Assessing the Forecast Impact of a Geostationary Microwave Sounder Using Regional and Global OSSEs

Derek J. Posselt aJet Propulsion Laboratory, California Institute of Technology, Pasadena, California

Search for other papers by Derek J. Posselt in
Current site
Google Scholar
PubMed
Close
,
Longtao Wu aJet Propulsion Laboratory, California Institute of Technology, Pasadena, California

Search for other papers by Longtao Wu in
Current site
Google Scholar
PubMed
Close
,
Mathias Schreier aJet Propulsion Laboratory, California Institute of Technology, Pasadena, California

Search for other papers by Mathias Schreier in
Current site
Google Scholar
PubMed
Close
,
Jacola Roman aJet Propulsion Laboratory, California Institute of Technology, Pasadena, California

Search for other papers by Jacola Roman in
Current site
Google Scholar
PubMed
Close
,
Masashi Minamide bUniversity of Tokyo, Tokyo, Japan

Search for other papers by Masashi Minamide in
Current site
Google Scholar
PubMed
Close
, and
Bjorn Lambrigtsen aJet Propulsion Laboratory, California Institute of Technology, Pasadena, California

Search for other papers by Bjorn Lambrigtsen in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Forecast observing system simulation experiments (OSSEs) are conducted to assess the potential impact of geostationary microwave (GeoMW) sounder observations on numerical weather prediction forecasts. A regional OSSE is conducted using a tropical cyclone (TC) case that is very similar to Hurricane Harvey (2017), as hurricanes are among the most devastating of weather-related natural disasters, and hurricane intensity continues to pose a significant challenge for numerical weather prediction. A global OSSE is conducted to assess the potential impact of a single GeoMW sounder centered over the continental United States versus two sounders positioned at the current locations of the National Oceanic and Atmospheric Administration Geostationary Operational Environmental Satellites (GOES) East and West. It is found that assimilation of GeoMW soundings result in better characterization of the TC environment, especially before and during intensification, which leads to significant improvements in forecasts of TC track and intensity. TC vertical structure (warm core thermal perturbation and horizontal wind distribution) is also substantially improved, as are the surface wind and precipitation extremes. In the global OSSE, assimilation of GeoMW soundings leads to slight improvement globally and significant improvement regionally, with regional impact equal to or greater than nearly all other observation types.

Significance Statement

This work seeks to determine the impact of a new geostationary microwave (GeoMW) sounder on tropical cyclone forecasts in particular, and on weather forecasts in general. It does so by assimilating simulated GeoMW sounder data into two different forecast models: one global and one regional. The data have a small positive impact globally, and a significant positive impact over the region viewed by the GeoMW instrument. In particular, assimilation of GeoMW data has a significant and positive impact on forecasts of tropical cyclone track, strength, and structure.

© 2022 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Derek J. Posselt, Derek.Posselt@jpl.nasa.gov

Abstract

Forecast observing system simulation experiments (OSSEs) are conducted to assess the potential impact of geostationary microwave (GeoMW) sounder observations on numerical weather prediction forecasts. A regional OSSE is conducted using a tropical cyclone (TC) case that is very similar to Hurricane Harvey (2017), as hurricanes are among the most devastating of weather-related natural disasters, and hurricane intensity continues to pose a significant challenge for numerical weather prediction. A global OSSE is conducted to assess the potential impact of a single GeoMW sounder centered over the continental United States versus two sounders positioned at the current locations of the National Oceanic and Atmospheric Administration Geostationary Operational Environmental Satellites (GOES) East and West. It is found that assimilation of GeoMW soundings result in better characterization of the TC environment, especially before and during intensification, which leads to significant improvements in forecasts of TC track and intensity. TC vertical structure (warm core thermal perturbation and horizontal wind distribution) is also substantially improved, as are the surface wind and precipitation extremes. In the global OSSE, assimilation of GeoMW soundings leads to slight improvement globally and significant improvement regionally, with regional impact equal to or greater than nearly all other observation types.

Significance Statement

This work seeks to determine the impact of a new geostationary microwave (GeoMW) sounder on tropical cyclone forecasts in particular, and on weather forecasts in general. It does so by assimilating simulated GeoMW sounder data into two different forecast models: one global and one regional. The data have a small positive impact globally, and a significant positive impact over the region viewed by the GeoMW instrument. In particular, assimilation of GeoMW data has a significant and positive impact on forecasts of tropical cyclone track, strength, and structure.

© 2022 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Derek J. Posselt, Derek.Posselt@jpl.nasa.gov

1. Introduction

The term observing system simulation experiment (OSSE) can be used to refer to any activity that is designed to evaluate a future observing system (Zeng et al. 2020). In the atmospheric sciences, OSSEs most commonly refer to evaluation of the potential impact of an observing system on a numerical weather prediction forecast (Hoffman and Atlas 2016). The key in any OSSE is to measure the information provided by a set of new observations, relative to the current observing system (if possible). All OSSEs, regardless of form, require a few key components:

  1. A reference dataset against which the observations may be compared. In a forecast OSSE, this is referred to as a nature run (NR), and consists of a free-running high fidelity (and often high resolution) simulation designed to mimic real-world processes and conditions as realistically as possible (Reale et al. 2007; Gelaro et al. 2015).

  2. Simulators capable of producing synthetic observations as if measurements had been made of the NR. For remote sensing measurements, this consists of one or more radiative transfer models, simulation of a retrieval system, or both. In any case, it is important to simulate not only the measurements, but also their errors and spatial and temporal resolution (Privé et al. 2021).

  3. A quantitative methodology with which the information in the candidate measurements can be assessed. In a forecast OSSE, this consists of a data assimilation and forecast system, and the impact of observations is assessed by comparing a forecast run without the new observations to the NR, then comparing a forecast that has assimilated the new observations to the NR.

Forecast OSSEs may be further subdivided into two general types: regional (McNoldy et al. 2017) and global (McCarty et al. 2021). Global forecast OSSEs evaluate the impact of an observing system on global numerical weather prediction. They have the advantage of measuring the global impact of a new set of observations versus all other currently available observations (provided they are simulated from the NR with appropriate error characteristics). The disadvantages of a global OSSE are the following: 1) global NR spatial and temporal resolution are typically not sufficient to represent fine-scale and high-impact weather, 2) global forecast models commonly have quite coarse spatial resolution and infrequent assimilation intervals, so that the impact of high temporal and/or spatial resolution data is limited, and 3) global NRs and forecast models are often run with simplified representations of cloud and precipitation processes, making it difficult to evaluate observation impact on key applications (e.g., water availability, flooding, etc.). In addition, the computational expense of running a global OSSE is high (often requiring a minimum of a month to run end-to-end).

Regional OSSEs utilize a limited area model for both the NR and forecast system, and assess the impact of measurements on specific regions and/or weather systems. They have the advantages of: 1) more realistic representation of high-impact weather systems in the NR and forecast model, 2) more effective use (in data assimilation) of high-resolution observations, and 3) more realistic treatment of parameterized physical processes. However, because of their limited regional focus, and the fact that they are often run over shorter periods of time, regional OSSEs may not represent the full (global) impact of the measurements. Because they are typically limited to case studies, it can also be more difficult to assess the significance of regional OSSE results.

During the latter half of 2020, the National Oceanographic and Atmospheric Administration (NOAA) commissioned several studies to assess the prospective performance of a number of new geostationary observing concepts. One such concept was a geostationary microwave (GeoMW) sounder, based on the Geostationary Synthetic Thinned Aperture Radiometer (GeoSTAR) instrument developed at the Jet Propulsion Laboratory. In contrast to hyperspectral infrared sounders, a GeoMW sounder has the advantage of being able to produce estimates of atmospheric temperature and water vapor profiles in clear, cloudy, and light rain conditions. Such observations have long been available from low Earth orbiting spacecraft, but the temporal revisit is relatively long (every 12 h). In contrast, a GeoMW sounder has the potential to take observations as frequently as every 15 min, providing rapid refresh observations of fast evolving weather events.

As part of the NOAA study, we conducted both regional and global OSSEs to assess the impact of geostationary microwave soundings on numerical weather prediction forecasts. Because the study was done with NOAA’s future geostationary observing system in mind, we assessed both a single and dual satellite configuration in our global OSSE (with the dual satellite configuration corresponding to the approximate positions of the current Geostationary Operational Environmental Satellites; GOES). Because, as was noted above, global OSSEs require significant computational resources and time, we conducted only two GeoMW global OSSEs. The first OSSE uses a single GeoMW instrument centered at 90°W longitude and with the instrument field of view angled slightly northward so that the center of the viewing region is located at 15°N latitude. The second examines the impact of two GeoMW instruments, one stationed at the position of GOES-East and the other at the position of GOES-West, each with the field of view of the instrument centered (pointed) at 15°N latitude.

In addition to the global OSSEs, we also conducted several regional OSSEs. These explored the impact of GeoMW assimilation on forecasts of a strong tropical cyclone. We describe each experiment in more detail below, but in brief, we performed the following experiments:

  1. We produced a new hurricane nature run by initializing the Weather Research and Forecasting (WRF) Model using initial conditions from a successful ensemble forecast of Hurricane Harvey (2017) that had assimilated all-sky satellite radiances from GOES-16 (Minamide and Zhang 2019).

  2. We then ran several forecast experiments using a coarser resolution version of the WRF and with different parameterizations and initial and boundary conditions, and with assimilation of observations in an ensemble Kalman filter system:

  3. A free-running forecast initialized from NOAA NCEP Global Forecast System (GFS) initial conditions.

  4. An experiment assimilating only conventional observations (see description below)

  5. Several experiments in which conventional + GeoMW observations were assimilated.

The remainder of this paper is organized as follows. Since synthetic GeoMW data are used in a similar manner in both regional and global OSSEs, we begin by providing a brief description of how the synthetic GeoMW data are generated, along with the procedure we use to ensure that it contains similar information expected from the on-orbit sensor. We then provide a brief description of the two OSSEs, focusing primarily on the regional OSSE, as the global system has been extensively documented elsewhere. Since we have a more comprehensive library of results from the regional OSSEs, we describe these experiments first. Next, we present results from the global OSSEs then conclude with a brief summary and discussion.

2. Data and methods

a. Production of synthetic GeoMW profiles

For any forecast OSSE to be credible (e.g., to return an observation impact consistent with the real observation impact), there are two key requirements. First, the NR (reference state) must realistically represent the processes and features of interest, and should be as different as possible from the forecast (assimilating) model, since the real atmosphere will surely differ from the numerical forecast. Second, the measurement uncertainties must be as consistent with the real measurement uncertainties as possible, including errors due to spatial scales/misrepresentation, instrument noise, and forward model/retrieval errors and uncertainties. We provide a description of the nature runs and forecast models below, and here focus on the simulated observations.

In modern data assimilation systems, when ingesting observations from spaceborne remote sensing platforms it is common to assimilate radiances (or brightness temperatures) rather than retrieved geophysical variables (e.g., temperature and water vapor profiles). This is because retrievals contain information not only from the measurements, but also from the retrieval system. Since the retrieval system typically incorporates ancillary information (e.g., from climatology), and many retrievals do not effectively propagate uncertainties, there may be unknown sources of bias or random error that make it difficult to successfully utilize the information in the measurements. For this reason, it is typically viewed as more internally consistent to use an observation simulator to calculate satellite brightness temperatures from the model, and then assimilate measured brightness temperatures. In an OSSE, this is not necessarily the right choice, for a number of reasons. First, assimilation of brightness temperatures requires tuning of the radiative transfer model, both for forward simulating the observations from the model state, and also for mapping from observations back to state space. This process is time consuming and commonly requires several iterations before observations can be effectively utilized in the data assimilation system. Second, in an OSSE, the true state is available (from the NR) typically at higher spatial resolution than the remote sensing measurements. If the retrieval system is available such that the resolution of the NR temperature and water vapor profiles can be suitably coarsened in the horizontal (e.g., using the instrument point spread function) and also in the vertical (using estimates of the information content and weighting functions from the retrieval) and assigned appropriate uncertainties, then assimilation of retrievals is more straightforward and likely a more realistic option.

In our experiments, we used a hybrid approach. First, we used an instrument simulator (the NASA Earth Observing System Simulator Suite, NEOS3; Tanelli et al. 2012) to compute synthetic brightness temperatures from the NR. We then used a Bayesian optimal estimation (OE) system, which has heritage in MW thermodynamic profile and rain rate estimates (Ferraro et al. 2000; Rosenkranz 2001; Laviola and Levizzani 2011) from the Advanced Microwave Sounding Unit (AMSU), Advanced Technology Microwave Sounder (ATMS), and the High Altitude MMIC Sounding Radiometer (HAMSR), to estimate vertical averaging kernels and retrieval errors. Both the NEOS3 and the retrieval system account for scattering of microwave radiation by clouds and precipitation hydrometeors. The OE-provided averaging kernels and corresponding degrees of freedom were used to blur the NR temperature and water vapor profiles in the vertical. We select discrete layers for assimilation with layer locations equal to the peaks in the retrieval averaging kernels. Profiles are averaged in the horizontal using a Gaussian point spread function with 25 km full-width-at-half-max, consistent with the GeoMW spatial resolution. The specific vertical levels assimilated are 850, 700, 600, 450, 300 hPa, and the uncertainties in temperature and relative humidity are listed in Table 1 and are derived from the MW retrieval system. The combination of horizontal and vertical blurring, along with uncertainties estimated from the OE retrieval system, provides a set of synthetic observations with information content that is expected to be very similar to the on-orbit GeoMW performance.

Table 1

The uncertainties in temperature, relative humidity, and wind for data assimilated in the regional and global OSSEs.

Table 1

Assimilation of temperature and water vapor profiles forms the basis for nearly all of our OSSEs; however, we also run two additional regional OSSEs; one in which we assimilate wind vectors that are assumed to be obtained by tracking retrieved water vapor fields, and another in which we assimilate the TC minimum central pressure using a microwave sounder-based estimate. Note that the AMV assimilation is based on previous experiments in which we have documented the capability of obtaining AMVs from time sequences of synthetic MW measurements (Lambrigtsen et al. 2018; Posselt et al. 2019). In our experiments, synthetic AMVs are located on the same levels as the temperature and relative humidity variables but at 100-km horizontal spacing. More information on the AMV tracking algorithm and method for estimating uncertainties can be found in Lambrigtsen et al. (2018) and Posselt et al. (2019).

b. Regional OSSEs using the WRF-EnKF system

While we do use the WRF Model for both NR and forecast, we ensure differences among them by running the forecast model with: an older model version, coarser spatial resolution, simpler microphysical parameterization, and a different source of initial conditions. The comparison among the NR and forecast model characteristics is presented in Table 2.

Table 2

Details of the model configurations used for the NR and the forecast model.

Table 2

1) Hurricane nature run

The NR in our OSSE consists of a free-running simulation on four two-way nested domains (Fig. 1a) using version 3.9.1 of the WRF Model. The simulation is initialized at 0000 UTC 23 August 2017 from the initial state that produced the third strongest member of an ensemble forecast of Hurricane Harvey, whose ensemble initial conditions were created by assimilating the conventional set of observations and all-sky satellite brightness temperatures (BTs) channel 8 from GOES-16 (similarly in Minamide et al. 2020; Zhang et al. 2019; Fig. 2). The NR simulation is run for 5 days (until 0000 UTC 28 August 2017), and is driven on the boundaries of the outermost domain by analysis fields obtained from the fifth-generation European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis (ERA5). The NR produces a category-5 hurricane, with minimum sea level pressure less than 920 hPa and maximum 10-m winds in excess of 80 m s−1 (>155 kt). In addition, it rapidly intensifies; during the 24 h between 1200 UTC 24 August and 1200 UTC 25 August, the minimum sea level pressure decreases by >40 hPa and the storm intensity increases from category 1 to 4. Storm structure on the highest-resolution (innermost) domain is highly realistic, both in terms of wind and humidity (Fig. 3).

Fig. 1.
Fig. 1.

Depiction of the WRF domains used for the (a) nature run and (b) forecast system. Note that the outermost domain (d01) remains fixed in the same position while the inner domains [d02, d03, and (in the nature run) d04] follow the position of the tropical cyclone. As such, the position of the inner domains is valid at the time shown (1200 UTC 23 Aug 2017), but will have moved to a different position later in the simulations. Color shading depicts the topography, and any differences are due to the differences in the ECMWF vs GFS boundary conditions.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

Fig. 2.
Fig. 2.

(a) Minimum sea level pressure and (b) maximum 10-m wind speed from a 60-member ensemble forecast of Hurricane Harvey (2017) initialized from a WRF-EnKF analysis. The control forecast initialized from the ensemble mean is shown in the black dashed line, while the solid black line with cross marks represents the National Hurricane Center best track analysis. The ensemble member whose initial conditions are used to initialize the nature run is indicated in gray, while the NR intensity is overlaid as a thick black line.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

Fig. 3.
Fig. 3.

Output from the NR, valid at 1200 UTC 25 Aug 2017. (left) Wind speed (color contours) and vectors (arrows) on the 750-hPa pressure level. (right) Relative humidity (percent; color contours) on the 750-hPa pressure level.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

2) WRF Model and ensemble Kalman filter data assimilation

We utilize an older version (3.6.0) of the WRF Model, run on a different (coarser resolution) domain (Fig. 1b) and with different physical parameterizations and boundary conditions, as our forecast and assimilating model (Table 2). This model version and configuration is also nearly identical to the one that has been used for the convective-scale hurricane initializations in previous studies that have successfully assimilated geostationary infrared brightness temperatures and airborne Doppler radar winds (Minamide and Zhang 2017, 2018, 2019; Zhang et al. 2009, 2016). We refer the reader to these references for more detail on the model configuration and performance. An ensemble Kalman filter (EnKF) is used to assimilate conventional data and GeoMW profiles. We utilize the same conventional dataset as has been used in previous studies (surface METARs, ship observations, radiosondes, and satellite winds), but rather than using real observations, we instead simulate observations from the NR. This is done by first extracting data from locations identical to the real measurements, then applying observation errors consistent with uncertainties estimated for the real measurements. We utilize a 60-member ensemble for the EnKF, and initialize from an initial state obtained from the NCEP GFS analysis valid at 0000 UTC 23 August 2017. The initial ensemble is generated using the WRF data assimilation system’s RANDOMCV methodology (so-called CV 3 option that is described in the WRF Users Guide, Chapter 6, Barker et al. 2004), which applies random perturbations to the model initial conditions in control variable space.

3) Description of forecast and data assimilation experiments

(i) Baseline experiments

As a baseline, we run two experiments that do not assimilate synthetic GeoMW temperature and water vapor profiles; one in which no observations are assimilated (NoDA), and another in which only “conventional data” are assimilated (Conv or Conventional). Conv is meant to replicate the results that would be obtained when all currently available conventional observations are assimilated, including ships, surface METARs (where available), radiosondes, aircraft, and atmospheric motion vectors obtained from geostationary satellite data (SATWIND). In a global OSSE, most of the assimilated measurements come from spaceborne observing platforms. However, in our regional OSSEs, the domain is nearly entirely over the Gulf of Mexico during the assimilation time period. Because of the regional domain, the relatively short assimilation time interval (24 h), and the intermittent overpass times of low Earth orbiting satellites, we do not assimilate any current LEO satellite data. We expect that the impact of GeoMW assimilation may be tempered a bit if LEO satellite data were assimilated.

(ii) Geostationary MW experiments

As mentioned above, the majority of our experiments assimilate temperature and water vapor profiles consistent with GeoMW retrievals. In contrast to infrared sounders, microwave sounders are able to retrieve profiles in cloudy regions (though the presence of precipitation increases the retrieval uncertainty). To analyze the effect of assimilating information in cloudy and precipitating regions, we conduct the following experiments:

  1. Only assimilate profiles in clear-sky regions—those with broadband outgoing longwave radiation > 220 W m−2. This effectively screens out all clouds above the boundary layer.

  2. Assimilate all clear-sky profiles (as in experiment 1), and add profiles in cloudy regions with precipitation rates up to 1 mm h−1. The uncertainties in cloudy and lightly precipitating regions are expected to be the same as those for cloud-free profiles, and are based on level-dependent error estimates from the retrieval system.

  3. Assimilate clear sky profiles as in experiment 1, and add profiles in cloud regions with precipitation rates up to 10 mm h−1. This is the maximum rain rate under which we expect to be able to retrieve temperature and water vapor. To account for increasing error in the retrievals, we inflate the assumed uncertainty in the observations according to precipitation rate, with 0.1–1.0 mm h−1 = 2x retrieval error estimate, 1.0–10 mm h−1 = 3x retrieval error estimate.

  4. A visual depiction of the locations of the boundaries of the OLR < 220 W m−2 region, as well as the precipitation rates exceeding 1 and 10 mm h−1 can be found in Fig. 4. As mentioned above, we also run two additional experiments:

  5. As in experiment 2 above, but also assimilate synthetic geostationary atmospheric motion vectors. These consist of winds obtained from the nature run, blurred to the GeoMW horizontal and vertical resolution, and then thinned to 100 km horizontal spacing. Errors are listed in Table 1.

  6. As in experiment 2 above, but also assimilate estimates of the tropical cyclone central pressure obtained from the GeoMW-estimated TC warm core thermal perturbation. Uncertainties on the minimum sea level pressure (SLP) estimates are set to 11.9 hPa.

Fig. 4.
Fig. 4.

Plot of simulated outgoing longwave radiation (W m−2; gray shading) at 0600 UTC 24 Aug 2017 on the innermost domain of the nature run (d04). Overlaid are contours encircling regions with OLR < 220 W m−2 (red), and precipitation rate > 1 mm h−1 (blue) and >10 mm h−1 (orange).

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

Experiments 1–5 assimilate GeoMW data hourly. We also conduct a sixth experiment, in which experiment 2 is repeated with 15-min data assimilation intervals. This is the highest frequency with which we can retrieve temperature and water vapor and maintain the expected uncertainties and horizontal resolution. Note that GeoMW observations can be made as frequently as every 1.5 min, but the measurements will have greater noise.

(iii) Validation of the nature run and regional OSSE

One way in which an OSSE system may be validated is by comparing the impact of assimilating synthetic conventional observations from the nature run versus assimilation of real conventional observations. The impact is commonly measured via the use of the observation minus background (OB) statistics, with larger departures associated with larger impact. We have computed the radiosonde OB statistics for a set of experiments that assimilated real conventional observations for every named storm in the 2017 Atlantic hurricane season. We have also computed the OB differences every 3 h from our assimilation of synthetic radiosonde data from the nature run. Comparison between the real and synthetic OB statistics (Fig. 5) indicates the impact of assimilating nature run based observations is quite consistent with the impact of assimilating real observations.

Fig. 5.
Fig. 5.

Observation minus background (OB) statistics for wind vector components, temperature, and water vapor from assimilation of real radiosonde data for all named storms in the 2017 hurricane season (box-and-whisker plots) and from the conventional assimilation experiment in our study (green circles).

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

c. Description of the global OSSE system

Our global OSSE uses the NASA Goddard Global Modeling and Assimilation Office (GMAO) OSSE system described in Errico et al. (2017), which is based on a global 7-km grid spacing nature run with 72 vertical levels that spans two simulated years starting 1 May 2005. While the lower boundary conditions are from 2005 to 2006, the simulation is free running and is not intended to represent the real state of the atmosphere during the 2-yr integration. The NR data are available every 30 min, and the results have been extensively validated (Gelaro et al. 2015). Simulated observations from surface, ship, radiosondes, aircraft, and satellites consistent with the 2015 operational observing system are extracted from the NR, and the coverage and uncertainty have been calibrated to match the statistics of real observations so that the simulated data impact is the same as the impact of real observations (Errico et al. 2017).

The model used to assimilate both the 2015 observing system and the GeoMW observations is the Global Earth Observing System (GEOS) model (Molod et al. 2015) atmospheric data assimilation system (Rienecker et al. 2008). The model is run on a cubed-sphere domain with 25-km horizontal grid spacing and 72 vertical levels. Data assimilation is conducted using 3D-Var via the NCEP Gridpoint Statistical Analysis system (GSI; Kleist et al. 2009). While an upgrade has recently been made to the GMAO OSSE system to use the NCEP four-dimensional ensemble variational (4DEnVar) system, the new configuration had not yet been tested and validated prior to our experiments.

In contrast to the regional OSSEs, which assumed that the GeoMW field of regard would follow the hurricane track, we fixed the location of the GeoMW viewing domain and used a larger (5000-km diameter) field of regard in the global OSSEs. In the single GeoMW configuration, the field of view is centered at 15°N latitude and 90°W longitude and covers the Gulf of Mexico and much of the southern and southeastern continental United States (Fig. 6a). In the dual GeoMW configuration, the eastern and western fields of view are centered at 15°N latitude and at 75° and 135°W longitude, respectively (Fig. 6b). As in the regional OSSE experiments, we extract GeoMW profiles of temperature and water vapor in regions with precipitation rates less than 1 mm h−1, and average the NR fields to 25-km spatial resolution. While GeoMW data are expected to be available hourly, the GMAO OSSE system conducts data assimilation every 6 h. Our experiments assimilate data every 6 h from 23 June 2006 to 22 August 2006 (2 months). This spans the period for which calibrated observations were available for the GMAO OSSE system. Errors in the GeoMW soundings were specified as in the regional OSSEs (Table 1). In section 4 below, we describe the results of two OSSEs, corresponding to single and dual satellite configurations.

Fig. 6.
Fig. 6.

Plots showing the viewing regions (red) for the (a) single GeoMW configuration and (b) dual GeoMW configuration. Overlaid in red dots are the locations of the GeoMW observations assimilated at 0000 UTC 1 Jul 2006.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

3. Regional OSSE results

a. TC intensity

Results from a control forecast (no DA), as well as a forecast that assimilates conventional data, are shown in Fig. 7, and it can be seen that neither is able to successfully capture the NR hurricane development. In contrast, assimilation of GeoMW soundings resulted in demonstrable improvement (Fig. 8), with the clear-sky-only experiment (Figs. 8a,b) producing a relatively weaker storm compared with assimilation of all-sky profiles without (Figs. 8c,d) and with (Figs. 8e,f) precipitation dependent error. Assimilation of profiles in regions with larger rain rates (>1 mm h−1; Figs. 8e,f) led to little improvement at later initialization times (0000 and 0600 UTC 24 August; darker blue lines), but significantly improved storm forecasts at longer lead times (1200 and 1800 UTC 23 August; lighter blue lines). In contrast, assimilation of water vapor AMVs led to larger improvements at later times (Figs. 9a,b), while assimilation of tropical cyclone central pressure estimates (Figs. 9c,d) led to significant improvements at all lead times. The most realistic (strongest) storm was produced from assimilation of 15-min interval GeoMW data (Figs. 9e,f), with the storm reaching strong category-4 intensity and with intensification rates very similar to the nature run.

Fig. 7.
Fig. 7.

Time series of (a) minimum sea level pressure; and (b) maximum 10-m wind speed for the hurricane NR (black), free-running forecast (noDA; blue colors), and forecast that assimilates conventional observations (warm colors). Different colors correspond to different initialization times, every 6 h starting at 1200 UTC 23 Aug 2017 and ending at 0600 UTC 24 Aug 2017.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

Fig. 8.
Fig. 8.

Comparisons of forecast intensity in terms of (a),(c),(e) minimum sea level pressure and (b),(d),(f) maximum 10-m wind speed for the nature run (black), assimilation of conventional observations (warm colors) and assimilation of three different GeoMW configurations (cool colors). (top) Clear-sky profiles only, (middle) all-sky profiles, with precipitation rates < 1 mm h−1, and (bottom) all-sky profiles, with precipitation rates up to 10 mm h−1 and precipitation-dependent error.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

Fig. 9.
Fig. 9.

Colors as in Fig. 4. As in the middle panel in Fig. 4, all experiments shown assimilate temperature and water vapor profiles with clear-sky errors in all-sky conditions with precipitation rates up to 1 mm h−1. However, in this case, we make the following modifications. (a),(b) Add assimilation of atmospheric motion vectors; (c),(d) add assimilation of minimum sea level pressure estimates; and (e),(f) assimilate temperature and water vapor profiles every 15 min.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

b. TC track error

Another metric commonly used to assess tropical cyclone forecast accuracy is the track position of the center of the storm, commonly represented as a deviation (or error) from the reference (or best track) position. In our case, we have exact knowledge of the storm center position from the NR, and compute hourly great circle distances between the NR TC center and the TC center in each of our data assimilation experiments. Plots of the track error are shown in Fig. 10 for each of the experiments, and it can be seen that track errors in the experiment with conventional data assimilated (warm colors in each plot) are on the order of 50–300 km. Assimilation of GeoMW data has a similar effect on track for all configurations tested, with track errors smallest for experiments that assimilated AMVs in addition to temperature and water vapor profiles. Note that the majority of the impact on the storm track comes from assimilation of clear-sky profiles, as track errors for the clear-sky-only assimilation (Fig. 10a) are comparable to the track errors for the all-sky assimilation (Figs. 10b–e).

Fig. 10.
Fig. 10.

Comparisons of forecast track error for the control run (assimilating conventional observations; warm colors) and assimilation of various GeoMW configurations (cool colors). (a) Assimilation of clear-sky-only T and RH profiles. (b) Assimilation of all-sky T and RH profiles in regions with precipitation < 1 mm h−1. (c) Assimilation of all-sky T and RH profiles in regions with precipitation < 10 mm h−1 and with precipitation-dependent error. (d) As in (b), but with the addition of water vapor AMVs. (e) As in (b), but with the addition of GeoMW-based minimum SLP estimates.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

c. Storm vertical structure

In addition to metrics of storm intensity, it is also useful to examine the storm vertical structure. Two common measures are the temperature deviation from the mean, which reveals the positive temperature perturbation in the storm’s inner core, and the tangential and radial components of the wind. We have computed both of these metrics for the NR, control simulation, and each of the GeoMW experiments. For compactness, we plot only the 48-h forecast results from the 0000 UTC 24 August 2017 initialization time, valid 48 h later (0000 UTC 26 August 2017).

Examination of the temperature perturbation (Fig. 11a, top row) reveals a strong positive thermal anomaly in the upper troposphere in the NR, consistent with observations of strong tropical cyclones. Near the time of peak intensity at 0000 UTC 26 August, the perturbation is deep and strong. A weaker negative perturbation is evident near the surface, consistent with low-level cooling due to evaporating precipitation. By comparison, while the conventional assimilation experiment does produce a warm core anomaly, it is relatively weak and peaks at a lower level (Fig. 11b, bottom row). In addition, the near-surface negative temperature perturbation is more widespread, extending to greater than 300 km from the storm center. Assimilation of GeoMW profiles results in improvements to storm structure in all cases, though the 15-min assimilation cadence appears to produce the storm structure that best matches the nature run.

Fig. 11.
Fig. 11.

Color contours of (a) warm core temperature perturbation and (b) warm core temperature deviation from the nature run for the 48-h forecast initialized at 0000 UTC 24 Aug 2017 and valid at 0000 UTC 26 Aug 2017. Experiment labels can be found atop each of the subplots.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

We have computed the radial and azimuthal winds using the procedure documented by Ahern and Cowan (2018). By convention, the azimuthal winds are positive in the cyclonic (counterclockwise) direction and the radial winds are positive in the direction pointed away from the cyclone center. The NR mean azimuthal winds (Fig. 12a, top row) exhibit a deep region of strong cyclonic flow extending from near the storm center to a radius of 100 km, with weak winds aloft and anticyclonic flow at a distance of 350 km and greater, reflecting the presence of upper-level outflow. Consistent with the thermal perturbations, the conventional experiment’s cyclonic winds are too weak (Fig. 12b, second row), and do not extend high enough into the troposphere. Each GeoMW assimilation experiment produces improved wind structure, with the 15-min assimilation cadence case, and the case in which TC central pressure is assimilated, producing the most realistic winds. The radial wind structure in the nature run (Fig. 12c, third row) shows shallow strong inflow at low levels and deep outflow with strongest radial winds in the upper troposphere. All forecasts have weaker inflow (Fig. 12d, bottom row), but GeoMW assimilation yields stronger inflow in all cases. All experiments produce similar outflow structure, with all GeoMW experiments producing improvements over the conventional DA experiment.

Fig. 12.
Fig. 12.

Color contours of (a) azimuthal wind, (b) azimuthal wind deviation from the nature run, (c) radial wind, and (d) radial wind deviation from the nature run. As in Fig. 10, all fields are from the 48-h forecast initialized at 0000 UTC 24 Aug 2017 and valid at 0000 UTC 26 Aug 2017. Experiment labels can be found atop each of the subplots.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

d. Effect on the storm environment

In addition to the effect of assimilating GeoMW profiles on the tropical cyclone characteristics, we also wish to know whether profile assimilation resulted in improvements in the storm environment. We measure this by computing the root mean squared error (RMSE) for the mean profiles of temperature, water vapor, and wind between each forecast experiment and the NR. To ensure consistency, we compute RMSE between the forecast and the 3-km resolution NR domain (domain 03, Fig. 1a), and reduce the size of the NR domain to match the forecast domain. The outcomes are quite similar for all GeoMW assimilation experiments; hence, we show only the results from the all-sky GeoMW assimilation experiment in which precipitation rates are limited to <1 mm h−1. The time series of domain-average RMSE for temperature, water vapor, and wind components from 1200 UTC 24 August to 0000 UTC 27 August 2017 is shown in Fig. 13.

Fig. 13.
Fig. 13.

Time series of (inner; d03) domain averaged root mean squared error (RMSE) between the all-sky GeoMW assimilation experiment vs domain 03 of the NR, reduced to the size and position of domain 03 of the forecast. Temperature, specific humidity, and zonal (u direction), and meridional (υ direction) winds are shown from top to bottom. In each panel, red represents the RMSE for the conventional forecast while blue represents the RMSE for the GeoMW assimilation experiment(s).

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

It can be seen that GeoMW data assimilation results in small improvements to environmental temperature, and a slight degradation in water vapor. The most notable improvement is to the environmental winds, for which the RMSE decreases by more than 50%. Analysis of the mean vertical profiles of RMSE over the same time period (Fig. 14) reveals that the temperature improvements are maximized in the lower and upper troposphere, with a slight degradation in the middle troposphere (400–600 hPa), and that these improvements are not statistically significant. Specific humidity improvements are concentrated in the lower to middle free troposphere (∼700 hPa) with small (but statistically significant) degradations near the surface and in the upper troposphere. Wind improvements are largest in the middle and upper troposphere between 750 and 250 hPa, with a decrease of 50% for zonal wind RMSE and 72% for meridional wind RMSE.

Fig. 14.
Fig. 14.

Vertical profiles of (inner; d03) domain and time (1200 UTC 24 Aug–0000 UTC 27 Aug) averaged root mean squared error (RMSE) between each GeoMW assimilation experiment vs domain 03 of the NR, reduced to the size and position of domain 03 of the forecast. Shown are the RMSE profiles for temperature, specific humidity, and zonal (u direction), and meridional (υ direction) winds. As in Fig. 8, in each panel, red represents the RMSE for the conventional forecast while blue represents the RMSE for the GeoMW assimilation experiment(s). Percent deviations between the GeoMW assimilation experiment and the control are represented as numbers in the top-right corner of each plot, with an asterisk (*) indicating the differences are statistically significant.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

While the storm structure and intensity are both improved by assimilation of GeoMW profiles, there are small, but statistically significant, degradations in the representation of the specific humidity. It is possible that the coarser resolution representation of convection in the assimilating model resulted in errors in the vertical transport of water vapor, but additional analysis would be needed to determine the specific processes involved.

e. Rain rate and surface wind speed statistics

The impact of a tropical cyclone on human lives and infrastructure is expressed through the strong winds and large amounts of precipitation it produces. It is the extremes in wind and precipitation that often cause the most damage. We measure the effect of assimilation of GeoMW data on precipitation rates, and on the 99th percentile of the near surface winds to determine whether there are improvements in rain and extreme wind. Figure 15 shows the histogram of rain rate over the 3-km inner domain in the WRF forecast, and over the corresponding region in the nature run 3-km domain. It is clear that the conventional experiment produces rain rates that are generally too large at early forecast times (1200 UTC 24 August–1200 UTC 25 August), and rain rates that are not strong enough (and not frequent enough) at 0000 UTC 26 August. The GeoMW assimilation experiment also produces too much heavy rain at earlier times, and in fact produces even more heavy rain than in the conventional DA experiment in the 12-h forecast. However, assimilation of GeoMW data results in a larger number of grid points with rain, and an improved overall distribution of rain at 0000 UTC 26 August. The smaller number of light and moderate rain rates (1–10 mm h−1) in both control and GeoMW runs is likely due to the coarser resolution of the forecast model. Note that, while we compared the WRF forecasts versus the 3-km grid spacing nature run domain (D03) over an identical geographic region, the values on the 3-km grid spacing NR domain are actually overwritten by data calculated on the finer resolution 1-km domain (D04) via the WRF Model’s two-way nesting procedure. As such, while the grid reporting interval and comparison domain is identical between forecast and nature run, the effective resolution of the processes that produced the nature run data are 3 times finer.

Fig. 15.
Fig. 15.

Histograms of precipitation rate from the nature run (black), control forecast (red), and GeoMW assimilation experiment (blue) at four different forecast times, each separated by 12 h (times are listed on the right side of each panel). The results are derived from the forecast initialized at 0000 UTC 24 Aug 2017.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

Examination of the wind extremes (Fig. 16) reveals that the no DA and conventional assimilation experiment are unable to reproduce the strong winds in the storm, especially at early times (prior to 0000 UTC 26 August). Assimilation of GeoMW data improves the wind speed statistics, with 15-min assimilation yielding a very close match to the nature run wind extremes through most of the storm’s evolution.

Fig. 16.
Fig. 16.

Time series plot of the 99th percentile of the near-surface wind speeds in the nature run (black), no-DA experiment (blue), control experiment (green), GeoMW with (yellow), and without (red) minSLP assimilation, and GeoMW 15-min assimilation frequency (purple).

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

4. Global OSSE results

a. Assimilation of GeoMW thermodynamic profiles, single spacecraft configuration

Figure 17 depicts the impact of GeoMW profile assimilation on the RMSE of temperature, water vapor, and the wind components. RMSE is computed between the GEOS analysis and the nature run over the GeoMW field of view (Fig. 6a), and for the entire 2-month assimilation period. It can be seen that GeoMW temperature and water vapor profile assimilation has a small positive impact on the temperature and water vapor, and a modest impact on the winds. Note that no wind data was assimilated in these experiments; all impact on the GEOS wind fields comes through the effect of temperature and water vapor assimilation. It is notable that the positive impact on winds is consistent through the depth of the troposphere, and extends above and below the region in which GeoMW data are assimilated (300–850 hPa). In contrast, the impact on temperature and water vapor is largely confined to the region of assimilation, with the largest improvements in the lower troposphere. We point out that the thermodynamic state of the lower free troposphere has recently been highlighted for its importance to convection and convective organization (Schiro and Neelin 2019).

Fig. 17.
Fig. 17.

Plots of the GEOS analysis RMSE calculated over the GeoMW field of regard (see text for details) (a),(b) for zonal and meridional wind components and (c),(d) for temperature and specific humidity. The error for the control experiment is shown in red, while the error for the experiment assimilating GeoMW profiles is shown in blue.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

Impact of candidate observations in a global OSSE can be compared quantitatively versus the impact of the current observing system using forecast sensitivity observation impact (FSOI; Langland and Baker 2004; Zhu and Gelaro 2008; Gelaro and Zhu 2009) metric. Specifically, FSOI quantifies the effect of each observation type on the 24-h forecast error using a moist energy norm (Holdaway et al. 2014). We compute this norm versus the nature run, and for two different regions: for the entire global model domain, and for a region surrounding the GeoMW field of view. In the single-GeoMW configuration, this region spans 10°S–40°N latitude and 245°–295°E longitude. In the two-GeoMW configuration, this region spans 10°S–40°N latitude and 200°–310°E longitude. The observation impact of the GeoMW thermodynamic profiles is shown in Fig. 18, and it can be seen that the GeoMW observations have the second and third highest total impact regionally (with the caveat that the impact of SatWind AMVs has been known to be overestimated in the GMAO OSSE system; Privé and Errico 2019). The impact per observation is moderate, and as such the total impact reflects the fact that the GeoMW observations are always available over the region of interest (relative to the intermittent observations from polar orbiting satellites). The region of view is limited, and as such the global impact of the GeoMW observations is much smaller than the regional impact (bottom row, Fig. 18). It is, however, notable that the impact is higher than for many of the polar orbiting sounders and comparable to AIRS in per-observation-impact.

Fig. 18.
Fig. 18.

Results of the FSOI analysis of forecast impact by observation type. Negative energy units indicate positive forecast impact (error reduction). (left) Total impact and (right) impact per observation is shown. (a),(b) The impact of each observation type over the GeoMW field of regard, and (c),(d) the global impact. The separate impact of GeoMW temperature and water vapor profiles is highlighted in red.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

b. Assimilation of GeoMW thermodynamic profiles, dual spacecraft configuration

Currently, U.S. geostationary imagers/sounders are located in two positions; one covering the western portion of the United States and the other covering the eastern portion. To examine the potential impact of assimilating data from two GeoMW sounders, we conducted a second global OSSE that assimilates 6-hourly GeoMW profiles over two 5000-km diameter fields of regard; one located at 75°W longitude, and the other at 135°W longitude (Fig. 6b). As in the single spacecraft configuration, the center of the field of regard is located at north of the sub-satellite point at 15°N latitude for both satellites.

Figure 19 shows the RMSE of temperature, water vapor, and the wind components for both single and dual spacecraft configurations. As in the single-GeoMW results, RMSE is computed between the GEOS analysis and the nature run over the two GeoMW fields of view, and for the entire 2-month assimilation period. Addition of a second observing platform has a small positive impact on the meridional winds and a negligible impact on the zonal winds. The second platform does produce an additional temperature error reduction in the lower troposphere, but the impacts in the upper troposphere are mixed. Water vapor uncertainty is also reduced in the lower troposphere with mixed results aloft. We note that it is not possible to make a direct comparison between the single and dual satellite configurations, as the fields of regard are placed in different locations.

Fig. 19.
Fig. 19.

As in Fig. 16, but adding the results from the two-spacecraft configuration (2 GeoMW; green lines).

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

FSOI results for the two-spacecraft configuration are shown in Fig. 20. Comparison between Figs. 18 and 20 reveals that the addition of a second GeoMW instrument has an additional small positive impact on the regional results; the GeoMW water vapor is now the highest impact observation in the regions covered by the GeoMW observations. In the global results, the impact of GeoMW observations is ranked significantly higher; 9th and 13th for water vapor and temperature, respectively, compared with 13th and 14th in the single-satellite configuration.

Fig. 20.
Fig. 20.

As in Fig. 18, but for the 2-GeoMW configuration.

Citation: Monthly Weather Review 150, 3; 10.1175/MWR-D-21-0192.1

5. Summary and discussion

We have conducted a set of OSSEs to explore the potential impact of assimilating GeoMW data on regional and global numerical weather prediction. The regional GeoMW OSSE was conducted for a single tropical cyclone case that is analogous to Hurricane Harvey (2017). We generated a new NR with 1-km horizontal grid spacing on a 1200 km × 1200 km domain that featured a rapidly intensifying storm that ultimately reached category-5 intensity. Synthetic conventional observations and GeoMW soundings were obtained from the NR, with GeoMW soundings averaged in the horizontal according to the instrument ground spatial distance, and blurred in the vertical using instrument-configuration-specific averaging kernels obtained from the retrieval system. The well-tested WRF-based ensemble forecast and ensemble Kalman filter data assimilation system was used to produce a control forecast that assimilated only conventional data, plus three additional forecasts that also assimilated GeoMW temperature and water vapor profiles, a forecast that added estimates of TC central pressure, and one in which GeoMW AMVs were also assimilated.

We found that the control forecast was unable to produce the NR tropical cyclone structure or intensity, resulting in development of a weak category-1 storm. In contrast, assimilation of GeoMW profiles significantly and uniformly improved the TC forecast intensity, TC structure, and storm track. While there was limited impact on the environmental temperature and water vapor, winds outside of the TC were greatly improved, with errors less than half of those in the control experiment.

Global OSSEs were conducted using the GMAO OSSE system, assimilating two months of 6-hourly GeoMW temperature and water vapor profile data. Two separate experiments were performed; one that assimilated data obtained over a single fixed 5000-km diameter region centered over the Gulf of Mexico, and another that assimilated data obtained over two regions approximately consistent with the positions of GOES east and west. Impact was assessed over the field(s) of view of GeoMW, as well as globally using the forecast sensitivity observation impact metric. It was found that GeoMW thermodynamic profile assimilation improved the forecasts of temperature and water vapor, with largest improvements in the lower troposphere. Winds were also improved, with positive impacts evident through the depth of the troposphere. The FSOI metric for the region covered by GeoMW observations showed that single-GeoMW temperature and water vapor profile assimilation had the second and third largest (respectively) impact over the GeoMW field of view, and first and third largest impact in the dual-GeoMW configuration. The global impact for both single and dual GeoMW configurations was comparable to other MW and IR sounders.

There are a few caveats that should be kept in mind when interpreting the results of this study. First, effective data assimilation of a new set of observations requires an extensive period of observation tuning and calibration, during which optimal channels are selected and forward model errors are mitigated. In addition, assimilation of synthetic retrievals is not the norm in modern data assimilation; rather, most operational centers prefer to ingest radiances and use the forward operators in the data assimilation system to convert back and forth from the model state variables. The use of synthetic temperature and water vapor profiles with uncertainties that are consistent with the information provided by the observing system is likely to produce similar results to radiance assimilation, but this assertion would need to be tested by running companion profile/radiance OSSEs. Since radiance assimilation would have required extensive modification to, and tuning of, the data assimilation systems we have left such analysis for future work.

Acknowledgments.

Will McCarty and Nikki Privé at the NASA Goddard Space Flight Center Global Modeling and Assimilation Office provided assistance with configuring and running the GEOS OSSE system. A portion of this research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration.

Data availability statement.

The GEOS5 nature run data are publicly available from NASA’s Global Modeling and Assimilation Office via an online (http or OpenDAP) interface. A description of the dataset and data access instructions can be found at https://gmao.gsfc.nasa.gov/global_mesoscale/7km-G5NR/. Datasets produced in the course of this research (WRF nature run and ensemble forecasts and the GEOS5 OSSE datasets) are too large to be publicly archived. All model and OSSE configuration files are available from the lead author, and the model and experiment data have been archived on the NASA mass storage system.

REFERENCES

  • Ahern, K., and L. Cowan, 2018: Minimizing common errors when projecting geospatial data onto a vortex-centered space. Geophys. Res. Lett., 45, 12 03212 039, https://doi.org/10.1029/2018GL079953.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Barker, D. M., W. Huang, Y.-R. Guo, A. J. Bourgeois, and Q. N. Xiao, 2004: A three-dimensional variational data assimilation system for MM5: Implementation and initial results. Mon. Wea. Rev., 132, 897914, https://doi.org/10.1175/1520-0493(2004)132<0897:ATVDAS>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Errico, R. M., and Coauthors, 2017: Description of the GMAO OSSE for Weather Analysis Software Package: Version 3. Tech. Rep. NASA/TM-2017-104606, Technical Report Series on Global Modeling and Data Assimilation, Vol. 48, NASA, 156 pp.

    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., F. Weng, N. C. Grody, and L. Zhao, 2000: Precipitation characteristics over land from the NOAA‐15 AMSU sensor. Geophys. Res. Lett., 27, 26692672, https://doi.org/10.1029/2000GL011665.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gelaro, R., and Y. Zhu, 2009: Examination of observation impacts derived from observing system experiments (OSEs) and adjoint models. Tellus, 61A, 179193, https://doi.org/10.1111/j.1600-0870.2008.00388.x.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gelaro, R., and Coauthors, 2015: Evaluation of the 7-km GEOS-5 nature run. Tech. Rep. NASA/TM-2014-104606, Technical Report Series on Global Modeling and Data Assimilation, Vol. 36, NASA, 305 pp.

    • Search Google Scholar
    • Export Citation
  • Hoffman, R. N., and R. Atlas, 2016: Future observing system simulation experiments. Bull. Amer. Meteor. Soc., 97, 16011616, https://doi.org/10.1175/BAMS-D-15-00200.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Holdaway, D., R. Errico, R. Gelaro, and J. G. Kim, 2014: Inclusion of linearized moist physics in NASA’s Goddard Earth observing system data assimilation tools. Mon. Wea. Rev., 142, 414433, https://doi.org/10.1175/MWR-D-13-00193.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kleist, D. T., D. F. Parrish, J. C. Derber, R. Treadon, W.-S. Wu, and S. Lord, 2009: Introduction of the GSI into the NCEP Global Data Assimilation System. Wea. Forecasting, 24, 16911705, https://doi.org/10.1175/2009WAF2222201.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Lambrigtsen, B., H. Van Dang, F. J. Turk, S. M. Hristova-Veleva, H. Su, and Y. Wen, 2018: All-weather tropospheric 3-D wind from microwave sounders. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., 11, 19491956, https://doi.org/10.1109/JSTARS.2018.2814540.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Langland, R. H., and N. L. Baker, 2004: Estimation of observation impact using the NRL atmospheric variational data assimilation adjoint system. Tellus, 56A, 189201, https://doi.org/10.3402/tellusa.v56i3.14413.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Laviola, S., and V. Levizzani, 2011: The 183-WSL fast rain rate retrieval algorithm: Part I: Retrieval design. Atmos. Res., 99, 443461, https://doi.org/10.1016/j.atmosres.2010.11.013.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McCarty, W., D. Carvalho, I. Moradi, and N. C. Privé, 2021: Observing system simulation experiments investigating atmospheric motion vectors and radiances from a constellation of 4–5-μm infrared sounders. J. Atmos. Oceanic Technol., 38, 331347, https://doi.org/10.1175/JTECH-D-20-0109.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McNoldy, B., B. Annane, J. Delgado, L. Bucci, R. Atlas, and S. J. Majumdar, 2017: Impact of assimilating CYGNSS data on tropical cyclone analyses and forecasts in a regional OSSE framework. Mar. Technol. Soc. J., 51, 715, https://doi.org/10.4031/MTSJ.51.1.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Minamide, M., and F. Zhang, 2017: Adaptive observation error inflation for assimilating all-sky satellite radiance. Mon. Wea. Rev., 145, 10631081, https://doi.org/10.1175/MWR-D-16-0257.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Minamide, M., and F. Zhang, 2018: Assimilation of all-sky infrared radiances from Himawari- 8 and impacts of moisture and hydrometer initialization on convection-permitting tropical cyclone prediction. Mon. Wea. Rev., 146, 32413258, https://doi.org/10.1175/MWR-D-17-0367.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Minamide, M., and F. Zhang, 2019: An adaptive background error inflation method for assimilating all‐sky radiances. Quart. J. Roy. Meteor. Soc., 145, 805823, https://doi.org/10.1002/qj.3466.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Minamide, M., F. Zhang, and E. E. Clothiaux, 2020: Nonlinear forecast error growth of rapidly intensifying Hurricane Harvey (2017) examined through convection-permitting ensemble assimilation of GOES-16 all-sky radiances. J. Atmos. Sci., 77, 42774296, https://doi.org/10.1175/JAS-D-19-0279.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Molod, A., L. Takacs, M. Suarez, and J. Bacmeister, 2015: Development of the GEOS-5 atmospheric general circulation model: Evolution from MERRA to MERRA2. Geosci. Model Dev., 8, 13391356, https://doi.org/10.5194/gmd-8-1339-2015.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Posselt, D. J., and Coauthors, 2019: Quantitative assessment of state-dependent atmospheric motion vector uncertainties. J. Appl. Meteor. Climatol., 58, 24792495, https://doi.org/10.1175/JAMC-D-19-0166.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Privé, N. C., and R. M. Errico, 2019: Uncertainty of observation impact estimation in an adjoint model investigated with an observing system simulation experiment. Mon. Wea. Rev., 147, 31913204, https://doi.org/10.1175/MWR-D-19-0097.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Privé, N. C., R. M. Errico, and W. McCarty, 2021: The importance of simulated errors in observing system simulation experiments. Tellus, 73, 117, https://doi.org/10.1080/16000870.2021.1886795.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Reale, O., J. Terry, M. Masutani, E. Andersson, L. P. Riishojgaard, and J. C. Jusem, 2007: Preliminary evaluation of the European Centre for Medium-Range Weather Forecasts’ (ECMWF) nature run over the tropical Atlantic and African monsoon region. Geophys. Res. Lett., 34, L22810, https://doi.org/10.1029/2007GL031640.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rienecker, M. M., and Coauthors, 2008: The GEOS-5 Data Assimilation System—Documentation of versions 5.0.1, 5.1.0, and 5.2.0. Tech. Rep. Series on Global Modeling and Data Assimilation, Vol. 27, 109 pp.

    • Search Google Scholar
    • Export Citation
  • Rosenkranz, P. W., 2001: Retrieval of temperature and moisture profiles from AMSU-A and AMSU-B measurements. IEEE Trans. Geosci. Remote Sens., 39, 24292435, https://doi.org/10.1109/36.964979.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schiro, K. A., and J. D. Neelin, 2019: Deep convective organization, moisture vertical structure, and convective transition using deep-inflow mixing. J. Atmos. Sci., 76, 965987, https://doi.org/10.1175/JAS-D-18-0122.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tanelli, S., and Coauthors, 2012: Integrated instrument simulator suites for Earth science. Proc. SPIE, 8529, 85290D, https://doi.org/10.1117/12.977577.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zeng, X., and Coauthors, 2020: Use of observing system simulation experiments in the United States. Bull. Amer. Meteor. Soc., 101, E1427E1438, https://doi.org/10.1175/BAMS-D-19-0155.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, F., Y. Weng, J. Sippel, Z. Meng, and C. H. Bishop, 2009: Cloud-resolving hurricane initialization and prediction through assimilation of Doppler radar observations with an ensemble Kalman filter. Mon. Wea. Rev., 137, 21052125, https://doi.org/10.1175/2009MWR2645.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, F., M. Minamide, and E. E. Clothiaux, 2016: Potential impacts of assimilating all-sky satellite radiances from GOES-R on convection-permitting analysis and prediction of tropical cyclones. Geophys. Res. Lett., 43, 29542963, https://doi.org/10.1002/2016GL068468.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, F., M. Minamide, X. Chen, R. G. Nystrom, S.-J. Lin, and L. M. Harris, 2019: Improving Harvey forecasts with next-generation weather satellites: Advanced hurricane analysis and prediction with assimilation of GOES-R all-sky radiances. Bull. Amer. Meteor. Soc., 100, 12171222, https://doi.org/10.1175/BAMS-D-18-0149.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhu, Y., and R. Gelaro, 2008: Observation sensitivity calculations using the adjoint of the Gridpoint Statistical Interpolation (GSI) analysis system. Mon. Wea. Rev., 136, 335351, https://doi.org/10.1175/MWR3525.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
Save
  • Ahern, K., and L. Cowan, 2018: Minimizing common errors when projecting geospatial data onto a vortex-centered space. Geophys. Res. Lett., 45, 12 03212 039, https://doi.org/10.1029/2018GL079953.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Barker, D. M., W. Huang, Y.-R. Guo, A. J. Bourgeois, and Q. N. Xiao, 2004: A three-dimensional variational data assimilation system for MM5: Implementation and initial results. Mon. Wea. Rev., 132, 897914, https://doi.org/10.1175/1520-0493(2004)132<0897:ATVDAS>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Errico, R. M., and Coauthors, 2017: Description of the GMAO OSSE for Weather Analysis Software Package: Version 3. Tech. Rep. NASA/TM-2017-104606, Technical Report Series on Global Modeling and Data Assimilation, Vol. 48, NASA, 156 pp.

    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., F. Weng, N. C. Grody, and L. Zhao, 2000: Precipitation characteristics over land from the NOAA‐15 AMSU sensor. Geophys. Res. Lett., 27, 26692672, https://doi.org/10.1029/2000GL011665.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gelaro, R., and Y. Zhu, 2009: Examination of observation impacts derived from observing system experiments (OSEs) and adjoint models. Tellus, 61A, 179193, https://doi.org/10.1111/j.1600-0870.2008.00388.x.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gelaro, R., and Coauthors, 2015: Evaluation of the 7-km GEOS-5 nature run. Tech. Rep. NASA/TM-2014-104606, Technical Report Series on Global Modeling and Data Assimilation, Vol. 36, NASA, 305 pp.

    • Search Google Scholar
    • Export Citation
  • Hoffman, R. N., and R. Atlas, 2016: Future observing system simulation experiments. Bull. Amer. Meteor. Soc., 97, 16011616, https://doi.org/10.1175/BAMS-D-15-00200.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Holdaway, D., R. Errico, R. Gelaro, and J. G. Kim, 2014: Inclusion of linearized moist physics in NASA’s Goddard Earth observing system data assimilation tools. Mon. Wea. Rev., 142, 414433, https://doi.org/10.1175/MWR-D-13-00193.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kleist, D. T., D. F. Parrish, J. C. Derber, R. Treadon, W.-S. Wu, and S. Lord, 2009: Introduction of the GSI into the NCEP Global Data Assimilation System. Wea. Forecasting, 24, 16911705, https://doi.org/10.1175/2009WAF2222201.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Lambrigtsen, B., H. Van Dang, F. J. Turk, S. M. Hristova-Veleva, H. Su, and Y. Wen, 2018: All-weather tropospheric 3-D wind from microwave sounders. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., 11, 19491956, https://doi.org/10.1109/JSTARS.2018.2814540.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Langland, R. H., and N. L. Baker, 2004: Estimation of observation impact using the NRL atmospheric variational data assimilation adjoint system. Tellus, 56A, 189201, https://doi.org/10.3402/tellusa.v56i3.14413.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Laviola, S., and V. Levizzani, 2011: The 183-WSL fast rain rate retrieval algorithm: Part I: Retrieval design. Atmos. Res., 99, 443461, https://doi.org/10.1016/j.atmosres.2010.11.013.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McCarty, W., D. Carvalho, I. Moradi, and N. C. Privé, 2021: Observing system simulation experiments investigating atmospheric motion vectors and radiances from a constellation of 4–5-μm infrared sounders. J. Atmos. Oceanic Technol., 38, 331347, https://doi.org/10.1175/JTECH-D-20-0109.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McNoldy, B., B. Annane, J. Delgado, L. Bucci, R. Atlas, and S. J. Majumdar, 2017: Impact of assimilating CYGNSS data on tropical cyclone analyses and forecasts in a regional OSSE framework. Mar. Technol. Soc. J., 51, 715, https://doi.org/10.4031/MTSJ.51.1.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Minamide, M., and F. Zhang, 2017: Adaptive observation error inflation for assimilating all-sky satellite radiance. Mon. Wea. Rev., 145, 10631081, https://doi.org/10.1175/MWR-D-16-0257.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Minamide, M., and F. Zhang, 2018: Assimilation of all-sky infrared radiances from Himawari- 8 and impacts of moisture and hydrometer initialization on convection-permitting tropical cyclone prediction. Mon. Wea. Rev., 146, 32413258, https://doi.org/10.1175/MWR-D-17-0367.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Minamide, M., and F. Zhang, 2019: An adaptive background error inflation method for assimilating all‐sky radiances. Quart. J. Roy. Meteor. Soc., 145, 805823, https://doi.org/10.1002/qj.3466.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Minamide, M., F. Zhang, and E. E. Clothiaux, 2020: Nonlinear forecast error growth of rapidly intensifying Hurricane Harvey (2017) examined through convection-permitting ensemble assimilation of GOES-16 all-sky radiances. J. Atmos. Sci., 77, 42774296, https://doi.org/10.1175/JAS-D-19-0279.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Molod, A., L. Takacs, M. Suarez, and J. Bacmeister, 2015: Development of the GEOS-5 atmospheric general circulation model: Evolution from MERRA to MERRA2. Geosci. Model Dev., 8, 13391356, https://doi.org/10.5194/gmd-8-1339-2015.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Posselt, D. J., and Coauthors, 2019: Quantitative assessment of state-dependent atmospheric motion vector uncertainties. J. Appl. Meteor. Climatol., 58, 24792495, https://doi.org/10.1175/JAMC-D-19-0166.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Privé, N. C., and R. M. Errico, 2019: Uncertainty of observation impact estimation in an adjoint model investigated with an observing system simulation experiment. Mon. Wea. Rev., 147, 31913204, https://doi.org/10.1175/MWR-D-19-0097.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Privé, N. C., R. M. Errico, and W. McCarty, 2021: The importance of simulated errors in observing system simulation experiments. Tellus, 73, 117, https://doi.org/10.1080/16000870.2021.1886795.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Reale, O., J. Terry, M. Masutani, E. Andersson, L. P. Riishojgaard, and J. C. Jusem, 2007: Preliminary evaluation of the European Centre for Medium-Range Weather Forecasts’ (ECMWF) nature run over the tropical Atlantic and African monsoon region. Geophys. Res. Lett., 34, L22810, https://doi.org/10.1029/2007GL031640.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rienecker, M. M., and Coauthors, 2008: The GEOS-5 Data Assimilation System—Documentation of versions 5.0.1, 5.1.0, and 5.2.0. Tech. Rep. Series on Global Modeling and Data Assimilation, Vol. 27, 109 pp.

    • Search Google Scholar
    • Export Citation
  • Rosenkranz, P. W., 2001: Retrieval of temperature and moisture profiles from AMSU-A and AMSU-B measurements. IEEE Trans. Geosci. Remote Sens., 39, 24292435, https://doi.org/10.1109/36.964979.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schiro, K. A., and J. D. Neelin, 2019: Deep convective organization, moisture vertical structure, and convective transition using deep-inflow mixing. J. Atmos. Sci., 76, 965987, https://doi.org/10.1175/JAS-D-18-0122.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tanelli, S., and Coauthors, 2012: Integrated instrument simulator suites for Earth science. Proc. SPIE, 8529, 85290D, https://doi.org/10.1117/12.977577.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zeng, X., and Coauthors, 2020: Use of observing system simulation experiments in the United States. Bull. Amer. Meteor. Soc., 101, E1427E1438, https://doi.org/10.1175/BAMS-D-19-0155.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, F., Y. Weng, J. Sippel, Z. Meng, and C. H. Bishop, 2009: Cloud-resolving hurricane initialization and prediction through assimilation of Doppler radar observations with an ensemble Kalman filter. Mon. Wea. Rev., 137, 21052125, https://doi.org/10.1175/2009MWR2645.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, F., M. Minamide, and E. E. Clothiaux, 2016: Potential impacts of assimilating all-sky satellite radiances from GOES-R on convection-permitting analysis and prediction of tropical cyclones. Geophys. Res. Lett., 43, 29542963, https://doi.org/10.1002/2016GL068468.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, F., M. Minamide, X. Chen, R. G. Nystrom, S.-J. Lin, and L. M. Harris, 2019: Improving Harvey forecasts with next-generation weather satellites: Advanced hurricane analysis and prediction with assimilation of GOES-R all-sky radiances. Bull. Amer. Meteor. Soc., 100, 12171222, https://doi.org/10.1175/BAMS-D-18-0149.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhu, Y., and R. Gelaro, 2008: Observation sensitivity calculations using the adjoint of the Gridpoint Statistical Interpolation (GSI) analysis system. Mon. Wea. Rev., 136, 335351, https://doi.org/10.1175/MWR3525.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Depiction of the WRF domains used for the (a) nature run and (b) forecast system. Note that the outermost domain (d01) remains fixed in the same position while the inner domains [d02, d03, and (in the nature run) d04] follow the position of the tropical cyclone. As such, the position of the inner domains is valid at the time shown (1200 UTC 23 Aug 2017), but will have moved to a different position later in the simulations. Color shading depicts the topography, and any differences are due to the differences in the ECMWF vs GFS boundary conditions.

  • Fig. 2.

    (a) Minimum sea level pressure and (b) maximum 10-m wind speed from a 60-member ensemble forecast of Hurricane Harvey (2017) initialized from a WRF-EnKF analysis. The control forecast initialized from the ensemble mean is shown in the black dashed line, while the solid black line with cross marks represents the National Hurricane Center best track analysis. The ensemble member whose initial conditions are used to initialize the nature run is indicated in gray, while the NR intensity is overlaid as a thick black line.

  • Fig. 3.

    Output from the NR, valid at 1200 UTC 25 Aug 2017. (left) Wind speed (color contours) and vectors (arrows) on the 750-hPa pressure level. (right) Relative humidity (percent; color contours) on the 750-hPa pressure level.

  • Fig. 4.

    Plot of simulated outgoing longwave radiation (W m−2; gray shading) at 0600 UTC 24 Aug 2017 on the innermost domain of the nature run (d04). Overlaid are contours encircling regions with OLR < 220 W m−2 (red), and precipitation rate > 1 mm h−1 (blue) and >10 mm h−1 (orange).

  • Fig. 5.

    Observation minus background (OB) statistics for wind vector components, temperature, and water vapor from assimilation of real radiosonde data for all named storms in the 2017 hurricane season (box-and-whisker plots) and from the conventional assimilation experiment in our study (green circles).

  • Fig. 6.

    Plots showing the viewing regions (red) for the (a) single GeoMW configuration and (b) dual GeoMW configuration. Overlaid in red dots are the locations of the GeoMW observations assimilated at 0000 UTC 1 Jul 2006.

  • Fig. 7.

    Time series of (a) minimum sea level pressure; and (b) maximum 10-m wind speed for the hurricane NR (black), free-running forecast (noDA; blue colors), and forecast that assimilates conventional observations (warm colors). Different colors correspond to different initialization times, every 6 h starting at 1200 UTC 23 Aug 2017 and ending at 0600 UTC 24 Aug 2017.

  • Fig. 8.

    Comparisons of forecast intensity in terms of (a),(c),(e) minimum sea level pressure and (b),(d),(f) maximum 10-m wind speed for the nature run (black), assimilation of conventional observations (warm colors) and assimilation of three different GeoMW configurations (cool colors). (top) Clear-sky profiles only, (middle) all-sky profiles, with precipitation rates < 1 mm h−1, and (bottom) all-sky profiles, with precipitation rates up to 10 mm h−1 and precipitation-dependent error.

  • Fig. 9.

    Colors as in Fig. 4. As in the middle panel in Fig. 4, all experiments shown assimilate temperature and water vapor profiles with clear-sky errors in all-sky conditions with precipitation rates up to 1 mm h−1. However, in this case, we make the following modifications. (a),(b) Add assimilation of atmospheric motion vectors; (c),(d) add assimilation of minimum sea level pressure estimates; and (e),(f) assimilate temperature and water vapor profiles every 15 min.

  • Fig. 10.

    Comparisons of forecast track error for the control run (assimilating conventional observations; warm colors) and assimilation of various GeoMW configurations (cool colors). (a) Assimilation of clear-sky-only T and RH profiles. (b) Assimilation of all-sky T and RH profiles in regions with precipitation < 1 mm h−1. (c) Assimilation of all-sky T and RH profiles in regions with precipitation < 10 mm h−1 and with precipitation-dependent error. (d) As in (b), but with the addition of water vapor AMVs. (e) As in (b), but with the addition of GeoMW-based minimum SLP estimates.

  • Fig. 11.

    Color contours of (a) warm core temperature perturbation and (b) warm core temperature deviation from the nature run for the 48-h forecast initialized at 0000 UTC 24 Aug 2017 and valid at 0000 UTC 26 Aug 2017. Experiment labels can be found atop each of the subplots.

  • Fig. 12.

    Color contours of (a) azimuthal wind, (b) azimuthal wind deviation from the nature run, (c) radial wind, and (d) radial wind deviation from the nature run. As in Fig. 10, all fields are from the 48-h forecast initialized at 0000 UTC 24 Aug 2017 and valid at 0000 UTC 26 Aug 2017. Experiment labels can be found atop each of the subplots.

  • Fig. 13.

    Time series of (inner; d03) domain averaged root mean squared error (RMSE) between the all-sky GeoMW assimilation experiment vs domain 03 of the NR, reduced to the size and position of domain 03 of the forecast. Temperature, specific humidity, and zonal (u direction), and meridional (υ direction) winds are shown from top to bottom. In each panel, red represents the RMSE for the conventional forecast while blue represents the RMSE for the GeoMW assimilation experiment(s).

  • Fig. 14.

    Vertical profiles of (inner; d03) domain and time (1200 UTC 24 Aug–0000 UTC 27 Aug) averaged root mean squared error (RMSE) between each GeoMW assimilation experiment vs domain 03 of the NR, reduced to the size and position of domain 03 of the forecast. Shown are the RMSE profiles for temperature, specific humidity, and zonal (u direction), and meridional (υ direction) winds. As in Fig. 8, in each panel, red represents the RMSE for the conventional forecast while blue represents the RMSE for the GeoMW assimilation experiment(s). Percent deviations between the GeoMW assimilation experiment and the control are represented as numbers in the top-right corner of each plot, with an asterisk (*) indicating the differences are statistically significant.

  • Fig. 15.

    Histograms of precipitation rate from the nature run (black), control forecast (red), and GeoMW assimilation experiment (blue) at four different forecast times, each separated by 12 h (times are listed on the right side of each panel). The results are derived from the forecast initialized at 0000 UTC 24 Aug 2017.

  • Fig. 16.

    Time series plot of the 99th percentile of the near-surface wind speeds in the nature run (black), no-DA experiment (blue), control experiment (green), GeoMW with (yellow), and without (red) minSLP assimilation, and GeoMW 15-min assimilation frequency (purple).

  • Fig. 17.

    Plots of the GEOS analysis RMSE calculated over the GeoMW field of regard (see text for details) (a),(b) for zonal and meridional wind components and (c),(d) for temperature and specific humidity. The error for the control experiment is shown in red, while the error for the experiment assimilating GeoMW profiles is shown in blue.

  • Fig. 18.

    Results of the FSOI analysis of forecast impact by observation type. Negative energy units indicate positive forecast impact (error reduction). (left) Total impact and (right) impact per observation is shown. (a),(b) The impact of each observation type over the GeoMW field of regard, and (c),(d) the global impact. The separate impact of GeoMW temperature and water vapor profiles is highlighted in red.

  • Fig. 19.

    As in Fig. 16, but adding the results from the two-spacecraft configuration (2 GeoMW; green lines).

  • Fig. 20.

    As in Fig. 18, but for the 2-GeoMW configuration.

All Time Past Year Past 30 Days
Abstract Views 531 0 0
Full Text Views 344 208 16
PDF Downloads 342 189 12