AMERICANMETEOROLOGICALSOCIETY

Bulletin of the American Meteorological Society

EARLY ONLINE RELEASEThis is a preliminary PDF of the author-producedmanuscript that has been peer-reviewed and accepted for publication. Since it is being postedso soon after acceptance, it has not yet beencopyedited, formatted, or processed by AMSPublications. This preliminary version of the manuscript may be downloaded, distributed, andcited, but please be aware that there will be visualdifferences and possibly some content differences between this version and the final published version.

The DOI for this manuscript is doi: 10.1175/BAMS-D-12-00075.1

The final published version of this manuscript will replacethe preliminary version at the above DOI once it is available.

© 2013 American Meteorological Society

1

Driftsondes: Providing In-Situ Long-Duration Dropsonde Observations over Remote Regions 1

2

Stephen A. Cohn1, Terry Hock1, Philippe Cocquerez2, Junhong Wang1, Florence Rabier3, David 3

Parsons4, Patrick Harr5, Chun-Chieh Wu6, Philippe Drobinski7, Fatima Karbou3, Stéphanie 4

Vénel2, André Vargas2, Nadia Fourrié3, Nathalie Saint-Ramond3, Vincent Guidard3, Alexis 5

Doerenbecher3, Huang-Hsiung Hsu8, Po-Hsiung Lin6, Ming-Dah Chou9, Jean-Luc Redelsperger10, 6

Charlie Martin1, Jack Fox1,11, Nick Potts1, Kathryn Young1, and Hal Cole1 7

8

1National Center for Atmospheric Research, Boulder, CO, USA 9

2Centre National d’Etudes Spatiales, Toulouse, France 10

3CNRM-GAME, Météo-France and CNRS, Toulouse, France 11

4University of Oklahoma, Norman, OK, USA 12

5Naval Postgraduate School, Monterey, California, USA 13

6National Taiwan University, Taipei, Taiwan 14

7Ecole Polytechnique/CNRS, Palaiseau, France 15

8Research Center for Environmental Change, Academia Sinica, Taipei, Taiwan 16

9National Central University, Taiwan 17

10Laboratoire de Physique des Océans, CNRS, IFREMER, IRD, UBO, France 18

11now at Advanced Radar Corporation, Boulder, CO, USA 19

20

2

Corresponding author: Stephen A. Cohn, National Center for Atmospheric Research 21

PO Box 3000, Boulder, CO 80307-3000 USA 22

Email cohn@ucar.edu 23

For submission to the Bulletin of the American Meteorological Society 24

November 2012 25

Revised: February 2013 26

3

Abstract 27

Constellations of Driftsonde Systems, gondolas floating in the stratosphere and able to release 28

dropsondes upon command, have so far been used in three major field experiments from 2006 29

through 2010. With them, high-quality, high-resolution, in-situ atmospheric profiles were made 30

over extended periods in regions that are otherwise very difficult to observe. The 31

measurements have unique value for verifying and evaluating numerical weather prediction 32

models and global data assimilation systems; they can be a valuable resource to validate data 33

from remote sensing instruments, especially on satellites, but also airborne or ground based 34

remote sensors. These applications for models and remote sensors result in a powerful 35

combination for improving data assimilation systems. Driftsondes also can support process 36

studies in otherwise difficult locations, for example to study factors that control the 37

development or decay of a tropical disturbance, or to investigate the lower boundary layer over 38

the interior Antarctic continent. The Driftsonde System is now a mature and robust observing 39

system that can be combined with flight-level data to conduct multidisciplinary research at 40

heights well above that reached by current research aircraft. In this article we describe the 41

development and capabilities of the Driftsonde System, the exemplary science resulting from its 42

use to date, and some future applications. 43

44

4

Capsule: 45

A long-duration balloon-borne dropsonde platform brings innovative technology for multi-46

disciplinary science to hard-to-reach parts of the globe from heights unobtainable by research 47

aircraft. 48

5

1. Introduction 49

High quality in-situ measurements from radiosondes and dropsondes are the gold 50

standard for vertical profiles of fundamental atmospheric measurements such as wind, 51

temperature, and relative humidity. Satellite-borne remote sensors provide much-needed 52

global, long-term coverage, but do not match the ability of sondes to capture sharp transitions 53

and fine vertical structure, and have significant performance limitations (e.g., inability of 54

infrared sounders to penetrate clouds, poor accuracy in the boundary layer). Sondes are also a 55

trusted means to calibrate and validate remote sensors. However, it is challenging to launch 56

radiosondes from remote locations such as the ocean surface or the interior of Antarctica. 57

Aircraft release dropsondes above such locations, but are limited by the range and endurance 58

of the aircraft. The Driftsonde System fills an important gap in our ability to use sondes to 59

measure atmospheric profiles in remote locations. Although its creation was motivated by the 60

THORPEX (The Observing System Research and Predictability Experiment, e.g. Shapiro and 61

Thorpe 2004) to optimize the global observing system, it has contributed to varied 62

investigations ranging from understanding the development of tropical cyclones to validation of 63

satellite retrievals in Antarctica. 64

The Driftsonde is a unique, balloon-borne instrument that releases dropsondes to 65

provide high resolution in-situ profiles of atmospheric temperature, humidity, pressure, and 66

winds from the lower stratosphere down to the surface. It is ideal for applications over oceans 67

and remote polar and continental regions, filling critical gaps in data coverage where the 68

release of surface-based radiosondes is not possible. Figure 1 shows the Driftsonde System 69

6

concept in which a stratospheric balloon carries the Driftsonde gondola with a large number of 70

Miniature In-situ Sonde Technology (MIST) dropsondes for days to months. The balloon drifts 71

with the wind and sondes are released upon command. They parachute to the ground 72

providing high vertical resolution profiles. Data from each sonde are transmitted back to the 73

gondola, and from there to the ground via an Iridium satellite link. Commands from the ground 74

are also relayed to the gondola via satellite link. Data can be quality-checked in near real-time 75

and sent to the Global Telecommunications System (GTS), making it available for operational 76

and research uses such as Numerical Weather Prediction (NWP) models and to direct in-the-77

field experimental planning. The Driftsonde concept of launching dropsondes from balloons 78

was initially developed in the 1970s for the Global Atmospheric Research Program (GARP e.g., 79

Lally and Passi 1976). Although test flights were successfully completed, the instrument was not 80

used for GARP. The concept was revived for THORPEX in a discussion at the National Center for 81

Atmospheric Research (NCAR) between Melvin Shapiro, Vin Lally, Terry Hock, and Hal Cole, with 82

subsequent simulations by Rolf Langland (Naval Postgraduate School) confirming its potential 83

reach. 84

The Driftsonde System consists of the flight train shown in Figure 2, ground control 85

software which is web based, and ground control servers and associated hardware located at 86

NCAR in Boulder, Colorado. Web based software adds flexibility, so that an experiment 87

operations center can be located worldwide, or may rotate through several locations to support 88

the continuous (24-h, 7-day) nature of balloon flight operations. The latest version of the 89

gondola structure (Figure 3), made of insulating foam, contains up to 54 MIST dropsondes, a 90

custom electronics motherboard which acts as the brain of the system, lithium batteries to 91

7

power the gondola for the expected flight duration, radio equipment to communicate both 92

with a released dropsonde and with Iridium satellites, and electric heaters to maintain the 93

gondola electronics and batteries at an operational temperature. The heaters are powered by 94

solar panels mounted outside the gondola. Heating the sonde electronics and batteries before 95

they are released from the gondola ensures the sensors will operate normally. Early Driftsonde 96

tests were done with several ballooning partners1. Subsequent deployments have been a close 97

collaboration between NCAR and the French Centre National d’Etudes Spatiales (CNES) with 98

NCAR developing the Driftsonde measurement capability (gondola, MIST sondes, 99

communications, data quality, etc.), and CNES having responsibility for all ballooning 100

development and flight operations. 101

Characteristics of the Driftsonde’s MIST dropsondes are shown in Table 1. They are 102

physically smaller and lighter than current aircraft dropsondes but both use the Vaisala RSS921 103

sensor module, with the same pressure, temperature, and humidity sensors as the well 104

documented RS92-SGP radiosonde (Vaisala, 2013). Each MIST dropsonde undergoes a 105

calibration verification step at NCAR. While the MIST dropsondes make similar measurements 106

to aircraft dropsondes, the two platforms – Driftsonde gondola and aircraft - have notable 107

differences. Capabilities of the stratospheric balloons used to lift Driftsonde are central to its 108

strengths and limitations. Aircraft are maneuverable, but can remain aloft for only a few hours. 109

They also can precisely target specific locations. The Driftsonde is not maneuverable, but can 110

1 Before and between field projects, we received assistance with ballooning tests from Terry Deshler and the

University of Wyoming, Tim Lachenmeier and Near Space Corporation, Inc., and the NASA Columbia Scientific

Ballooning Facility.

8

remain aloft for several months. Sondes are released from much higher altitudes than most 111

aircraft (Figure 4), and multiple Driftsondes can be in flight simultaneously. In the just 112

completed Concordiasi experiment, a constellation of 13 Driftsondes were aloft simultaneously 113

for about two months, with drops controlled from the ground in McMurdo Station, Antarctica, 114

Toulouse, France, and Boulder, Colorado. In general, because stratospheric balloons drift with 115

the wind and have long duration, Driftsonde data can provide synoptic-scale or finer 116

observations with wide-ranging geographical coverage that would be difficult to obtain with 117

research aircraft. On the other hand, precise targeting of drops is limited by the accuracy of 118

balloon trajectory forecasts. 119

2. Field Experiments and Science Applications 120

As the Driftsonde System was developed (Figure 5), it was deployed in three field 121

experiments associated with THORPEX activities (Table 2). Each revealed and led to needed 122

improvements, and from these experiments we also learned better how to take advantage of 123

the system’s strengths for varied science applications. Details of the Driftsonde System 124

performance and scientific applications in each experiment are presented in the following 125

sections. 126

a. Driftsonde observations during the African Monsoon Multidisciplinary Analyses 127

The first field project experience with Driftsonde was in the African Monsoon 128

Multidisciplinary Analyses (AMMA) project2, both as a rigorous field test and for its scientific 129

2 AMMA, based on a French initiative, was organized by an international scientific group and is currently funded by

a large number of agencies, especially from France, UK, US and Africa. It has been the beneficiary of a major

9

value. AMMA was organized to advance understanding of the West African Monsoon system 130

and to improve predictions of its variability and the associated wide range of societal impacts. It 131

is a major international program led by France, but involving agencies and scientists located in 132

the United Kingdom, Germany, the United States, and other countries across Africa and Europe. 133

The Driftsonde deployment supported AMMA’s research focus on high-impact weather and 134

was undertaken through a collaboration between AMMA and THORPEX. The program is 135

summarized in Redelsperger et al. (2006), with the Driftsonde observing strategy described in 136

Rabier et al. (2008). 137

The AMMA measurement strategy included long-term observations from 2002 through 138

2010 to investigate the inter-annual variability of the West African Monsoon. Within this period 139

was an Extended Observing Period (EOP) from 2005-2007 to document the annual cycle, and 140

four Special Observing Periods (SOPs) during 2006 to provide specific observations of physical 141

processes and of weather systems. The Driftsonde operations took place during the 4th SOP 142

covering the late monsoon period in August and September 2006. 143

The Driftsonde deployment was considered a THORPEX Observing System Test (e.g., 144

Shapiro and Thorpe 2004). Thus, a large component of Driftsonde operations concentrated on 145

engineering tests, including the first major test of NCAR’s new smaller and lighter dropsonde 146

called MIST, which at 175 g was less than half the weight of the previous dropsondes and was 147

financial contribution from the European Community's Sixth Framework Research Programme. Detailed

information on scientific coordination and funding is available on the AMMA International web site

http://www.amma-international.org.

10

developed specifically for use in the Driftsonde. Miniaturization of the dropsonde was 148

necessary for ballooning where weight is more critical than for aircraft deployments. For 149

AMMA, the Driftsonde gondola held 49 MIST dropsondes. The AMMA deployment was the first 150

scientific use of the new CNES 12-m super pressure balloons, the new NCAR gondolas and the 151

MIST dropsondes. 152

Eight Driftsondes were launched from Zinder, Niger, and floated at about 20 km as they 153

drifted eastward reaching the Atlantic Ocean. The location was chosen to allow investigators to 154

study both African Easterly Waves over central and western Africa and the potential 155

intensification of these waves into tropical disturbances or even hurricanes over the subtropical 156

Atlantic. Flight trajectories and the dropsonde locations are shown in Fig. 6. The dropsondes 157

were able to sample both the Saharan Air Layer that is often advected over the tropical Atlantic 158

and precursor environments, and the near-storm environments associated with 2006 tropical 159

storm Florence and hurricanes Gordon and Helen. The first balloon was launched on 28 August 160

2006 and the termination of the final balloon over the central subtropical Atlantic occurred on 161

22 September. Two balloons had mission durations in excess of 8 days. Sondes were typically 162

deployed near 0000 and 1200 UTC as well as on demand for promising weather conditions. For 163

further details on the Driftsonde operations during AMMA, a description of the challenges 164

associated with balloon and dropsonde design and preliminary scientific results, refer to the 165

overview of the deployment presented in Drobinski et al. (2006) and Drobinski et al. (2012a). 166

Much was learned about the system’s operation and performance as 124 sondes were 167

successfully deployed from the 8 Driftsonde gondolas. While there was one premature failure 168

11

of a stratospheric balloon and lessons learned about Driftsonde launch procedures, most 169

engineering challenges highlighted the need to improve aspects of the Driftsonde gondola 170

design and usability of system software. There were many cases where sondes failed to launch, 171

as well as periods of lost communication between the ground control station and the gondola. 172

Another significant conclusion was the need to redesign the MIST sonde to include a pressure 173

sensor. A pressure sensor was not included in the AMMA version of the MIST sonde because of 174

the incorrect assumption that the pressure for the dropsonde profile could be obtained from 175

knowledge of the pressure and GPS altitude at launch and the hydrostatic equation. However, 176

the flight level pressure sensor on the gondolas did not have sufficient accuracy, so small errors 177

in the initial pressure were magnified by the downward integration of the hydrostatic equation. 178

A success of AMMA was the demonstrated ability to target measurements with 179

Driftsondes launched 4-5 days before the sampling window. The successful sampling of storms 180

was due to the accuracy of the upper-level winds predicted from operational NWP models, the 181

quasi-non-divergent nature of the flow at 20 km, and successful tropical storm genesis 182

forecasts by the AMMA team through combining experimental and operation products for 183

storm genesis. 184

Despite technical challenges, AMMA demonstrated the scientific value of Driftsonde, in 185

particular to evaluate operational model performance and identify specific areas for model 186

improvement. Drobinski et al. (2012b) use Driftsonde and other data to evaluate performance 187

of the ECMWF Integrated Forecast System (IFS) and the two versions of the Météo-France 188

ARPEGE operational forecast system through comparison of the dropsonde observations and 189

12

the model analysis and prediction. The concept was to improve performance in key regions and 190

different flow regimes using the special dropsonde observations to supplement the assessment 191

procedure employed by the operational centers. This technique was extended to evaluate the 192

impacts of recent upgrades in model physics and assimilation techniques. 193

The findings from Drobinski et al. (2012b) include that these models well represent the 194

complex vertical structure of humidity associated with the Saharan Air Layer (Fig. 7a). However, 195

the comparison identified temperature errors of several degrees C in both modeling systems 196

near the base of the Saharan Air Layer (Fig. 7b). The relatively large errors are likely because of 197

the lack of dust and the associated radiative impacts within NWP models. This result suggests 198

shortcomings in the assimilation system, perhaps due to the vertical resolution of the satellite 199

data, and argues for the inclusion of aerosols and their radiative effects in NWP models. In this 200

case, static stability errors in the vicinity of the Saharan dust could, in turn, impact the 201

likelihood, intensity and structure of convection. Considerable debate currently exists in 202

determining and explaining the potential impacts of the Saharan Air Layer on tropical cyclones 203

(e.g., see Braun 2010 and references therein). It was also found that within the analyses, the 204

zonal and meridional winds in the Saharan Air Layer cases have significant errors (Figs. 7c and 205

d). The Driftsonde observations were also valuable as “ground-truth” in data impact and data 206

assimilation experiments (Drobinski et al, 2012b). 207

b. Driftsonde observations during the THORPEX-Pacific Asian Regional Campaign 208

The Driftsonde experience in AMMA was a success both in identifying technical issues 209

which needed attention after the campaign and collecting scientifically valuable observations. 210

13

Prior to the next large field use in the THORPEX-Pacific Asian Regional Campaign (T-PARC)3 in 211

2008, many parts of the system were upgraded. In particular a pressure measurement was 212

added to the MIST sonde, robustness of the satellite communication link was improved, and 213

reliability of the sonde separation from the gondola when a drop is commanded was also 214

improved. 215

As a multi-national field campaign and research initiative, T-PARC addressed the 216

shorter-range dynamics and forecast skill of one region (eastern Asian and the western North 217

Pacific) and its downstream impact on the medium-range dynamics and forecast skill of another 218

region (eastern North Pacific and North America). High-impact weather events over the regions 219

examined in T-PARC have strong dynamical links downstream. For example, persistent deep 220

tropical convection or the extratropical transition of tropical cyclones can trigger downstream 221

responses over the eastern North Pacific, North America, and beyond via upper-tropospheric 222

wave packets on the primary midlatitude waveguides (Anwender et al. 2008, Harr et al. 2008). 223

Then, wave packets can be invigorated by subsequent downstream cyclogenesis events that are 224

often associated with reduced predictability. High-impact weather events over North America 225

driven by these processes can include intense extratropical cyclones, floods, severe weather 226

and hot, dry winds that increase the risk of wildfires and the severity of droughts. While T-PARC 227

objectives encompassed mesoscale and synoptic-scale processes associated with tropical 228

cyclones over the western North Pacific and eastern Asia, they also addressed medium-range 229

forecast skill associated with downstream impacts across the North Pacific and beyond. 230

3 T-PARC was supported financially by Germany, Canada, Japan, Australia, France, Korea, Taiwan, the U.K., the

European Centre for Medium-Range Weather Forecasts, and the United States.

14

Dropsondes were an important contribution to T-PARC. In addition to use of Driftsonde, 231

four aircraft from three countries [U.S.: USAF WC-130J and NRL P-3, Taiwan: DOTSTAR Astra 232

(Wu et al. 2005, 2007a), and Germany: DLR Falcon] were used to take special observations. The 233

collaborative program required an experimental design that covered a very wide geographic 234

range to address three primary components: (1) A tropical measurement strategy examined 235

circulations of the tropical western North Pacific monsoon environment as they related to 236

enhanced and reduced periods of widespread deep convection, tropical cyclone formation, 237

tropical cyclone intensification, and tropical cyclone structure change; (2) A measurement 238

strategy for the extratropical transition and its downstream impacts followed the poleward 239

movement of a decaying tropical cyclone and the intense cyclogenesis that results from its 240

interaction with the midlatitude circulation; and (3) A targeted observation strategy focused on 241

regions in which extra observations may reduce numerical forecast error growth (Wu et al. 242

2007b, 2009, Harnisch and Weissmann 2010, Reynolds et al. 2010, Weissmann et al. 2010, Chou 243

et al. 2011). In T-PARC, the targeted observations were aimed at reducing errors associated 244

with tropical cyclone track forecasts, which included whether a tropical cyclone would recurve, 245

the longitude of recurvature, and the orientation and speed along the track following 246

recurvature. 247

To accomplish the primary objectives of T-PARC, a complete tropical-to-extratropical 248

measurement strategy was necessary. For example, predictability associated with extratropical 249

transition depends on the intensity and structure of the tropical cyclone, where and when the 250

tropical cyclone arrives in the midlatitude westerlies, the characteristics of the midlatitude 251

15

waveguide that impact the extratropical transition-related cyclogenesis, and the downstream 252

propagation and evolution of the wave packets. 253

The motivation for deployment of Driftsondes in T-PARC was to provide measurements 254

over data sparse regions of the tropical central Pacific. The data from the Driftsonde 255

complemented satellite observations and provided calibration and validation data for new 256

satellite-based observations (Hawkins and Velden 2011) and global reanalysis products (Wang 257

et al. 2010). During T-PARC, 16 Driftsondes were launched from the southern end of the Big 258

Island of Hawaii between 15 August and 30 September 2008. Thirteen of the Driftsondes 259

traveled at an altitude of about 30 km for up to five days to reach the western North Pacific and 260

the primary T-PARC observation region (Figure 8). Throughout T-PARC, 254 dropsondes were 261

deployed from the Driftsondes. The location and timing of the dropsonde deployments were 262

coordinated from the T-PARC operations center at the Naval Postgraduate School in Monterey, 263

CA, taking advantage of the now web-based Driftsonde control and display software. During 264

each balloon flight, data were relayed to the T-PARC operations center, quality controlled, and 265

transmitted to the GTS for use at operational weather centers. 266

As an example of Driftsonde use during T-PARC, the fourth Driftsonde was launched on 267

24 August 2008 and on 29 August it reached a tropical disturbance that was being investigated 268

by the T-PARC aircraft (Figure 9). While the Driftsonde was over-flying the tropical disturbance 269

in the lower stratosphere two aircraft were deploying dropsondes from their respective flight-270

level altitudes. Seven dropsondes were deployed from the Driftsonde and provided 271

16

measurements of two upper-tropospheric cyclonic systems that were preventing the 272

development of the tropical disturbance. 273

Driftsondes in T-PARC were flown with zero-pressure balloons. These were designed to 274

float much higher than the super-pressure balloons used for AMMA (and later in Concordiasi), 275

but also had a shorter flight lifetime. Because of light winds and a flaw in the ballooning 276

technique several flights failed to advect far enough westward to enter the most interesting 277

measurement region. However, overall, data obtained from dropsondes released from the 278

Driftsondes provided valuable measurements of persistent deep convection with special 279

emphasis on the detailed vertical structure, impacts of vertical wind shear, and upper-level 280

divergent outflow. Because of the Driftsonde launch location and trajectories, data were 281

instrumental in monitoring tropical cloud clusters that migrated over the data-sparse region of 282

the tropical central Pacific until the clusters reached the region of T-PARC aircraft operations. 283

c. Driftsonde observations during the Concordiasi Field Experiment 284

The third major Driftsonde deployment was in 2010 for the Concordiasi field 285

experiment4 (Rabier et al, 2010 and 2013), a multi-disciplinary effort jointly conducted by 286

several groups in France and the United States to study the lower stratosphere and 287

4 Concordiasi was organized by an international scientific group and supported by the following agencies: Météo-

France, CNES, IPEV, PNRA, CNRS/INSU, NSF, NCAR, Concordia consortium, University of Wyoming, Purdue

University, and the University of Colorado. ECMWF also contributed to the project through computer resources

and support, and scientific expertise. The two operational polar agencies PNRA and IPEV are thanked for their

support at Concordia station.

17

troposphere above Antarctica. Concordiasi was one of the cluster of THORPEX projects 288

associated with the International Polar Year (e.g., Renfrew et al. 2008, Hanesiak et al. 2010, 289

Kristjánsson et al. 2011). The primary focus of Concordiasi was to validate the use of satellite 290

observations and to document which observing systems are most relevant for numerical 291

weather prediction over the polar areas. Concordiasi field experiments took place in Austral 292

springs 2008, 2009, and 2010, including surface measurements and radiosoundings at the 293

Concordia Antarctica station at Dome C, and radiosoundings at the Dumont d'Urville and 294

Rothera sites on Antarctica. In 2010 Driftsonde was part of an innovative constellation of 295

balloons that provided a unique set of measurements spanning a large spatial extent (both 296

horizontal and vertical) and time. The balloons drifted for several months in the lower 297

stratosphere around 18 km, circling over Antarctica in the polar vortex. The balloon flotilla 298

formed a regional observatory of the atmosphere. As in the earlier Driftsonde experiments, 299

hundreds of soundings were performed on command. The launch campaign took place from 300

the U.S. McMurdo Station, located at 78° south latitude. Nineteen balloons were launched 301

between 8 September and 26 October 2010. The mean flight duration was 69 days, while the 302

longest flight lasted 95 days. Thirteen balloons carried a Driftsonde, and six carried other 303

instruments for Concordiasi. The long flight duration of the super pressure balloons used for 304

Concordiasi, months rather than about a week for the previous Driftsonde use, was critical to 305

enable the project’s science. 306

To prepare for Concordiasi, the Driftsonde System was modified for the much longer 307

duration flights and challenging range of thermal conditions it would encounter. Early in this 308

high latitude project the gondolas were in total darkness, and later in the project they 309

18

transitioned to full sunlight. The software was also enhanced to allow for drops at pre-310

scheduled times. Overall, the 13 Driftsonde gondolas returned 644 high-quality profiles, with 311

only 14 failed drops. This is a much higher success rate than in either AMMA or T-PARC, and 312

resulted in an excellent spatial distribution of observations both over the Antarctic continent 313

and the surrounding ocean. Figure 10 shows the comprehensive coverage and distribution of 314

drop locations over the full experiment, as well as an example of the coverage of the 315

constellation on a single day. 316

Many dropsondes were released to coincide with Driftsonde overpasses of Concordia 317

station, allowing comparison of dropsonde and radiosonde profiles, and also to coincide with 318

overpasses of the MetOp satellite, allowing comparison with data from the Infrared 319

Atmospheric Sounding Interferometer (IASI). IASI is an advanced infrared sounder which has a 320

large impact in NWP systems in general. However, there are some difficulties in its use over 321

polar areas because the extremely cold polar environment makes it more difficult to extract 322

temperature information from infra-red spectra and makes it difficult to detect cloud 323

properties. As a consequence, IASI is currently under-utilized over Antarctica. 324

A number of important results have already come from the 2010 Concordiasi dataset, as 325

described in the Concordiasi workshop report (Rabier et al, 2013). Wang et al. (2013) compare 326

sonde profiles with satellite retrievals, using the NOAA PROducts Validation System (NPROVS) 327

to match Concordiasi dropsonde and radiosonde profiles with profiles from several satellite 328

products. A comparison of temperature profiles shows a cold bias present in all satellite data. 329

The cold bias has larger magnitude relative to the dropsonde data than the radiosonde for all 330

19

satellite products except the COSMIC (The Constellation Observing System for Meteorology, 331

Ionosphere, and Climate) (Fig. 11). The difference between the radiosonde and dropsonde bias 332

can be traced to a larger cold bias over the Antarctic continent than over the coast and ocean 333

since all radiosonde stations but two are located along the coast (Fig. 10a). The source of this 334

bias remains a topic of investigation. Aside from the bias, and an inability to resolve detailed 335

thermodynamic structures near the surface and tropopause, the satellite retrievals reproduce 336

the temperature profiles reasonably well. 337

In addition to temperature and humidity profiles, cloud properties such as cloud top 338

pressure and cloud effective emissivity (emissivity of an equivalent single layer cloud) can be 339

retrieved from IASI measurements. These retrievals are highly dependent on the quality of 340

temperature and humidity profiles. As reported in Rabier et al. (2013), detection of cloud 341

properties can be improved by using an accurate atmospheric profile provided by the 342

Concordiasi dropsondes rather than the atmospheric model. 343

Another result from this dataset comes from the use of Concordiasi Driftsonde 344

observations in real-time at NWP centers. As noted in Rabier et al (2013), large systematic 345

differences exist between various NWP analyses and forecasts for temperature over Antarctica, 346

and for winds on the surrounding oceans. Comparison between short-range forecasts and the 347

Concordiasi dropsonde data show that models poorly represent near-surface temperature over 348

the Antarctic high terrain. The strong thermal inversions are challenging because numerical 349

models need very good representations of both turbulent exchange processes and snow 350

processes to simulate this extreme atmospheric behavior. The difference between the 351

20

dropsonde and the model temperatures at the lowest model level is presented in Figure 12 for 352

the French global model. The model is too warm over the plateau in Antarctica, and it is too 353

cold over the surrounding sea-ice. 354

The impact of dropsondes on NWP models has also been studied with data denial 355

experiments and advanced data impact diagnostics. Dropsondes have a positive impact on the 356

forecast performance in different models, with an impact of the same order of magnitude as 357

that of radiosondes. Rabier et al (2013) report that the average error reduction per observation 358

is much larger for dropsondes than for satellite data, and that dropsonde observations have 359

greater impact when they are closer to the pole. In Figure 13, the impact of dropsonde 360

temperature and wind profiles is illustrated at high levels (pressure less than 400 hPa) and low 361

levels (pressure greater than 400 hPa), together with the number of observations. Overall, 362

temperature information contributes most at low levels, and wind information contributes 363

more at high levels. However on a per-observation basis both wind and temperature have 364

larger impact at low levels, where there are very few other observations. 365

These results from Driftsonde data in Concordiasi provide insight into improvements to 366

the global observing system that must be achieved to improve NWP over the polar areas. This is 367

important not only to improve forecast performance but also for producing more accurate re-368

analyses of the atmosphere to document climate change. 369

3. Conclusion 370

Thanks to the Driftsonde System, in-situ measurements were obtained in parts of the 371

world which are not accessible by any other means. This has provided invaluable information 372

21

about model strengths and weaknesses, and about which observations will be needed in the 373

future to monitor the climate, especially in polar areas. The excellent technical success 374

achieved during Concordiasi demonstrates that Driftsonde is now a mature, reliable, and 375

productive observing system. Its strengths include the ability to reach difficult parts of the 376

globe, to collect highly accurate, in-situ dropsonde profiles, to reliably release up to 54 377

dropsondes per system from the lower stratosphere, and to be deployed as a constellation with 378

many Driftsondes flying simultaneously. 379

Like aircraft dropsonde systems, field experiments using Driftsondes involve significant 380

cost and require much advance planning. It can take months to understand likely flight paths 381

and obtain permissions to overfly many countries. If a balloon drifts near a region where 382

overflight permissions have not been granted, it must be cut down. The ability of the Driftsonde 383

to observe specific phenomena depends critically upon finding a suitable launch site relative to 384

stratospheric wind patterns. For T-PARC, finding a subtropical Pacific island for which 385

stratospheric wind patterns intersected the climatological tracks of tropical cyclones was 386

surprisingly difficult. In contrast, finding a suitable launch site for AMMA within Africa was more 387

straightforward. Once a launch site is selected, the success of the Driftsonde to target a specific 388

event depends on accurate forecasts of both the wind field in the stratosphere and the 389

evolution and movement of the event to be targeted. Despite the ability of Driftsondes to 390

intercept tropical cyclones during AMMA and to a lesser extent during T-PARC, reliance upon 391

forecasts with lead times of several days is a disadvantage relative to aircraft dropsonde 392

deployment. At far longer lead times, the Driftsonde behavior from CONCORDIASI and 393

experience from ballooning campaigns during GARP suggest that the balloons tend to be 394

22

advected into the confluent, more dynamically active regions of the atmosphere. For many 395

aspects of weather research, this behavior is desirable. 396

Despite these complexities, scientifically, the Driftsondes are well suited to numerous 397

applications, especially because of their ability to operate in otherwise data-sparse regions. In 398

particular, they have unique value for verifying and evaluating NWP models, global reanalysis 399

models, and data assimilation approaches. The measurements are also a valuable resource to 400

validate remote sensors, especially on satellites but also airborne or ground based remote 401

sensors. Driftsondes also can support process studies in otherwise difficult locations. In AMMA 402

and T-PARC, examples are effects of the SAL and factors that control the development of a 403

tropical disturbance. There is also a potential role for Driftsonde operationally, although costs 404

and forecast impacts of this have not been considered in detail. For example, a concept 405

discussed in the early years of Driftsonde was that a series of Driftsondes could be released at 406

regular intervals from sites in Asia to provide synoptic data over the Pacific Ocean, or from the 407

east coast of the United States for regular observations over the Atlantic Ocean. In addition to 408

Driftsonde flights, the Concordiasi program included measurements of ozone-related processes, 409

microphysics of stratospheric clouds, and remote sensing using GPS occultation. It showed the 410

potential of combining such diverse flight level measurements with the vertical profiles from 411

dropsondes to carry out a multi-disciplinary experiment that would not be possible with 412

aircraft. 413

The Driftsonde System has been discussed as a possible contributor to future field 414

studies. One is a long-duration stratospheric balloon campaign at the equator intended to 415

23

study, among other goals, the dynamics of the equatorial middle atmosphere with a focus on 416

the quasi-biennial oscillation, and transport, dehydration, and clouds in the tropical tropopause 417

layer. A second field study suggested Driftsondes as one of a synergistic set of tools to study the 418

propagation and effects of orographically generated atmospheric gravity waves from near the 419

surface to the upper atmosphere. Those with interest in using Driftsondes in their research are 420

encouraged to contact the lead authors. 421

24

Acknowledgements 422

Development and field use of the Driftsonde System is a joint effort of many people, 423

institutions, and nations. We thank all who made the development and deployments to AMMA, 424

T-PARC, and Concordiasi successful. We are grateful to Rolf Langland, Mark Bradford, Joe 425

VanAndel, Dean Lauritsen, Chip Owens, Clayton Arendt, and Mary Hanson for contributions to 426

the development. 427

Melvyn Shapiro is especially acknowledged for his resurrection of the early ideas of the 428

late Vin Lally, which were developed decades before their time. In many regards, this paper is a 429

testament to Vin’s creativity. 430

We are grateful for funding support from the National Science Foundation’s Division of 431

Atmospheric and Geospace Sciences and Office of Polar Programs (Grants ATM-0301213, ATM-432

9732665, ANT-0733007, ANT-1002057, and AGS-0736003); National Oceanographic and 433

Atmospheric Administration (Grant NA17GP1376); National Science Council of Taiwan (Grants 434

NSC 96-2745-M-002-004, NSC 97-2111-M-002-005, and NSC 97-2111-M-002-016-MY3); Taiwan 435

Central Weather Bureau (Grant MOTC-CWB-97-6M-01); Office of Naval Research (Grant 436

N00173-08-1-G007 and N00014-09-WR20008), and the support of the national and 437

international THORPEX Project Offices. NSF further supported these field projects through their 438

support of the US THORPEX Project Office, and the Lower Atmospheric Observing Facilities. 439

The National Center for Atmospheric Research is sponsored by the National Science 440

Foundation. 441

25

References 442

Anwender, D., P. Harr, and S. Jones, 2008: Predictability associated with the extratropical 443

transition of tropical cyclones: Case studies, Mon. Wea. Rev., 136, 3226–3247. 444

Braun, S. A., 2010: Reevaluating the role of the Saharan air layer in Atlantic tropical cyclogenesis 445

and evolution, Mon. Wea. Rev., 138, 2007–2037. 446

Chou, K.-H., C.-C. Wu, P.-H Lin, S. D. Aberson, M. Weissmann, F. Harnisch, and T. Nakazawa, 447

2011: The impact of dropwindsonde observations on typhoon track forecasts in DOTSTAR 448

and T-PARC, Mon. Wea. Rev., 139, 1728–1743. 449

Drobinski, P., D. Parsons, P. Cocquerex, A. Cardonne, H. Cole, J. Fox, J.P. Lefevre, J.P. 450

Redelsperger, K. Romberg, J. VanAndel and S. Venel, 2006: Des ballons stratospheriques 451

traquent la mousson Africaine, La Meteorologie, 55, 2–3. 452

Drobinski, P., P. Cocquerez, A. Doerenbecher, T. Hock, C. Lavaysse, D. Parsons, J. L. 453

Redelsperger, and S. Venel, 2012a: Hurricane and monsoon tracking with driftsondes. 454

Encyclopedia of Sustainability Science and Technology, Springer, in press pp. 455

Drobinski, P., F. Karbou, P. Bauer, P. Cocquerez, C. Lavaysse, T. Hock, D. Parsons, F. Rabier, J.L. 456

Redelsperger and S. Venel, 2012b: Driftsonde observations to evaluate numerical 457

weather prediction of the late 2006 African Monsoon, J. Climate and Appl. Meteor., in 458

press. 459

Hanesiak, J., and Coauthors, 2010: Storm Studies in the Arctic (STAR). Bull. Amer. Meteor. Soc., 460

91, 47–68. 461

26

Harnisch, F., and M. Weissmann, 2010: Sensitivity of typhoon forecasts to different subsets of 462

targeted dropsonde observations, Mon. Wea. Rev., 138, 2664–2680. 463

Harr, P., D. Anwender, S. Jones, 2008: Predictability associated with the downstream impacts of 464

the extratropical transition (ET) of tropical cyclones: Methodology and a case study of TY 465

Nabi (2005), Mon. Wea. Rev., 136, 3205–3225. 466

Hawkins, J., and C. Velden, 2011: Supporting meteorological field experiment missions and post-467

mission analysis with satellite digital data and products, Bull. Amer. Meteor. Soc., 92, 468

1009–1022. 469

Kristjánsson, J. E., and Coauthors, 2011: The Norwegian IPY–THORPEX: Polar Lows and Arctic 470

Fronts during the 2008 Andøya Campaign. Bull. Amer. Meteor. Soc., 92, 1443–1466. 471

Lally, V. E. and R. M. Passi, 1976: Height Determination from the Carrier Balloon Dropsonde. J. 472

Appl. Meteor., 15, 337–345. 473

doi: http://dx.doi.org/10.1175/1520-0450(1976)015<0337:HDFTCB>2.0.CO;2 474

Rabier, F., A. Bouchard, E. Brun, A. Doerenbecher, S. Guedj, V. Guidard, F. Karbou, V.-H. Peuch, 475

L. El Amraoui, D. Puech, C. Genthon, G. Picard, M. Town, A. Hertzog, F. Vial, P. 476

Cocquerez, S. A. Cohn, T. Hock, J. Fox, H. Cole, D. Parsons, J. Powers, K. Romberg, J. 477

VanAndel, T. Deshler, J. Mercer, J. S. Haase, L. Avallone, L. Kalnajs, C. R. Mechoso, A. 478

Tangborn, A. Pellegrini, Y. Frenot, J.-N. Thépaut, A. McNally, G. Balsamo, and P. Steinle, 479

2010: The Concordiasi Project in Antarctica, Bull. Amer. Meteorol. Soc., 91, 69-86. doi: 480

http://dx.doi.org/10.1175/2009BAMS2764.1 481

27

Rabier, F., S. Cohn, P. Cocquerez, A. Hertzog, L. Avallone, T. Deshler, J. Haase, T. Hock, A. 482

Doerenbecher, J. Wang, V. Guidard, J.-N. Thépaut, R. Langland, A. Tangborn, G. Balsamo, 483

E. Brun, D. Parsons, J. Bordereau, C. Cardinali, F. Danis, J.-P. Escarnot, N. Fourrié, R. 484

Gelaro, C. Genthon, K. Ide, L. Kalnajs, C. Martin, L.-F. Meunier, J.-M. Nicot, T. Perttula, N. 485

Potts, P. Ragazzo, D. Richardson, S. Sosa-Sesma, and A. Vargas, 2013: The Concordiasi 486

field experiment over Antarctica: First results from innovative atmospheric 487

measurements, Bull. Amer. Meteorol. Soc., doi: http://dx.doi.org/10.1175/BAMS-D-12-488

00005.1 489

Rabier, F., P. Gauthier, P., C. Cardinali, R. Langland, M. Tsyrulnikov, A. Lorenc, P. Steinle, R. 490

Gelaro and K. Koizumi, 2008: An update on THORPEX-related research in data 491

assimilation and observing strategies, Nonlinear Proces. in Geophys., 15, 81-94. 492

Redelsperger, J. L., C. D. Thorncroft, A. Diedhiou, T. Lebel, D. J. Parker, and J. Polcher, 2006: 493

African monsoon multidisciplinary analysis: An international research project and field 494

campaign, Bull. Amer. Meteor. Soc., 87, 1739–1746. 495

Renfrew, I. A., and Coauthors, 2008: The Greenland Flow Distortion Experiment, Bull. Amer. 496

Meteor. Soc., 89, 1307–1324. 497

Reynolds, C. A., J. D. Doyle, R. M. Hodur, and H. Jin, 2010: Naval Research Laboratory multiscsale 498

targeting guidance for T-PARC and TCS-08, Wea. Forecasting, 25, 526–544. 499

Shapiro, M. A. and A. J. Thorpe, 2004: Thorpex international science plan. WMO/TD-N◦ 1246, 500

WWRP/THORPEX, N◦ 2, 51 pp. 501

28

Vaisala (2013): Vaisala Radiosonde RS92 data sheet, available at 502

http://www.vaisala.com/Vaisala%20Documents/Brochures%20and%20Datasheets/RS92SGP-503

Datasheet-B210358EN-E-LoRes.pdf 504

Wang, J., T. Hock, S. A. Cohn, C. Martin, N. Potts, T. Reale, B. Sun, and F. Tilley, 2013: 505

Unprecedented upper air dropsonde observations over Antarctica from Concordiasi: 506

Validation of satellite-retrieved temperature profiles, Geophys. Res. Lett. (submitted). 507

Wang, J., L. Zhang, P.-H. Lin, M. Bradford, H. Cole, J. Fox, T. Hock, D. Lauritsen, S. Loehrer, C. 508

Martin, J. VanAndel, C.-H. Weng and K. Young, 2010: Water vapor variability and 509

comparisons in subtropical Pacific from T-PARC Driftsonde, COSMIC and reanalyses, J. 510

Geophys. Res., 115, D21108, doi:10.1029/2010JD014494. 511

Weissmann, M., F. Harnisch, C. C. Wu, P. H. Lin, Y.Ohta, K. Yamashita, Y. H. Kim, E. H. Jeon, T. 512

Nakazawa, and S. Aberson, 2010: The influence of assimilating dropsonde data on 513

typhoon track and mid-latitude forecasts. Mon. Wea. Rev., 139, 908–920. 514

Wu, C.-C., and Co-authors, 2005: Dropwindsonde Observations for Typhoon Surveillance near 515

the Taiwan Region (DOTSTAR): An overview, Bulletin of Amer. Meteor. Soc., 86, 787-790. 516

Wu, C.-C., and Co-authors, 2009: Inter-comparison of targeted observation guidance for tropical 517

cyclones in the North western Pacific, Mon. Wea. Rev., 137, 2471-2492. 518

Wu, C.-C., K.-H. Chou, P.-H. Lin, S. D. Aberson, M. S. Peng, and T. Nakazawa, 2007a: The impact 519

of dropwindsonde data on typhoon track forecasts in DOTSTAR. Weather and Forecasting, 520

22, 1157-1176. 521

29

Wu, C.-C., J.-H. Chen, P.-H. Lin, and K.-S. Chou, 2007b: Targeted observations of tropical cyclones 522

based on the adjoint-derived sensitivity steering vector, J. Atmos. Sci., 64, 2611-2626. 523

Wylie, D.P., W. P. Menzel, H. M. Woolf, and K. I. Strabala, 1994: Four Years of Global Cirrus Cloud 524

Statistics Using HIRS., J. of Climate, 7, 1972-1986. 525

30

Tables 526

Table 1: MIST dropsonde characteristics 527

MIST sonde Weight: 175 g

Length: 30.5 cm

Diameter: 4.6 cm

Pressure, temperature, and dual-

humidity sensors (used for T-PARC

and Concordiasi, not AMMA)

Vaisala module RSS921 (same as in RS92 radiosonde)

Sample rate: 0.5 s

Resolution: P: 0.01 hPa, T: 0.01o, H: 0.1%

Wind speed and direction GPS

Sample rate: 0.25 s

Resolution: 0.01 ms-1 and 0.1o

Sonde fall speed Approx. 90 ms-1 at 30 km, 45 ms-1 at 20 km, and 10 ms-1

at sea level

Sonde fall time Approx. 19 minutes from 30 km, 16 minutes from 20 km

528

31

Table 2: Driftsonde evolution through three field projects 529

African Monsoon Multidisciplinary Analyses (AMMA) August-September, 2006

Launch site: Zinder, Niger

Superpressure balloon. 8 flights. 3-18 day duration. 20 km float level.

178 MIST sondes with T, RH, GPS winds. No pressure sensor

Ground control through a terminal modem program with simple text commands and

manual operation.

THORPEX-Pacific Asian Regional Campaign (T-PARC) August-October, 2008

Launch site: Hawaii (the big Island)

Zero pressure balloon. 15 flights. 3-6 day duration. 30 km float level.

254 MIST sondes with P, T, RH, GPS winds (equivalent to Dropsonde sensor suite)

Web-based ground control for drops and display position and sounding data

Concordiasi Field Experiment, September-December, 2010

Launch site: McMurdo Station, Antarctica

Superpressure balloon. 13 flights. 50+ day duration. 18 km float level.

644 MIST sondes with P, T, RH, GPS winds (equivalent to Dropsonde sensor suite)

Enhanced web-based ground control to schedule automatic drops and display position and

sounding data

530

32

Figure Captions 531

Figure 1: The Driftsonde System concept 532

Figure 2: AMMA Driftsonde flight train 533

Figure 3: Driftsonde gondola 534

Figure 4: Sample Driftsonde temperature profiles, with the drop altitudes of Driftsonde and the 535

T-PARC aircraft 536

Figure 5: Four images of Driftsonde deployments. (a) Launch from McMurdo Station, Antarctica 537

during Concordiasi (2010); (b) Shortly after launch during Concordiasi. Dropsondes are visible at 538

perimeter of the Driftsonde gondola below the CNES superpressure balloon; (c) Launch from 539

Hawaii during T-PARC (2008). The Driftsonde gondola is on its deployment sled; (d) Driftsonde 540

before launch during AMMA (2006). This earlier version of the Driftsonde gondola was 541

constructed from cardboard rather than hard foam. 542

Figure 6: Trajectories of the 8 Driftsondes launched from Zinder, Niger, located 640 km east of 543

Niamey. Squares indicate the location of a dropsonde release. 544

Figure 7: Vertical profiles of relative humidity (a), temperature difference (b), zonal wind (c) and 545

meridional wind (d) from the ARPEGE (ARP) and ECMWF Integrated Forecast System (IFS) 546

model analyses co-located with dropsondes (DROP) within the Saharan Air Layer. 547

Figure 8: Locations of dropsondes deployed from the 16 Driftsonde balloons launched from the 548

Big Island of Hawaii during T-PARC. Squares indicate the location of a dropsonde release. 549

33

Figure 9: Infrared satellite image of a tropical disturbance at 0500 UTC 29 August 2008 over 550

which the Driftsonde (yellow drift track, progressing east to west) was deploying dropsondes 551

(yellow balloon symbols). The Air Force C-130 (black flight track) also released dropsondes for 552

this event (black/white squares). Driftsonde releases occurred between 00 and 09 UTC, while 553

Aircraft releases were between 00 and 05 UTC. The NRL P-3 aircraft (not shown) also collected 554

data in the environment surrounding this disturbance. 555

Fig. 10: Locations of all 644 dropsondes (yellow squares) released over Antarctica during 556

Concordiasi and radiosonde stations (red circles) on the Antarctic continent (top); Flight tracks 557

of the 13 Driftsondes for the duration of the project (middle); and tracks of the Driftsondes 558

constellation during a single 12 hour period on November 26, 2010 showing the wide 559

distribution of possible release points (bottom). 560

Fig. 11 Mean (solid line) and standard deviation (dashed line) of temperature differences 561

between satellite and dropsonde (red line) or radiosonde (black line) data for NOAA IASI 562

instrument (left panel) and AIRS (right panel) products. 563

Figure 12: The difference between the dropsonde and the model temperatures at the lowest 564

model level for the French global model. Blue colors indicate that the dropsonde is colder than 565

the model, and red colors indicate that it is warmer. 566

Figure 13: The total impact of dropsonde temperature (T) and wind is illustrated at high levels 567

(pressure less than 400 hPa) and low levels (pressure less 400 hPa) in the left panel, together 568

with the number of observations in the right panel. Negative values of the impact indicate that 569

the data contribute to reducing the error in the NWP system. For observations, each individual 570

34

datum on each pressure level is counted; for winds, zonal and meridional components are 571

counted separately. The impact has been measured using adjoint-based sensitivity of the 24-572

hour forecast error with respect to observations. The forecast error has been defined using a 573

dry total energy norm over the polar area (south of 60oS) and from the surface to the top of the 574

model. The linear estimation has been computed using a second-order approximation. 575

35

Figures 576

577

Figure 1: The Driftsonde System concept 578

36

579

Figure 2: AMMA Driftsonde flight train 580

37

581

Figure 3: Driftsonde gondola 582

System Electronics

Lithium Batteries

Battery & Elect. Heater

Iridium Modem

MIST Sondes

GPS & Iridium Antenna

400 MHz Sonde Antenna

Solar Panels for Heater

NOT TO SCALE

Harness

38

583

Figure 4: Sample Driftsonde temperature profiles, with the drop altitudes of Driftsonde and the 584

T-PARC aircraft 585

39

586

Figure 5: Four images of Driftsonde deployments. (a) Launch from McMurdo Station, Antarctica 587

during Concordiasi (2010); (b) Shortly after launch during Concordiasi. Dropsondes are visible at 588

perimeter of the Driftsonde gondola below the CNES superpressure balloon; (c) Launch from 589

Hawaii during T-PARC (2008). The Driftsonde gondola is on its deployment sled; (d) Driftsonde 590

before launch during AMMA (2006). This earlier version of the Driftsonde gondola was 591

constructed from cardboard rather than hard foam. 592

d

a

c

b

40

593

Figure 6: Trajectories of the 8 Driftsondes launched from Zinder, Niger, located 640 km east of 594

Niamey. Squares indicate the location of a dropsonde release. 595

41

596

Figure 7: Vertical profiles of relative humidity (a), temperature difference (b), zonal wind (c) and 597

meridional wind (d) from the ARPEGE (ARP) and ECMWF Integrated Forecast System (IFS) 598

model analyses co-located with dropsondes (DROP) within the Saharan Air Layer. 599

Meridonal wind (m s-1)

Zonal wind (m s-1)

Temperature difference (K)

Relative Humidity (%)

42

600

Figure 8: Locations of dropsondes deployed from the 16 Driftsonde balloons launched from the 601

Big Island of Hawaii during T-PARC. Squares indicate the location of a dropsonde release. 602

43

603

Figure 9: Infrared satellite image of a tropical disturbance at 0500 UTC 29 August 2008 over 604

which the Driftsonde (yellow drift track, progressing east to west) was deploying dropsondes 605

(yellow balloon symbols). The Air Force C-130 (black flight track) also released dropsondes for 606

this event (black/white squares). Driftsonde releases occurred between 00 and 09 UTC, while 607

Aircraft releases were between 00 and 05 UTC. The NRL P-3 aircraft (not shown) also collected 608

data in the environment surrounding this disturbance. 609

44

610

611

612

Fig. 10: Locations of all 644 dropsondes (yellow squares) released over Antarctica during 613

Concordiasi and radiosonde stations (red circles) on the Antarctic continent (top); Flight tracks 614

of the 13 Driftsondes for the duration of the project (middle); and tracks of the Driftsondes 615

45

constellation during a single 12 hour period on November 26, 2010 showing the wide 616

distribution of possible release points (bottom). 617

46

618

Fig. 11 Mean (solid line) and standard deviation (dashed line) of temperature differences 619

between satellite and dropsonde (red line) or radiosonde (black line) data for NOAA IASI 620

instrument (left panel) and AIRS (right panel) products. 621

47

622

Figure 12: The difference between the dropsonde and the model temperatures at the lowest 623

model level for the French global model. Blue colors indicate that the dropsonde is colder than 624

the model, and red colors indicate that it is warmer. 625

48

626

Figure 13: The total impact of dropsonde temperature (T) and wind is illustrated at high levels 627

(pressure less than 400 hPa) and low levels (pressure less 400 hPa) in the left panel, together 628

with the number of observations in the right panel. Negative values of the impact indicate that 629

the data contribute to reducing the error in the NWP system. For observations, each individual 630

datum on each pressure level is counted; for winds, zonal and meridional components are 631

counted separately. The impact has been measured using adjoint-based sensitivity of the 24-632

hour forecast error with respect to observations. The forecast error has been defined using a 633

dry total energy norm over the polar area (south of 60oS) and from the surface to the top of the 634

model. The linear estimation has been computed using a second-order approximation. 635