17 4.45 4.93 6.22 268

ECMWF 4.84 4.50 4.68 5.45 268

NCEPII 6.58 5.17 6.75 7.43 268

BIAS

SW +3.64 +3.66 +4.19 +2.51 264

CFSv2 +4.41 +3.07 +4.38 +5.82 268

ECMWF +4.20 +3.25 +4.19 +5.17 268

NCEPII +5.47 +3.87 +5.84 +6.32 268

R^2

SW 0.48 0.32 0.16 0.32 264

CFSv2 0.76 0.56 0.34 0.54 268

ECMWF 0.75 0.52 0.40 0.60 268

NCEPII 0.60 0.27 0.26 0.27 268

July2015

Nov2015

CSIR2

vs

ALL

(m.s-1)

LOW

(WG ≤ 7 m.s-1)

MEDIUM

(7 ≤ WG ≤ 12 m.s-1)

HIGH

(WG ≥ 12 m.s-1) N

RMSE

SW 4.25 4.58 3.64 4.95 250

CFSv2 2.81 3.22 2.76 2.47 259

ECMWF 2.62 3.14 2.45 2.35 259

NCEPII 3.73 4.47 3.34 3.64 259

BIAS

SW -0.42 +3.14 -0.49 -3.80 250

CFSv2 -0.18 +1.09 -0.43 -0.97 259

ECMWF -0.40 +1.04 -0.54 -1.57 259

NCEPII +1.13 +2.18 -0.58 +1.17 259

R^2

SW 0.27 -0.03 0.05 0.01 250

CFSv2 0.74 0.21 0.34 0.63 259

ECMWF 0.76 0.17 0.40 0.73 259

NCEPII 0.71 0.07 0.42 0.74 259

Table 2 continued on next page.

Schmidt et al. [19]

Dec2015

Feb2016

CSIR1

vs

ALL

(m.s-1)

LOW

(WG ≤ 9 m.s-1)

MEDIUM

(9 ≤ WG ≤ 16 m.s-1)

HIGH

(WG ≥ 16 m.s-1) N

RMSE

SW 6.37 3.23 5.24 9.84 243

CFSv2 3.58 2.39 3.51 4.57 245

ECMWF 3.70 2.43 3.42 5.06 217

NCEPII 2.52 2.75 2.40 2.50 245

BIAS

SW -3.92 +0.60 -3.53 -9.23 243

CFSv2 -3.06 -1.55 -3.19 -4.32 245

ECMWF -3.16 -1.55 -3.16 -4.78 217

NCEPII -1.31 -0.75 -1.47 -1.54 245

R^2

SW 0.25 0.16 -0.08 0.24 243

CFSv2 0.93 0.58 0.79 0.81 245

ECMWF 0.93 0.59 0.81 0.77 217

NCEPII 0.90 0.35 0.73 0.82 245

Table 2 Continued

Figure 5 shows the correlation of the wind products with the in situ data collected

by WG CSIR1 from December 2013 - February 2014. Even though the SW product

exhibited the lowest errors and bias, the ECMWF and CFSv2 products had the best

correlation with the in situ data. However the low wind speeds of all wind products are

overestimated significantly which produces a skewed overall correlation. The NCEPII

product exhibited a lower correlation yet was more consistent across all wind speed

categories. Overall, the ECMWF product best represented the temporal variability of the

wind field during this time series.

Figure 6 shows the correlation of the wind products with the in situ data collected

by WG CSIR2 from July 2015 to November 2015. The SW product performed poorly

across all wind speed categories. All products poorly represented low wind speeds,

which is also shown in Table 2. CFSv2 and ECMWF similarly represented the medium

wind speeds, while NCEPII outperformed all other products in this category while

Schmidt et al. [20]

Figure 5: Comparison of wind products with in situ [CSIR1 WG] data collected December 2013

– February 2014 (austral summer). Wind speed categories are determined using the lower and

upper quartiles of the in situ data. Relationships for each wind speed category are shown using

the colored lines.

significantly overestimating the magnitude of high winds. CFSv2 and ECMWF similarly

best represent the high wind speed category and in general best represent the temporal

variability of the wind field during this time series.

Figure 7 shows the correlation of the wind products with the in situ data collected

by WG CSIR1 from December 2015 to February 2016. Similar to the CSIR2 time series,

the SW product performed poorly across all wind speed categories. ECMWF and

CFSv2 best represented low wind speeds, while NCEPII best represented medium and

Schmidt et al. [21]

Figure 6: Comparison of wind products with in situ [CSIR2 WG] data collected July 2015 –

November 2015. Wind speed categories are determined using the lower and upper quartiles of

the in situ data. Trendlines for each wind speed category are shown using the colored lines.

high wind speeds. Overall, CFSv2 performed most consistently and as such, best

represents the temporal variability of the wind field during this time series.

The residuals (or anomalies) are the differences between the expected and

measured wind speeds. As depicted in Figure 8, all wind products except SW show an

increasing trend in residuals with respect to an increasing wind speed. This is indicative

of low bias with low wind speed, and high bias with high wind speed. The SW product

depicted the opposite trend, with high bias attributed to low wind speeds, and lower bias

Schmidt et al. [22]

to high wind speeds. Figure 9 shows the residuals of the varying wind products

compared to the in situ data [WG CSIR2] collected from July 2015 to November 2015.

Figure 7: Comparison of wind products with in situ [CSIR1 WG] data collected December 2015

– February 2016 (austral summer). Wind speed categories are determined using the lower and

upper quartiles of the in situ data. Trendlines for each wind speed category are shown using the

colored lines.

The SW product still depicted a strong negative trend in bias with respect to

increasing wind speed. However, all other wind products also depicted a negative trend.

Most interestingly, the CFSv2, ECMWF, and NCEPII show increased positive residuals

for low and high wind speeds, but reduced residuals for medium wind speeds. In this

time series [CSIR2 July2015- Nov2015], the ECMWF product had the lowest residuals

Schmidt et al. [23]

for medium and high wind speeds. Figure 10 shows the residuals of the varying wind

products compared to the in situ data [WG CSIR1] collected from December 2015 to

February 2016. All products display similar trends seen from Figure 9. Again, NCEPII

product showed positive residuals for low and high wind speeds, and negative residuals

for medium wind speeds. For this time series [CSIR1 Dec2015- Feb2016], the CFSv2

and ECMWF products similarly had the lowest residuals for all wind speed categories.

Figure 8: Wind speed residuals (wind product minus in situ data) for WG CSIR 1 time series

between December 2013 – February 2014. Linear fit is displayed in gray.

In order to assess which of the wind products offers a characterization of the WG

wind regime closest to reality, a Weibull PDF is used. The PDFs of each wind product

are shown in Figure 11. For the CSIR1 time series which spanned from Dec 2015 –

Feb2014, the Sea Winds product showed better performance for the wind speed

Schmidt et al. [24]

temporal variability (RMSE, R^2), along with low errors in the estimation of the wind

speed frequency distribution due to its lower biases for the wind speed. As such, the

SW PDF curve is closest to the WG PDF curve. All wind products significantly

overestimate the frequency of wind speeds greater than 6 m.s-1 and underestimate the

frequencies of wind speeds below 4-6 m.s-1. This behavior is portrayed by a shift in the

Weibull curves to the right side of the wind speed axis for all products with respect to

the WG CSIR1 [December 2014 – November 2014] distribution curve. For the second

time series [CSIR2 July 2015- November 2015], the ECMWF product is almost

Figure 9: Wind speed residuals (wind product minus in situ data) for WG CSIR 2 time series

between July 2015 – November 2015. Linear fit is displayed in gray.

Schmidt et al. [25]

Figure 10: Wind speed residuals (wind product minus in situ data) for WG CSIR 1 time series

between December 2015 – February 2016. Linear fit is displayed in gray.

identical to the ECMWF frequency distribution for wind speeds. Lastly, the wind speed

frequency distribution of the NCEPII product is almost identical to the of the third time

series [CSIR1 December 2015- February 2016]. The PDFs shown in Figure 11 show

interesting trends. While the in situ wind speed frequency distribution varies significantly

with respect to time, the wind products used in this comparison do not. When compared

to the other products, the SW product systematically represented the wind fields having

higher frequencies of low winds. The ECMWF and CFSv2 products had a more even

frequency distribution across all wind speed categories, while the NCEPII product

tended to introduce a lower variability in the wind speed frequency distribution. It is also

interesting to note that the higher the maximum wind speed of the wind product, the

greater the variability and the more even the distribution of wind speeds.

Schmidt et al. [26]

Figure 11: Wind speed frequency distribution (Weibull PDF) for all products (SW, ECMWF,

CFSv2, and NCEPII) from all time series of in situ wind collection by WGs CSIR1 and CSIR2.

5. Discussion

The intent of the study was to identify which product most accurately represents

the variability of the wind field in the Southern Ocean in comparison to in situ. Previous

studies have concluded that globally, satellite products, such as QuikSCAT, OSCAT,

and CCMP tend to overestimate wind speed exponentially with respect to wind intensity

(Carvalho et al., 2013), and have the highest margins of error during low and high wind

events (Sudha et al., 2013; Alvarez et al., 2014; Jayaram et al., 2014). However,

because the systems have not been rigorously inter-calibrated, any comparison

between satellite wind data, in situ data, or numerical weather predictions will only give

relative information on accuracy (Vogelzang et al., 2012).

5.1 Product Performance

Based on the results from Table 2 and as seen in Figures 5-11, performance of

each product varied with respect to time. For the first time series [December 2013-

February 2014], while the SW product best represented the mean wind state with the

lowest errors (Table 2), ECMWF had the highest and most consistent correlation

coefficients across all wind speed categories. Figure 5 and 8 show that the ECMWF

Schmidt et al. [27]

product, albeit with high residuals with respect to the SW product, best represents the

temporal variability of the wind field for the first time series. During the second time

series [July 2015 – November 2015], the ECMWF product was consistent across all

wind speeds at best representing the mean wind state. The CFSv2 product was a close

second and compared to the WG data, had the most similar overall mean wind speed

with respect to the in situ data. The WG all wind speed average was 9.98 m.s-1 while

the CFSv2 all wind speed average was 9.80 m.s-1 (Table 1). ECMWF all wind speed

average was slightly less (9.58 m.s-1). The SW product significantly underestimated all

wind speeds, which is why it appears to be a good representative of the all wind speed

average for the second time series. As shown in Figure 2, ECMWF had the lowest

errors and was a close second to CFSv2 in the lowest bias with respect to the WG

winds. Regarding the correlation of the products with the July 2015 – November 2015

WG series, ECMWF had the highest overall correlation and was marginally second to

the CFSv2 and NCEPII products for the other wind speed categories. It should be noted

that all products performed very poorly in their representation of the temporal variability

of low winds (Table 2, Figure 6). Additionally, as depicted in Figure 6, the NCEPII

product derives its high correlation from a gross overestimation of high wind speeds and

underestimation of low wind speeds. For these reasons as well as due to its low errors

and bias and consistently high correlation across all wind fields, the ECMWF product

best represented both the mean wind state and temporal variability of the wind field

sampled from July 2015 – Nov 2015. The final time series [December 2015 – February

2016] was best represented by two products; ECMWF and NCEPII. During this time

series, all wind products underestimated the wind fields observed by the WG CSIR1.

NCEPII performed best in representing the mean wind state. It also had the lowest

errors and biases in all categories except the low wind speed category. NCEPII was

only outperformed by ECMWF with regard to its correlation to the temporal variability of

the observed wind field. Even though ECMWF outperformed NCEPII in this category,

overall the NCEPII product performed best for the third time series [December 2015 –

February 2016].

Regarding the first time series, the overestimation of all wind products forces the

SW product to in some cases have the lowest residuals, even though it clearly does not

Schmidt et al. [28]

represent the wind fields as well as other wind products. For the following two time

series, the SW product grossly underestimated the wind fields and systematically had

poor correlation across all wind speeds; as such it is not considered to be a viable

product for use in fine-scale modelling. This can be attributed to the fact that while SW

data are available at a relatively high spatial resolution (0.25⁰ x 0.25⁰), it is a gridded

satellite product. This product uses a simple objective analysis method (spatial-

temporally weighted interpolation) to generate a gridded and blended product (Zhang,

2006). This method of blending satellite observations minimizes but does not completely

eliminate data gaps. As such, the high resolution of spatial and temporal sampling

conducted by WG technology may not be represented accurately. The high

performance of both the ECMWF and CFSv2 products can be attributed to the high

spatial resolution of the datasets (0.125⁰ x 0.125⁰, and 0.205⁰ x 0.204⁰ respectively).

Four of the ECMWF data grid locations fall either directly on or within a tenth of a

degree away from the circumference of the circular sampling pattern (Figure 2) of the

WGs for the three time series. The high bias of the NCEPII product can be attributed to

its low resolution, as the nearest point of reference to the majority of the in situ

measurements is roughly 100 kilometers away (see Figure 2). Overall, the ECMWF

product is best at consistently representing the temporal wind field variability.

For the first two time series, measures of error and bias for ECMWF and CFSv2

were lower than for NCEPII for all wind speed categories. However, for the third time

series the NCEPII product clearly outperforms all other products with regards to error

and bias. While the CFSv2 product is better than ECMWF at representing the mean

wind state of low and medium wind speeds, ECMWF is better at representing the

temporal variability of the wind field magnitude. Upon closer inspection of the

performance of the NCEPII product, it best represented mean wind state in that it

exhibited similar trends in standard deviation across wind speed categories. For

example, in the first time series NCEPII had similar deviation from the mean of the low

and medium wind speeds, and a higher deviation for the high wind speed category. A

similar trend is observed regarding the deviation from the mean of the in situ data

across wind speed categories. This similar trend observed between the WG and

NCEPII data is reflected across all three time series, while all other products exhibit

Schmidt et al. [29]

opposing trends in deviation from the mean. Due to these factors, the NCEPII product is

best at representing the mean wind state and deviation from the mean state, especially

during times of high mean winds. However this product must be used with caution, as

Figures 9 and 10 show that NCEPII derives its overall good correlation through its

overestimation of low and high wind speeds.

The ability of a product to best represent the wind speed frequency distribution

also varied over the different time series. Findings by Pickett et al. (2003) and Tang et

al. (2004) suggest that satellites measures higher frequencies of strong winds and lower

frequencies of weak winds. It is curious to see that the frequency distribution of the in

situ wind speeds varies among times series, while the frequency distribution of the wind

product wind speeds remains fairly constant across the three time series. As such, a

different product for each time series best represented the frequency distribution of

observed wind speeds. The SW product consistently showed higher frequencies of low

wind speeds, while ECMWF and CFSv2 similar showed a fairly uniform distribution of

wind speed frequencies. NCEPII consistently showed a more broad distribution of

speed frequencies. However, the similarity of the WG and SW PDFs during the time

frame of the first sampling period is not indicative of a better ability of the SW to

represent the wind speed frequency spectrum. In fact, it is indicative of a gross

overestimation of wind speed by all wind products, which forces the product with the

lowest bias to have the most similar wind speed frequency distribution. During the

second time series, the average mean wind state increased by more than 5 m.s-1, which

caused the product which best represents low and medium wind speeds to accurately

represent the speed frequency distribution – in this case, ECMWF. There is also an

increase in the mean wind speed of the third time series. This increase of more than 3

m.s-1 (with respect to the second time series) caused the NCEPII product to best

represent the frequency distribution of the wind speed.

Schmidt et al. [30]

5.2 Sampling Period Sensitivities

Satellite derived wind data as well as reanalysis and model products can be very

different from in-situ anemometer data in that they tend to represent synoptic winds. A

study by Atlas et al. (1999) showed that satellite products are able to detect mesoscale

features, however this representation is limited due to the fact that satellite and

numerical weather prediction products are of lower spatial and temporal resolution

(Carvalho et al., 2013). Anemometer wind data is closer to a temporal average of

instantaneous moment, while satellite winds represent a spatial average of

instantaneous moments (Pansieri et al., 2010). Studies have shown that in cases where

a limited amount of data existed, such as from a single satellite, any interpolations/

extrapolations were inaccurate in representing the true strength of mesoscale events

(Isaksen and Stoffelen, 2000). Bearing this in mind, over most SW grids the number of

observational input data points is over 40 (Zhang, 2006). As such a multiple satellite

blending scheme can be advantageous in regions of lower mean wind states in the

Southern Ocean. ECMWF is a reanalysis product, however and uses a sequential data

assimilation scheme, advancing forward in time using 12 hourly analysis cycles (Dee et

al., 2011). A temporally short-scale assimilation scheme could give ECMWF an

advantage over other reanalysis products such as NCEPII in representing the temporal

variability on shorter temporal scales, as is the scale of this study. CFSv2 is a fully

coupled ocean–land–atmosphere dynamical model using a global ocean data

assimilation system (GODAS) operational at NCEP (Saha et al., 2014). This product is

unique in that it makes retrospective forecasts to calibrate operational subsequent real

time sub-seasonal and seasonal predictions (Saha et al., 2014). This calibration

scheme could be why CFSv2 is a more rounded product and does well in representing

both the temporal variability and the mean wind state. The NCEPII product is a model

simulation as well (forecast/hindcast reanalysis) however the data assimilation scheme

is slightly different to ECMWF which is forced to conform to observational input. Using

4-D data assimilation methodology, observational input from ships, satellite, radiosonde,

aircraft, and meteorological station observations are incorporated into NCEPII (Kalnay

et al., 1999). An assimilation methodology with high input gives NCEPII ability in

Schmidt et al. [31]

representing regional climate dynamics (Kanamitsu et al., 2002), even though it is of the

coarsest spatial resolution compared to all other wind products.

5.3 Implications:

Ocean circulation modelling, climate change estimation, and numerical weather

prediction is heavily dependent on the accuracy of input. Upper ocean dynamics are

heavily linked to atmospheric variability in the Southern Ocean. As mentioned

previously, the mid-latitude regions of the Southern Ocean are host to the strongest

wind fields at the ocean surface. If these winds are not accurately represented, models

will incorrectly simulate sea surface exchanges of heat and moisture flux, lateral

advection, stratification and mixed layer dynamics (frontal formation and instabilities),

and Ekman pumping and transport for example. However, the present study among

others shows that in many cases the wind fields are not accurately represented.

Upper ocean models which include satellite-derived wind speed input will not

accurately represent variance in MLD and SST. Ocean circulation models have shown

the sensitivity of the MLD to the gustiness in the wind (Lee et al., 2008). Turbulent

mixing and subsequent buoyancy forcing caused by sea surface winds strongly

influences the MLD. As the MLD shallows, high wind events are observed to have a

greater impact on the deepening of the mixed layer (Carranza and Gille, 2015).

Carranza and Gille (2015) observed that during high wind events, water from below the

mixed layer is entrained in the upper ocean as the mixed layer drops. Swart et al.

(2014) claim that this variability of the MLD is driven by synoptic storms (4-9 days)

where turbulent mixing deepens the surface mixed layer, while quiescent wind episodes

draw the MLD up rapidly. This wind-driven mixing also has an effect on SST, where cold

water from below the MLD is entrained in the upper ocean. This in turn influences the

mixed layer heat budget (Bonekamp et al., 1999) and produces cold SSTs. The study

by Carranza and Gille (2015) found a statistically significant negative correlation

between wind speed and SST anomalies, and that wind speed alone can explain as

much as 80% of the variance observed in SSTs.

There are many important biological applications of wind data which require high

temporal and spatial quality. Knox (2007) observed the MLD to be the shallowest in the

Schmidt et al. [32]

summer at less than 90 meters. This is close to the bottom of the euphotic zone;

beyond which photosynthesis can no longer be supported by solar radiation. Deepening

of the MLD either facilitates phytoplankton growth via a greater input of nutrients, or

brings the chlorophyll (Chl-a) maximum up to a shallower depth. Since the mixed layer

is deepened by strong winds, high wind speeds and high Chl-a are generally positively

correlated (Carranza and Gille, 2015). Carranza and Gille (2015) observe on short time

scales (days to weeks), nutrient and chlorophyll entrainment to the mixed layer follows

wind variability with a time lag of just a few hours. Areas of the Southern Ocean which

showed positive correlations between wind speed and Chl-a tended to persist on

smaller, weekly timescales and were associated with synoptic storms. However, on

longer time scales the Southern Ocean does not exhibit a positive correlation between

high wind and high Chl-a. On monthly to seasonal timescales, correlations between

wind speed and Chl-a tend to be negative, especially over regions with deep winter

mixed layers (Carranza and Gille, 2015). Their rationale for this is that wind-induced

mixing action is strong enough to force the Chl-a out of the euphotic zone. In high

latitudes of the Southern Ocean, the seasonality of the mixed-layer depth is

fundamental for explaining seasonal chlorophyll bloom characteristics (such as its

initiation and decline) (Sverdrup, 1953). Thomalla et al. (2011) investigated the

seasonality of Chl-a in the Southern Ocean. Along with Fauchereau et al. (2011), this

study concluded that variability in Chl-a bloom events on a sub-seasonal scale is a

result of highly-variable wind stress in addition to buoyant forcing induced stabilization

of the upper water column via mesoscale to sub-mesoscale dynamics. A similar study

by Swart et al. (2014) attributes gaps in biogeochemistry/ phytoplankton distribution and

abundance to a lack of good understanding of the wind field in the Southern Ocean.

Swart et al. (2014) emphasize that poor knowledge of the wind field could potentially

inhibit an understanding of physical and biogeochemical interactions in the region.

Thus, quantifying the impact of wind on stratification is needed before the seasonal

scale variability of Chl-a can be modelled. The physical forcing mechanisms responsible

for enhanced Chl-a are still unclear (Thomalla et al., 2011).

Carranza and Gille (2015) further investigated the impacts of winds and

buoyancy forcing on Chl-a abundance in the Southern Ocean. With respect to their sea

Schmidt et al. [33]

surface wind analysis, Cross Calibrated Multi-Platform (CCMP) data provided gridded

surface wind fields at a 6-hour temporal resolution. CCMP wind data is different from

reanalysis data in that it falls under the class of blended scatterometer data, where

multiple datasets from different platforms are blended to fill in spatial and temporal gaps

(similar to SW). Global studies by Kahru et al. (2010) have shown that sea surface

winds, such as the ones represented by CCMP data, have a measurable influence on

Chl-a variability. Klein and Coste (1984) find that entrainment of chlorophyll at the mixed

layer is directly related to variability of wind speed. Carranza and Gille (2015) use the

following formula to calculate the entrainment rate at the base of the mixed layer

following the methods of De Szoeke (1980); Musgrave et al. (1988); and Close and

Goosse. (2013):

𝑊𝑒 = 𝜕ℎ

𝜕𝑡+ 𝑊𝐸𝑘(ℎ) (7)

Turbulent (𝜕ℎ

𝜕𝑡) and non-turbulent (𝑊𝐸𝑘) processes which facilitate entrainment are

represented in this equation. The turbulent parameter is modified by wind induced

mixing and will vary with variability of the wind field (Carranza and Gille, 2015).

Additionally, the vertical displacement of the mixed layer is represented by the non-

turbulent parameter. This displacement is a result of divergences in the mixed layer

driven solely by wind stress curl (τ) (Carranza and Gille, 2015). This is shown in the

calculation of the non-turbulent parameter (𝑊𝐸𝑘 = ∇ × τ

𝜌𝑓). If sea surface wind fields are

not correctly represented by satellite data (in this case CCMP data), both turbulent and

non- turbulent processes will consequently be misrepresented in the calculation of

entrainment.

A better representation of variability in the upper ocean dynamics such as winds,

SST, and MLD will yield a greater understanding of trends in primary productivity.

Findings by Nicholson et al. (submitted 2016) emphasize the need for high resolution

wind input in biogeochemical models. They created a model with idealized storm-driven

vertical mixing scenarios in an attempt to show how summer primary production can be

sustained over seasonal time scales. The premise of their study is that inertial (wind)

Schmidt et al. [34]

driven subsurface mixing beneath the surface-mixed layer makes significant

contributions to the refurbishment of the subsurface reservoir of iron which available to

be entrained by deepening mixed-layers. As such, Nicholson et al. (submitted 2016)

suggest that the intra-seasonal mixed layer perturbations linked to mid-latitude storms

may offer relief from iron limitation in the summer bloom period. If wind products are of

low resolution and miss synoptic storms, it is possible that correlations between wind

speed and Chl-a during summer blooms are inaccurate. Swart et al. (2014) use high

resolution glider observations (similar to those used in the present study) in the Sub-

Antarctic Zone (SAZ) and provide evidence to suggest that transient wind events drive

intra-seasonal variability in Chl-a. During sustained blooms, their observations of

collocated wind stress with MLD yielded a correlation coefficient of greater than 0.40

(Swart et al. 2014). Such observations from gliders emphasizes that the variability of

wind stress will drive the MLD variability, which has since been observed to correlate

with Chl-a. There is need to examine these relationships further whilst considering the

shortcomings of NWP models as well as blended satellite wind products ability to

represent mesoscale cyclones (Yuan et al. 2009).

A hemispheric scale process that has important climatic implications is the

Southern Annular Mode (SAM). The SAM reflects the north-south transfer of

atmospheric mass and pressure between the Southern Hemisphere mid- and high

latitudes and hence strongly influences the variability in the westerly winds over the

Southern Ocean. Thompson et al. (2011) show that changes in Antarctic ozone levels

has driven poleward displacement of the SAM, which results in summertime Southern

Hemisphere climate change. In the summer, over high latitudes there is an acceleration

of the prevailing eastward wind which drives eastward directional departures from the

long term mean (Thompson et al., 2011). Thompson et al. (2011) state that anomalous

eastward flow at high latitudes results in increased Ekman transport over much of the

Southern Ocean and upwelling south of 60⁰S. They observe the opposite at middle

latitudes in the Southern Ocean, where the southward movement of the SAM results in

a deceleration of the prevailing eastward wind. This deceleration results in westward

anomalies of surface flow, which consequently causes downwelling along with variability

in temperature and precipitation in the midlatitude region. Since the SAM exists on all

Schmidt et al. [35]

time scales from synoptic, intraseasonal to interannual and longer and is fundamental to

Southern Hemisphere climate, it is important to choose a wind product that represents

this mode accurately.

Variability in the air-sea interface not only affects the physical oceanography of a

region, but also the biogeochemical cycling of gaseous compounds such as CO2. A

number of studies (Wanninkhof 1992, 1999, 2009; Nightingale et al., 2000; McGillis et

al., 2001; Ho et al., 2006) have developed methodology to determine gas transfer rates,

processes which control them, and quantifying flux. These studies (Wanninkhof 1992;

Sweeney et al., 2007; Wanninkhof 2009) often find a squared or cubic relationship

between observed fluxes and wind speed. This means that an incorrect

parameterization of wind speeds could lead to significant errors in simulations of global

CO2 flux. In their study, Takahashi et al. (2009) state that due to the dependence of the

sea–air gas transfer piston velocity and the transfer coefficient on wind speed, with a

constant scaling factor a systematic increase in wind speed by 10% could consequently

increase the global CO2 flux by as much as 20%. Wanninkhof (2009) declare that as

such, much methodology to determining gaseous exchange is still in developmental

stages due to significant biases in different wind speed products.

5.4 Caveats and Future Work

Seasonal variability of the mean wind state was not a factor which was

considered in the statistical analysis, but could possibly explain why some wind

products performed better than others at different times of the year. However, the first

and third time series were measured over the same season (December – February) and

differences in the results of the statistical analysis are significant enough to debunk this

theory. Thus this discrepancy may be a result of poor sensor performance specific to

the mission. Either the sensor may not be calibrated properly, or salt from sea spray

could have plugged the vent on the Airmar sensor, preventing the mylar pressure from

stabilizing and thus impacting the measurements taken by the wind sensor. This theory

is under investigation and will be confirmed if pressure levels measured during the first

time series indicate there was in fact a blocked vent. Additional further study should

investigate the seasonal, annual and interannual variability of the wind field in the study

Schmidt et al. [36]

region prior to analysis. Seasons or years where a lower mean wind speed state (less

than 6 m.s-1) is prevalent could cause a product such as ECMWF or CFSv2 to be best

at representing the wind fields. Seasons or years where the mean wind state is high

(greater than 10 m.s-1) could allow a product such as NCEPII to be best at representing

wind fields. Furthermore, highly variable residuals were observed for most wind

products and for every wind speed category. Due to the small profile of the WG in the

water, it is possible that periods of the year with increased gustiness (extreme wind

events), high seas and wave sheltering could lead to inaccuracies in measurements.

The WG samples wind speed at an instantaneous location, while scatterometers

measure wind relative to a moving ocean surface (Sudha and Rao, 2013). Dickinson et

al. (2001) observed that during alignment (both in same and opposite direction) of both

wind and surface ocean current, scatterometer data tend to be under/overestimated.

This could partially explain the bias observed by all products. Furthermore, there

appears to be a signal - noise ratio. From Table 1, the highest standard deviation of the

wind products is exhibited with low wind speeds (except for NCEPII). This could be

indicative of poor scatterometer ability to detect weak wind speeds. It is possible that

the small detection limits of the sensor are ineffective in measuring backscatter from

weak wind speeds. Further study should be conducted to determine if there is a

relationship between weak wind variability and the backscatter detection sensitivity of

scatterometer sensors. Additionally, the stability of the atmospheric boundary layer

above the sea surface can potentially affect the variability of wind speeds of satellite-

derived products. At a fixed height of 10m above sea surface, the generation of

smallscale waves via wind stress varies with respect to atmospheric stability (Ebuchi et

al., 2002). This in turn affects radar backscatter which is used to derive wind speed. In

addition to this discrepancy in satellite product data, the vertical transformation of WG

data lacked atmospheric stability input. It is possible that data transformation could be of

0.7 m.s-1 greater magnitude (Singh et al., 2013). For future studies, atmospheric stability

information should be included when possible to reduce bias. This can come from the

use of WG technology, which provides a means to achieve continuous high resolution

observations of atmospheric stability in data sparse regions like the Southern Ocean.

Schmidt et al. [37]

6. Summary

Predictions of heat, moisture, and momentum exchanges between the ocean and

the atmosphere are inaccurate and the Southern Ocean is a region where there is large

uncertainty in many model estimates. Detailed observations such as ones from Wave

Glider technology are crucial for future work in that they provide valuable insight to the

sensitivity and variability of the ocean-atmosphere physical and biological response. In

this study it was found that the NCEPII product best represented the mean wind state,

while ECMWF most consistently represented the temporal wind field variability. There is

a great need to correctly parameterize transient, synoptic scale storm events to depict

decadal and century scale changes in ocean-atmosphere relationships. Studies suggest

that with changes in the SAM and subsequent higher winds in the higher latitudes,

primary productivity will increase and enhance atmospheric CO2 drawdown. Changes to

the biological pump will have a global impact on climate and it is imperative that wind

products which correctly parameterize the wind field are used in order to model and

predict climate variability and ecosystem functioning.

Schmidt et al. [38]

ACKNOWLEDGEMENTS

The Wave Glider wind data were provided by the Southern Ocean Carbon & Climate

Observatory, Council for Science and Industrial Research (CSIR) from a series of three

SOSCEx missions. The SeaWinds blended altimetry data, CFS climate forecast, and

NCEPII reanalysis wind product was downloaded from NOAA’s National Centers for

Environmental Information (NOAA/NCDC). The ECMWF reanalysis wind product was

downloaded from the European Centre for Medium-Range Weather Forecasts open

access database. I would like to thank University of Cape Town Applied Marine Science

course convener Dr. Cecile Reed for support throughout the duration of the program.

Thanks to PhD candidate Luke Gregor for his assistance with understanding the CO2

fluxes of the Southern Ocean.

Schmidt et al. [39]

REFERENCES:

Alvarez, I., M. Gomez-Gesteira, M. deCastro, and D. Carvalho. (2014). Comparison of different

wind products and buoy wind data with seasonality and interannual climate variability in the

southern Bay of Biscay (2000–2009). Deep-Sea Research II. 106, 38-48. DOI: 10.1016/

j.dsr2.2013.09.028

Andrews, P. L. and R. S. Bell. (1998). Optimizing the United Kingdom Meteorological Office

data assimilation for ERS-1 scatterometer winds. Monthly Weather Review. 126:3, 736-746.

DOI: 10.1175/1520-0493

Atlas, R., S. C. Bloom, R.N. Hoffman, E. Brin, J. Ardizzone, J. Terry, D. Bungato, and J.C.

Jusem. (1999). Geophysical validation of NSCAT winds using atmospheric data and

analyses. Journal of Geophysical Research. 104:C5, 11, 405-411, 424.

Bentamy, A., Y. Quilfen, and P. Flament. (2002). Scatterometer wind fields: A new release over

the decade 1991-2001. Canadian Journal of Remote Sensing. 28:3, 431-449

Bentamy, A., D. Croize-Fillon, and C. Perigaud. (2008). Characterization of ASCAT

measurements based on buoy and QuikSCAT wind vector observations. Ocean Science. 4,

265–274.

Bentamy, A. and D. Croize-Fillon. (2012). Gridded surface wind fields from Metop/ASCAT

measurements. International Journal of Remote Sensing. 33:6, 1729-1754. DOI: 10.1080/

01431161.2011.600348

Bonekamp, H., A. Sterl, and G. J. Komen. (1999). Interannual variability in the Southern Ocean

from an ocean model forced by European Centre for Medium-Range Weather Forecasts

reanalysis fluxes. Journal of Geophysical Research. 104:C6, 13317-13331. DOI: 10.1029/

1999JC900052

Bourassa, M. A., D. G. Vincent, and W. L. Wood. (1999a). A flux parameterization including the

effects of capillary waves and sea state. Journal of Atmospheric Science. 56, 1123–1139

Bourassa, M. A., D. M. Legler, J. J. O’Brien, and S. R. Smith. (2003) SeaWinds validation with

research vessels. Journal of Geophysical Research. 108:C2, DOI: 10.1029/ 2001JC001028

Carranza, M. M., and S. T. Gille. (2015). Southern Ocean wind-driven entrainment enhances

satellite chlorophyll-a through the summer. Journal of Geophysical Research: Oceans. 120,

304–323. DOI: 10.1002/2014JC010203.

Schmidt et al. [40]

Carvalho, D., A. Rocha, M. Gomez-Gesteira, I. Alvarez, and C. Silva Santos. (2013).

Comparison between CCMP, QuikSCAT and buoy winds along the Iberian Peninsula coast.

Remote Sensing of the Environment. 137, 173-183. DOI: 10.1016/j.rse.2013.06.005.

Chakraborty, A. and R. Kumar (2013). Generation and validation of analysed wind vectors over

the global oceans. Remote Sensing Letters. 4:2, 114-122. DOI: 10.1080/ 2150704X. 2012.

701344

Chelton, D. B., M. G. Schlax, M. H. Freilich, and R. F. Milliff. (2004). Satellite Measurements

Reveal Persistent Small-Scale Features in Ocean Winds. Science. 303, 978-983. DOI:

10.1126/science.1091901

Close, S. E., and H. Goosse. (2013). Entrainment-driven modulation of Southern Ocean mixed

layer properties and sea ice variability in CMIP5 models. Journal of Geophysical Research:

Oceans. 118, 2811–2827. DOI:10.1002/jgrc.20226.

De Szoeke, R. (1980). On the effects of Horizontal Variability of Wind Stress on the Dynamics

of the Ocean Mixed Layer. Journal of Physical Oceanography. 10, 1439 – 1454. DOI:

10.1175/1520-0485

Dee D. P., Uppala S. M., Simmons A. J., Berrisford P. et al. (2011). The ERA-Interim

reanalysis: configuration and performance of the data assimilation system. Quarterly Journal

of the Royal Meteorological Society. 137, 553–597. DOI:10.1002/qj.828

Dickinson, S., K. A. Kelly, M. J. Caruso, and M. J. McPhaden, 2001: Comparisons between the

TAO buoy and NASA scatterometer wind vectors. Journal of Atmospheric Oceanic

Technology. 18, 799–806. DOI: 10.1175/1520-0426

Domingues, R., G. Goni, S. Swart, and S. Dong. (2014). Wind forced variability of the Antarctic

Circumpolar Current south of Africa between 1993 and 2010. Journal of Geophysical

Research: Oceans. 119, 1123–1145. DOI: 10.1002/2013JC008908.

Ebuchi, N., H. C. Graber, and M. J. Caruso. (2002). Evaluation of Wind Vectors Observed by

QuikSCAT/ Seawinds Using Ocean Buoy Data. Journal of Atmospheric and Oceanic

Technology. 19:12, 2049–2062. DOI: 10.1175/1520-0426 (2002) 0192.0.

Fauchereau, N., A. Tagliabue, L. Bopp, and P. M. S. Monteiro. (2011). The response of

phytoplankton biomass to transient mixing events in the Southern Ocean. Geophysical

Research Letters. 38:L17601. DOI: 10.1029/2011GL048498

Schmidt et al. [41]

Herrera, J. L., S. Piedracoba, R. A. Varela, G. Roson. (2005). Spatial analysis of the wind field

on the western coast of Galicia (NW Spain) from in situ measurements. Continental Shelf

Research. 25, 1728-1748. DOI:10.1016/j.csr.2005.06.001

Hilburn, K.A., M.A. Bourassa, and J.J. O’Brien. (2003) Development of scatterometer-derived

surface pressures for the Southern Ocean. Journal of Geophysical Research. 108:C7, 32-

44. DOI: 10.1029/2003JC001772.

Ho, D. T., C. S. Law, M. J. Smith, P. Schlosser, M. Harvey, and P. Hill. (2006). Measurements

of air-sea gas exchange at high wind speeds in the Southern Ocean: Implications for global

parameterizations. Geophysical Research Letters. 33:L16611. DOI:10.1029/2006GL026817

Isaksen, L., and A. Stoffelen. (2000). ERS-Scatterometer wind data impact on ECMWF’s

tropical cyclone forecasts. Transactions on Geoscience and Remote Sensing. 38, 1885-

1892. DOI: 10.1109/36.851771

Jayaram, C., T. Udaya-Bhaskar, D. Swain, E. Pattabhi, R. Rao, S. Bansal, D. Dutta and K.

Hanumantha Rao. (2014). Daily composite wind fields from Oceansat-2 scatterometer.

Remote Sensing Letters. 5:3, 258-267. DOI: 10.1080/2150704X.2014.898191

Kahru, M., S. T. Gille, R. Murtugudde, P. G. Strutton, M. Manzano-Sarabia, H. Wang, and B. G.

Mitchell. (2010). Global correlations between winds and ocean chlorophyll. Journal of

Geophysical research. 115:C12040. DOI: 10.1029/2010JC006500

Kalnay, E., M. Kanamitsu, R. Kistler, W. Collins, D. Deaven, L. Gandin, M. Iredell, S. Saha, G.

White, J. Woollen, Y. Zhu, A. Leetmaa, R. Reynolds, M. Chelliah, W. Ebisuzaki, W. Higgins,

J. Janowiak, K. C. Mo, C. Ropelewski, J. Wang, R. Jenne, D. Joseph. (1996). The

NCEP/NCAR 40-year reanalysis project. Bulletin of the American Meteorological Society.

77, 437-470. DOI: 10.1175/1520-0477.

Kanamitsu, M., W. Ebisuzaki, J. Woollen, S. K. Yang, J. J. Hnilo, M. Fiorino, and L. Potter.

(2002). NCEP-DOE AMIP-II Reanalysis (R-2). American Meteorological Society. DOI:

10.1175/BAMS-83-11-1631

Klein, P. and B. Coste. (1984). Effects of wind-stress variability on nutrient transport into the

mixed layer. Deep-Sea Research. 31, 21-37.

Knox, G. A. (2007). Biology of the Southern Ocean. 2nd edition. Boca Raton, FL : CRC

Press/Taylor & Francis

Schmidt et al. [42]

Lee, T., O. Wang, W. Tang, and W. T. Liu. (2008). Wind stress measurements from the

QuikSCAT-SeaWinds scatterometer tandem mission and the impact on an ocean model.

Journal of Geophysical Research. 113:C12019. DOI:10.1029/2008JC004855.

Liu, W. T., and W. Tang. (1996). Equivalent Neutral Wind. Jet Propulsion Laboratory

Publication 96-17.

Liu, W.T., W. Tang, and X. Xie. (2008). Wind power distribution over the ocean. Geophysical

Research Letters. 35:L13808. DOI: 10.1029/2008GL034172.

McGillis, W. R., J. B. Edson, J. D. Ware, J. W. H. Dacey, J. E. Hare, C. W. Fairall, and R.

Wanninkhof. (2001). Carbon Dioxide flux techniques performed during GasEx-98. Marine

Chemistry. 75, 267-280. DOI: S0304-4203 01 00042-1

Mears, C. A., D. K. Smith, and F. J. Wentz. (2001). Comparison of Special Sensor Microwave

Imager and buoy-measured wind speeds from 1987 to 1997. Journal of Geophysical

Research. 106:11, 719–11,729. DOI: 1999JC000097.0148-0227/01/1999JC000097

Monteiro, P. M. S., L. Gregor, M. Lévy, S. Maenner, C. L. Sabine, and S. Swart. (2015).

Intraseasonal variability linked to sampling alias in air-sea CO2 fluxes in the Southern

Ocean. Geophysical Research Letters. 42, DOI: 10.1002/2015GL066009.

Musgrave, D. L., J. Chou, and W. J. Jenkins. (1988). Application of a model of upper-ocean

physics for studying seasonal cycles of oxygen. Journal of Geophysical Research. 93:15,

679-15, 700. DOI: 10.1029/JC093iC12p15679

NCEP_Reanalysis 2 data provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA,

from their web site at http://www.esrl.noaa.gov/psd/

Nicholson, S. A., M. Levy, J. Llort, S. Swart, and P. M. S. Monteiro. (Submitted 2016). Storms

sustain summer primary productivity in the Sub-Antarctic Ocean via enhanced iron

entrainment.

Nightingale, P.D., Malin, G., Law, C.S., Watson, A.J., Liss, P.S., Liddicoat, M.I., Boutin, J. and

Upstill-Goddard, R.C. (2000). In situ evaluation of air-sea gas exchange parameterizations

using novel conservative and volatile tracers. Global Biogeochemical Cycles. 14. DOI:

10.1029/1999GB900091.

Schmidt et al. [43]

Patoux, J., X. Yuan, and C. Li. (2009). Satellite-based midlatitude cyclone statistics over the

Southern Ocean: 1. Scatterometer-derived pressure fields and storm tracking. Journal of

Geophysical Research. 114:D04105. DOI: 10.1029/2008JD010873.

Pickett, M. H., Tang, W., Rosenfeld, L. K., & Wash, C. H. (2003). QuikSCAT satellite

comparisons with nearshore buoy wind data off the U.S. West Coast. Journal of

Atmospheric and Oceanic Technology. 20, 1869–1979.

Qing, W.U., G. Chen. (2015). Validation and intercomparison of HY-2A/ MetOp-A/ Oceansat-2

scatterometer wind products. Chinese Journal of Oceanology and Limnology. 33:5, 1181-

1190.

Ruti, P. M., S. Marullo, F. D’Ortenzio, and M. Tremant. (2008). Comparison of analyzed and

measured wind speeds in the perspective of oceanic simulations over the Mediterranean

basin: Analyses, QuikSCAT and buoy data. Journal of Marine Systems. 70, 33-48.

DOI:10.1016/j.jmarsys.2007.02.026

Saha, S., S. Moorthi, X. Wu, J. Wang et al. 2014: The NCEP Climate Forecast System Version

2. Journal of Climate. 27, 2185–2208. DOI:10.1175/JCLI-D-12-00823.1

Satheesan, K., A. Sarkar, A. Parekh, M. R. R. Kumar, and Y. Kuroda. (2007). Comparison of

Wind Data from QuikSCAT and Buoys in the Indian Ocean. International Journal of Remote

Sensing. 28:10, 2375–2382. DOI: 10.1080/01431160701236803.

Singh, P., A. Parekh, and R. Attada. (2013). Comparison of a single logarithmic and equivalent

neutral wind approaches for converting buoy-measured wind speed to the standard height:

special emphasis to North Indian Ocean. Theoretical and Applied Climatology. 111, 455-

463. DOI: 10.1007/s00704-012-0674-2

Stoffelen, A. and D. Anderson. (1997) Scatterometer data interpretation: Estimation and

Validation of the transfer function CMOD4. Journal of Geophysical Research. 102:C3, 5767-

5780. DOI: 96JC02860.0148-0227/97/96JC-0286050

Sudha, A.K., and C.V. Prasada Rao (2013). Comparison of Oceansat-2 scatterometer winds

with buoy observations over the Indian Ocean and the Pacific Ocean. Remote Sensing

Letters. 4:2, 171-179. DOI: 10.1080/2150704X.2012.713140

Sverdrup, H. U. (1953). On conditions for the vernal blooming of phytoplankton. ICES Journal

of Marine Science. 18, 287–295.

Schmidt et al. [44]

Swart, S., S.J. Thomalla, and P.M.S. Monteiro. (2014). The seasonal cycle of mixed layer

dynamics and phytoplankton biomass in the Sub-Antarctic Zone: A high-resolution glider

experiment. Journal of Marine Systems. DOI:10.1016/j.jmarsys.2014.06.002

Sweeney, C., E. Gloor, A. R. Jacobson, R. M. Key, G. McKinley, J. L. Sarmiento, and R.

Wanninkhof. (2007), Constraining global air-sea gas exchange for CO2 with recent bomb

14C measurements. Global Biogeochemical Cycles. 21:GB2015. DOI:10.1029/ 2006GB

002784.

Takahashi, T., S. C. Sutherland, R. Wanninkhof, C. Sweeney et al. (2009). Climatological

mean and decadal change in surface ocean pCO2, and net sea-air CO2 flux over the global

oceans. Deep-Sea Research II. 56, 554-577. DOI: 10.1016/j.dsr2.2008.12.009

Takahashi, T., C. Sweeney, B. Hales, D. W. Chipman, T. Newberger, J. G. Goddard, R. A.

Iannuzzi, and S. C. Sutherland. (2012). The changing carbon cycle in the Southern Ocean.

Oceanography. 25:3, 26–37. DOI:10.5670/oceanog.2012.71.

Tang, W., W. T. Liu, and B. W. Stiles. (2004). Evaluation of High-Resolution Ocean Surface

Vector Winds Measured by QuikSCAT Scatterometer in Coastal Regions. IEEE

Transactions on Geoscience and Remote Sensing. 42:8, 1762. DOI: 10.1109/ TGRS.2004.

831685

Thomalla, S.J., N. Faucherau, S. Swart, and P.M.S. Monteiro. (2011). Regional scale

characteristics of the seasonal cycle of chlorophyll in the Southern Ocean. Biogeosciences.

8, 2849–2866. DOI: 10.5194/bg-8-2849-2011

Thompson, D. W. J., S. Solomon, P. J. Kushner, M. H. England, K. M. Grise, and D. J. Karoly.

(2011). Signatures of the Antarctic ozone hole in Southern Hemisphere surface climate

change. Nature Geoscience. 4, 741-749. DOI: 10.1038/NGEO1296

Vozelgang, J., and A. Stoffelen. (2012). Scatterometer wind vector products for application in

meteorology and oceanography. Journal of Sea Research. 74, 16-25. DOI:10.1016/

j.seares.2012.05.002

Wanninkhof, R. (1992). Relationship between gas exchange and wind speed over the ocean.

Journal of Geophysical Research. 97:C5, 7373–81. DOI: 92JC00188.0148-0227/92/92 J C-

00188505.00

Schmidt et al. [45]

Wanninkhof, R. and W. R. McGillis. (1999). A cubic relationship between air-sea CO2

exchange and wind speed. Geophysical Research Letters. 26:13, 1889-1892. DOI:

1999GL900363.0094-8276/99/1999GL900363

Wanninkhof, R., W. E. Asher, D. T. Ho, C. Sweeney, and W. R. McGillis. (2009). Advances in

Quantifying Air-Sea Gas Exchange and Environmental Forcing. Annual Review of Marine

Science. 1, 213-244. DOI: 10.1146/annurev.marine.010908.163742

Yang, X., G. Liu, Z. Li, Y. Yu. (2014). Preliminary validation of ocean surface vector winds

estimated from China’s HY-2A scatterometer. International Journal of Remote Sensing.

35:11-12, 4532-4543. DOI: 10.1080/ 01431161.2014.916049

Yuan, X. (2004). High-wind-speed evaluation in the Southern Ocean. Journal of Geophysical

Research. 109:D13101. DOI:10.1029/2003JD004179

Yuan, X., J. Patoux, and C. Li (2009), Satellite-based midlatitude cyclone statistics over the

Southern Ocean: 2. Tracks and surface fluxes. Journal of Geophysical Research.

114:D04106. DOI: 10.1029/ 2008JD010874.

Zhang, Huai-Min. (2006). Blended and Gridded High Resolution Global Sea Surface Winds

from Multiple Satellites. NOAA NESDIS National Climatic Data Center. 1-24