Dependence of positive refractivity bias of GPS RO cloudy profiles on cloud fraction along GPS RO limb tracks
- First Online:
We collected global COSMIC RO data from 2009 and 2010 using measurements from the NOAA-18 Advanced Microwave Sounding Unit-A (AMSU-A) along GPS RO limb tracks and also using data from the CloudSat Cloud Profiling Radar (CPR). The collocated AMSU-A liquid water path (LWP) retrieval over the ocean is then used to quantify the dependence of fractional refractivity bias (N-bias) of GPS RO profiles in cloudy conditions on AMSU-A data points with nonzero LWP along GPS RO limb tracks. The collocated CPR cloud type dataset is used for selecting GPS RO profiles in cloudy conditions and for grouping GPS RO cloudy profiles into seven different cloud types. It is shown that the positive fractional N-bias varies with cloud fraction along the COSMIC GPS RO limb tracks. It reaches a value between 1 and 2 % when cloud fraction is as high as 90–100 % for altocumulus, altostratus, cirrus, cumulus, and deep convection clouds. For nimbostratus and stratocumulus clouds, large biases are found at any value of cloud fraction. The positive fractional N-bias can be more than 2 % even if cloud fraction is less than about 50 % for nimbostratus and stratus clouds.
KeywordsGPS RO Cloud type Data assimilation
Global measurements of the vertical profiles of the atmospheric refractivity made by the GPS radio occultation (RO) technique are derived from the measurements of carrier phase delay of two known L-band frequencies, L1 at 1.57542 GHz and L2 at 1.22760 GHz, transmitted from the GPS satellites. The phase delay measurements are first converted to the bending angles of radio wave trajectories and then to the vertical profiles of the atmospheric refractivity (Zou et al. 1999). GPS RO measurements are applicable to both weather and climate studies.
COSMIC GPS RO data during 2009 and 2010 are first collocated with CloudSat CPR to identify cloudy ROs as well as cloud types and then collocated with global cloud liquid water path (LWP) data derived from two window channels of the Advanced Microwave Sounding Unit-A (AMSU-A) onboard the polar-orbiting satellite NOAA-18. The collocated LWP data along the 600-km GPS limb track of and centered at the average tangent point of each GPS RO profile are used to provide an approximation of cloud fraction information. The dependence of the positive N-bias of GPS RO cloudy profiles on cloud fraction along GPS RO limb tracks is quantified for seven different cloud types based on a collocated CPR cloud type dataset. Since the AMSU-A LWP retrievals are available over ocean only, the present study is restricted to ocean clouds only.
The following four data types are employed in this study: (i) COSMIC GPS RO refractivity, (ii) CloudSat CPR cloud type, (iii) NOAA-18 AMSU-A LWP, and (iv) ECMWF temperature and humidity. COSMIC GPS RO data have been provided to the international community since April 2006 (Anthes et al. 2008). We use refractivity data of 2009 and 2010 that are data made available by CDAAC (UCAR COSMIC Data Analysis and Archival Center) (Kuo et al. 2004). The daily RO data count is about 2000 during this period. The GPS RO refractivity profile data are provided as a function of altitude at a vertical resolution of approximately 200 m. The horizontal resolution of GPS RO data is approximately 1.5 km in the cross-track direction and 300 km in the along-track direction (Kursinski et al. 1996).
The cloudy and clear-sky COSMIC GPS ROs were selected from those COSMIC GPS ROs that were collocated with CloudSat data. CloudSat data have been available since June 2, 2006. The CloudSat satellite orbits the earth approximately every 1.5 h. A 94-GHz and nadir-pointing CPR is the primary observing instrument onboard. It measures the returned backscattered power from clouds as a function of the distance from the CPR. The along-track temporal sample interval equals 0.16 s, the along-track spatial resolution is about 1.1 km, and the effective field of view (FOV) is approximately 1.4 km × 1.7 km (Tanelli et al. 2008). CloudSat provides a total of 30,000 vertical profiles of radar reflectivity, liquid water content (LWC), ice water content (IWC), a maximum of five cloud layers, cloud type, as well as the altitudes of cloud tops and cloud bases along each of its complete orbits (Stephens et al. 2002).
Further selection is made for NOAA-18 AMSU-A-derived LWP data that are collocated with the limb tracks of all the COSMIC and CloudSat-collocated GPS ROs. Microwave measurements at the two lowest-frequency window channels of AMSU-A are functions of cloud LWP and water vapor path. The LWP and water vapor path can be retrieved using an emission-based radiative transfer model (Greenwald et al. 1993; Weng and Grody 1994; Grody et al. 2001). The cloud LWP used in this study were obtained from measurements at AMSU-A channels 1 and 2 over the ocean using the retrieval algorithm developed by Weng et al. (2003).
The ECMWF analysis of temperature, water vapor, and pressure at model grid points was interpolated to CloudSat data resolution and was included in the CloudSat auxiliary data products. These ECMWF profiles were generated from an ECMWF analysis. It is worth mentioning that COSMIC data above 4 km are included in ECMWF analyses (Healy and Thépaut 2006).
Collocation among three types of data
The COSMIC GPS ROs were selected from the subset that was collocated with CloudSat data. The temporal and spatial collocation criteria between GPS RO and CloudSat cloud are <3 h and 50 km. The cloud type of each collocated RO profile is determined by the CloudSat profile that is closest to the mean position of the RO profile. There are seven different cloud types: deep convection (Dc), cumulus (Cu), cirrus (Ci), altocumulus (Ac), stratocumulus (Sc), altostratus (As), and nimbostratus (Ns). The cloud type information used in this study is obtained from the 2B-CLDCLASS datasets. The total number of collocated cloudy GPS RO profiles is 6593.
Once COSMIC GPS ROs collocated with CloudSat CPR data have been selected, further collocation is made between NOAA-18 AMSU-A derived LWP data and the GPS tangent line of ±300 km length for all the COSMIC and CloudSat-collocated GPS RO profiles. The temporal and spatial collocation criteria are 3 h and 30 km, respectively, from the 300-km-long tangent line that is tangent to the GPS RO limb track at the mean GPS RO tangent point. It is noted that the global cloud LWP data derived from AMSU-A represent the total cloud liquid water amount in a vertical column over a nearly circular area of about 50–100 km in diameter, but the height of the existing clouds is not known from the satellite imagery. On the other hand, the RO signal is influenced by the medium existing along the ray-tracing path, i.e., line-of-sight medium. Therefore, the collocation conditions between RO data and LWP data are still a crude way for assessing cloud impacts on GPS RO biases.
Figure 4 (bottom) provides the data counts of COSMIC GPS RO profiles collocated with different numbers of LWP data along each of the GPS RO tangent lines in cloudy and clear-sky conditions. For most of GPS RO tangent lines, there are about 6–8 collocated LWP data points. This is expected given the above-mentioned AMSU-A data resolution and the presence of AMSU-A swath gaps in the tropics.
Dependence of positive N-bias on cloud fraction along GPS ray path
Lin et al. (2010) compared large-scale analysis biases in clear-sky and cloudy conditions and found that the COSMIC GPS RO refractivity observations are systematically greater than the refractivity calculated from ECMWF analyses, which is to be referred as a positive N-bias in clouds. By using LWC observations made available by CloudSat CPR, Yang and Zou (2012) quantified contributions to atmospheric refractivity from LWC to show that GPS signals could be modulated by cloud and precipitation to a level exceeding the GPS RO observation error. Evaluating the GPS N-biases by using a collocated single CloudSat profile to represent a cloud environment along a GPS RO limb track has limitation. The global LWP data from AMSU-A onboard the polar-orbiting satellite NOAA-18, which has a much wider swath than that of CPR onboard CloudSat, provide an approximation of cloud fraction information so that the dependence of positive N-bias of GPS RO cloudy profiles on cloud fraction along GPS RO limb tracks can be assessed. The evaluation of the dependence of positive N-bias of GPS RO cloudy profiles on cloud fraction along GPS RO limb tracks will be conducted by grouping all the collocated cloudy ROs into seven different cloud types that are determined at the average tangent point of each RO. Although a single CloudSat profile is likely not representative of cloud type over the entire GPS RO ray path that may contain segments of other clouds and clear-sky, the effects of the same type of clouds for collocated cloudy ROs are likely to be larger than other cloud types. Also, most contributions to GPS RO come from the atmosphere near the tangent point. Results are provided below.
The accuracy of the GPS RO refractivity profiles in the lowermost troposphere is about 1 %, which comes mainly from the Abel inversion which derives GPS refractivity from the excess phase delay of the GPS radio signals (Kursinski et al. 1995; Kursinski and Hajj 2001), The horizontal variations of atmospheric refractivity have an effect on GPS RO signals, but cannot be represented in the standard Abel transform (Healy 2001). The 1–2 % positive N-biases are similar or slightly greater than the accuracy of refractivity estimated in Fig. 6.
The standard deviations of the fractional N differences for seven single cloud types aligned at the cloud base as functions of cloud fraction along the ray path are also provided in Fig. 6. It is pointed out that the number of GPS ROs in cumulus and deep convection clouds is an order of magnitude smaller than those in other cloud types. A large variability of the fractional N differences is found near the cloud top for altocumulus, cirrus, cumulus, deep convection, and stratocumulus clouds. The standard deviations of the fractional N differences are small and do not vary much with respect to cloud fraction and altitude for both altostratus and nimbostratus clouds.
Impacts of cloud absorption on fractional N-bias
Although the impacts of liquid and ice clouds on the GPS refractivity are more than two orders of magnitude smaller than those of dry and water vapor parts, the positive N-bias is robustly known. Given the small percentage of cloud impacts on GPS RO mean fractional differences, it becomes even more important to accurately simulate cloud liquid and ice water mixing ratios for applying GPS RO data for NWP and climate studies.
Summary and conclusions
The positive N-biases in clouds are estimated so that they may be removed from GPS RO data assimilation in cloudy conditions. A two-year collocation between GPS ROs, CloudSat, and microwave radiance data from AMSU-A instruments onboard the polar-orbiting meteorological satellite NOAA-18 allowed the impacts of cloud fraction along GPS RO ray paths on N-biases be estimated for the first time.
Taking advantages of COSMIC with six LEO satellites to provide GPS ROs, CloudSat with a 94-GHz, nadir-pointing CPR to provide different cloud types, and LWC and IWC within clouds, and NOAA-18 with a cross-track AMSU-A to provide global LWP over global oceans, two years of collocated cloudy RO profiles were selected to estimate the mean differences between COSMIC GPS RO refractivity retrievals and ECMWF analysis within collocated clouds. Specifically, using CloudSat CPR-measured information on cloud base, cloud top and cloud type and using NOAA-18 LWP-provided cloud fraction information along GPS RO ray paths, a consistent positive N-bias could be found within all types of clouds (e.g., deep convection, cumulus, cirrus, altocumulus, stratocumulus, altostratus, nimbostratus) at any cloud fraction. Most cloudy GPS ROs have a high cloud fraction along their ray paths. The mean fractional differences of refractivity between COSMIC GPS ROs and ECMWF analyses vary from 0.2 to 2 % depending on cloud fractions and cloud types. The largest positive N-bias is found within nimbostratus clouds for all cloud fractions (0–100 %) between 2 and 4 km above cloud bases (or 2.5–4.5 km above the ocean surface).
Future work includes extending the present study to a longer data period from 2011 to 2015 and to also include collocations with AMSU-A data from other polar-orbiting satellites including NOAA-19, MetOp-A/B, and Advanced Technology Microwave Sounder (ATMS) onboard Suomi NPP.
This work is jointly supported by National Natural Science Foundation of China (Project No. 41375013), Chinese Ministry of Science and Technology (Project No. 2015CB452805), and China Special fund for Meteorological Research in the Public Interest (Project No. GYHY201406008). The data used for this research can be obtained by emailing to email@example.com.
- Tarantola A (1987) Inverse problem theory: methods for data fitting and model parameter estimation. Elsevier Sci, New YorkGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.