This study extends the relative calibration adjustment technique for calibration of weather radars to higher-frequency radars as well as range–height indicator (RHI) scans. The calibration of weather radars represents one of the most dominant sources of error for their use in a variety of fields including quantitative precipitation estimation and model comparisons. While most weather radars are routinely calibrated, the frequency of calibration is often less than required, resulting in miscalibrated time periods. While full absolute calibration techniques often require the radar to be taken offline for a period of time, there have been online calibration techniques discussed in the literature. The relative calibration adjustment (RCA) technique uses the statistics of the ground clutter surrounding the radar as a monitoring source for the stability of calibration but has only been demonstrated to work at S- and C-band for plan-position indicator (PPI) scans at a constant elevation.
In this work the RCA technique is modified to work with higher-frequency radars, including Ka-band cloud radars. At higher frequencies the properties of clutter can be much more variable. This work introduces an extended clutter selection procedure that incorporates the temporal stability of clutter and helps to improve the operational stability of RCA for relatively higher-frequency radars. The technique is also extended to utilize range–height scans from radars where the elevation is varied rather than the azimuth. These types of scans are often utilized with research radars to examine the vertical structure of clouds.
The newly extended technique (eRCA) is applied to four Department of Energy Atmospheric Radiation Measurement (DOE ARM) weather radars ranging in frequency from C- to Ka-band. Cross comparisons of three co-located radars with frequencies C, X, and Ka at the ARM Cloud, Aerosol, and Complex Terrain Interactions (CACTI) site show that the technique can determine changes in calibration. Using an X-band radar at the ARM Eastern North Atlantic (ENA) site, we show how the technique can be modified to be more resilient to clutter fields that show increased variability, in this case due to sea clutter. The results show that this technique is promising for a posteriori data calibration and monitoring.
The Department of Energy Atmospheric Radiation Measurement (DOE ARM) program deploys weather radars around the world to observe a variety of weather regimes. These radars operate in remote regions with limited on-site monitoring. Without the ability to conduct routine calibrations, weather radars can experience a degradation in calibration accuracy over time. To mitigate this, a robust technique to monitor and correct calibration drifts is needed as well as a tool to monitor for possible periods of reduced performance. Additionally, for historical campaign data where retroactive calibrations cannot be performed a posteriori, there is a need to flag periods of miscalibration. Previous and existing calibration techniques rely on the detection or measurement of a known target (i.e., metallic spherical targets;
The RCA technique utilizes changes in the probability distribution of ground clutter reflectivity at the lowest elevation to identify changes in the radar system. This technique allows for tracking of radar calibration without having to take a radar offline, which is important for research radars operating during field campaigns when observation periods are limited and with historical datasets for which the capability to retroactively calibrate is limited. While the reflectivity of individual clutter targets can change significantly depending on environmental conditions (rain-free, light or heavy precipitation, anomalous propagation), the 95th percentile of the probability distribution of ground clutter reflectivity remains stable in the absence of engineering changes, system failures (i.e., failure of radar components), or pointing-angle errors
Previously,
The basis for the RCA technique relies on the assumption that any variation in ground clutter reflectivity is caused by a change in radar calibration. Persistent ground clutter echoes at low elevation angles generally come from features in the landscape surrounding the radar. Sources of ground clutter can include trees, buildings, towers, or terrain. Often the radar signatures for these features are high reflectivity (
The RCA technique in its current form is not compatible with all ARM radars and their deployments. In this paper, the RCA technique is adapted for and applied to three radar frequencies, C-, X-, and Ka-band, operating at two ARM sites. This extended RCA (eRCA) technique aims to support a wider range of radar frequencies and scan types by modifying the clutter selection procedure. The standard scan strategies of these research radars include plan-position indicator (PPI) and range–height indicator (RHI) scans. The RCA technique, as originally proposed, does not support the use of RHI scans. This paper demonstrates the development and validation of the RCA technique for RHIs using data collected with a scanning C-band radar deployed during a DOE ARM field campaign in the Sierras de Córdoba of Argentina. The RCA technique is then expanded and evaluated for use on higher-frequency (X- and Ka-band) radars, the first known application of its kind. The application to RHI scans is of special interest, as ARM radars do not always conduct PPI scans for field campaigns.
Section 2 describes the datasets used in this study. Section 3 outlines the clutter selection procedure for developing a clutter map, followed by the introduction of an extended clutter selection procedure that helps to improve the operational stability of eRCA for higher-frequency radars, ending with requirements for baseline selection and eRCA calculation. Section 4 presents usage of the extended clutter map selection procedure; a comparison of calibration adjustments derived from the same radar using PPIs and RHIs; and a comparison of RHI-derived eRCA values from C-, X-, and Ka-band radars located at the same site, before discussing limitations of the algorithm. Section 5 summarizes conclusions and suggests future applications of the new methods created in this study.
The United States DOE ARM user facility runs multiple fixed and mobile atmospheric observing facilities spread throughout the world. A wide variety of weather radars are routinely deployed at these sites as part of long-term measurements as well as shorter field campaigns. While the choice of radar configuration at each site depends on the science questions of that site, it is common to have scanning precipitation and cloud radars deployed at the majority of the sites. This work will focus on two sites in particular to highlight improvements made to the RCA algorithm: the Eastern North Atlantic (ENA) site and Argentine site during the Cloud, Aerosol, and Complex Terrain Interactions (CACTI) field campaign. Both of these ARM sites are located in regions with prominent and persistent ground clutter signatures, making them ideal for eRCA technique testing and evaluation.
The RCA technique relies on ground clutter statistics obtained from radar reflectivity. As such, all data used are prior to any clutter filtering applied by the radar processing. All scans are pulled from regular operations of the radar.
The ENA ARM measurement site has been located on Graciosa Island in the Azores archipelago west of Portugal since 2013. Observations from this site provide a rare long-term dataset of marine stratocumulus clouds and their interactions with atmospheric aerosols. A suite of ground-based in situ and remote-sensing instruments are used to characterize meteorological conditions, aerosols, surface state, clouds, precipitation, and radiation.
Characteristics and specifications for ARM radars used in this study.
The site has hosted a high-powered, dual-polarization X-band Scanning Precipitation Radar (XSAPR2) to provide spatial context and measure drizzle and rain rates since 2017. The details of this radar can be found in Table
We utilized almost 3 months of data collected with XSAPR2 in 2018 (13 January–31 March) to evaluate the eRCA technique. Immediately prior to this time period, the radar was visited by engineering staff, and maintenance was performed just before the start of the Aerosol and Cloud Experiments in the Eastern North Atlantic (ACE-ENA) field campaign
The Eastern North Atlantic XSAPR2 PPI from 23 January 2018 at 14:45:10 UTC. The radar returns in the northern sector near the radar are sea clutter, not precipitation. The sea clutter varies day to day, with wind conditions presenting a challenge for the traditional RCA algorithm. The southern sector is not scanned due to mountain blockage as shown by the outline of the island.
The CACTI field campaign
The KaSACR–XSACR deployed at the mobile facility outside of Villa Yacanto, Argentina. The radar is a dual-frequency radar with two antennas mounted on a single pedestal.
The C-band Scanning ARM Precipitation Radar (CSAPR2) is a dual-polarization Doppler weather radar that operates in a simultaneous transmit and receive mode, splitting the transmit signal so that the power is transmitted in both horizontal and vertical polarizations at the same time. The CSAPR2 operates at a frequency of 5.7
The CSAPR2 scan strategy includes hemispherical RHIs at six azimuths (0–180, 30–210, 60–240, 90–270, 120–300, 150–330
The CACTI CSAPR2 PPI from 9 November 2018 at 22:45:03 UTC. The Sierras de Córdoba lie west of the radar, visible in the ground clutter signature. There is also significant clutter due to the hills to the east of the radar. There are also three interference sources visible to the northeast and the northwest. The northwestern source is attributable to a transmission tower, while the northeastern source is the town of Córdoba.
The Scanning ARM Cloud Radars (SACRs) are dual-frequency, dual-polarization Doppler radars mounted on a common scanning pedestal as shown in Fig.
The RCA technique relies on the stability of the radar returns due to ground clutter. It was originally developed in
The simplified form of the radar equation for meteorological targets
The following sections will detail the implementation of RCA, but generally the technique can be broken down into two stages. The first is generation of a clutter map to isolate known clutter points. This forms a baseline from which all changes are taken in a relative sense. The second step is to track the clutter over time and compare it to the baseline time period.
Thresholds and ranges for different radar frequencies for generation of the composite clutter map.
n/a: not applicable.
The RCA technique utilizes the change in the probability distribution of ground clutter reflectivity compared to a baseline clutter reflectivity distribution to identify changes in the radar system and ultimately make an adjustment in the calibration. The adjustment is based on the 95th-percentile clutter area reflectivity, defined in Select all PPIs in a day with no precipitation within 5 Create a fixed polar grid–array
(FPG array) with a resolution of Using the Save the FPG array and repeat for all PPIs in a day. Calculate a percentage of occurrence for each FPG element by summing the saved FPG arrays and dividing by the total number of arrays used. This value may be called “percent on” (PCT_ON) and indicates how often an FPG element contains at least one PPI pixel that exceeds the reflectivity threshold. The final clutter map is defined as the range and azimuth locations where PCT_ON
These steps are translated into a workflow in Fig.
Workflow for designing the clutter map used to calculate the baseline value. The first chart shows how a clutter map is created from a single day of data. The second chart shows how multiple daily maps can be combined into a composite map used in this work that is more resistent to temporal fluctuations.
The clutter map calculated from the CACTI CSAPR2 PPI scans on 9 November 2018. The map predominantly uses the clutter caused by the hills to the west of the radar. The clutter to the east of the radar is out of range of the 10
Figure
To capture the most stable clutter signature and create a clutter map that is used to yield the most stable 95th percentile of the clutter area reflectivity distribution over an observing period, it is beneficial to create a composite clutter map using several single-day clutter maps from the observing period. The composite clutter map method ensures that only the Create daily clutter maps by following the steps outlined in Sect. For each day's clutter map, assign a value of 1 to all FPG elements that are considered clutter points (PCT_ON Sum the FPG elements of all days' clutter maps and divide by the number of days used (
These steps are translated into a workflow in Fig.
Applications of the RCA technique thus far have only included the use of PPI scans
The clutter map calculated from the RHI scans from 9 November 2018 for the CACTI CSAPR2 radar by azimuth angle. Angles for the RHI were every 30
Figure
After a clutter map is constructed, the next step is to determine the baseline value. The baseline is the benchmark with which all other RCA values are compared. The use of a baseline is what makes this a
Using the selected baseline day, we calculate the PDF and cumulative distribution function (CDF) of the clutter area reflectivity using a single-day clutter map (original technique) or the composite clutter map (extended technique). While the FPG clutter map is on a fixed coarsened grid, the PDF and CDF are calculated using all of the range gates within each FPG element that was determined to be clutter. The PDF and CDF show the distribution of the clutter area reflectivity, which may differ from day to day but stabilizes at the 95th percentile of the CDF
Workflow for calculating a baseline value from the composite clutter map and radar files. The second diagram shows how the baseline value can be used to calculate a daily adjustment to reflectivity.
The workflows for baseline and RCA calculation are illustrated in Fig.
Probability density functions and cumulative distribution functions shown from each hour during the selected baseline day, 25 January 2018, at ENA XSAPR2. The 95th percentile is marked with the corresponding reflectivity value also marked at 55.4
Figure
One concern when using RCA is the level of initial calibration. If the initial calibration state of the radar is significantly far off (say
The RCA technique in its current form has some limitations. Since the technique's strength lies in the persistence of ground clutter, the absence of ground clutter makes this technique weak or useless in regions where there is a lack of ground clutter. Though no minimum has been determined, the technique requires enough persistent ground clutter points to function properly over a long period of time.
Daily clutter map for XSAPR2 PPIs at the ENA site for 13 March 2018
RCA time series for XSAPR2 at the ENA site calculated using 13 March 2018 for the clutter map and the baseline
Composite clutter map for XSAPR2 PPIs at the ENA site, created using clutter points from every fifth observation day that occurred at least 80 % of the time. The black denotes those time periods that were used for the composite map. The light gray denotes points that made it into a daily clutter map but were not stable across multiple maps.
The extended RCA technique developed in Sect.
It is important that the generated clutter map only contain actual ground clutter. This can often be made more challenging in the presence of artifacts that can behave similar to ground clutter, such as sea clutter. Failing to estimate a good clutter map can severely affect the calculated RCA values for a radar. If for instance sea clutter points are selected in the map, the day-to-day changes in sea state will be reflected in the 95th percentile of reflectivity over these points.
To demonstrate the importance of clutter map selection Fig.
This problem can be mitigated by using the previously introduced composite clutter map technique. Figure
Using the composite clutter map for ENA XSAPR2 (Fig.
While the RCA method has been previously demonstrated to work for PPIs at S- and C-band
RCA time series for CSAPR2 PPIs and RHIs at the COR site during CACTI. RCA calculated using the composite clutter maps and 20 November 2018 as the baseline. The two radars are within the statistical noise of the eRCA technique. The gaps in the time series are due to periods where the radar was offline.
Figure
Histogram of daily mean RCA values for COR CSAPR2 between 8 November 2018 and 3 March 2019. The variability in RCA values is demonstrated to fall well within
Histograms of RCA values from both PPI and RHI scans are shown in Fig.
One strong limitation of the eRCA technique for RHI scans is that the angles chosen for the RHI must include ground clutter. Similarly, the extent of scanning for the radar needs to include sufficiently low elevations to measure the surrounding ground clutter. For instance, of the 12 azimuths included in the CSAPR2 scan strategy, only 6 of these included points chosen for the ground clutter map. Similarly there needs to be enough points in the RHI scans included as clutter to ensure robust statistics.
Even though the same reflectivity threshold and clutter map development method are used, the clutter map points of the CSAPR2 PPI clutter map and the CSAPR2 RHI clutter map are not necessarily the same points in space due to the differences in scanning. In addition to sample size differences, this difference in points can also account for much of the small-scale variability in the RCA values in Figs.
While we previously showed that the eRCA technique could be run on an X-band radar (XSAPR2 at ARM ENA site), the unique setup at the CACTI site allows for a comparison between the effectiveness of the eRCA technique and direct reflectivity cross calibration. A comparison of reflectivity between systems is one of the most commonly used techniques to cross calibrate co-located radar systems in field campaigns.
RCA time series for CACTI XSACR. Panel
The scatter plot of the difference of RCA values between CSAPR2 and XSACR compared to the difference in mean filtered reflectivities between the two radars. This shows that RCA captures the change in reflectivity calibrations as well as the direct comparison.
The XSACR ran only RHI-type scans for the majority of the CACTI field campaign. As such, the RHI form of the eRCA technique was run on the XSACR radar for the duration of the CACTI campaign and is shown in Fig.
To quantify the uncertainty in the RCA tracking, the difference in RCA values between CSAPR2 and XSACR was calculated and compared to the difference of reflectivity values for matched scans. The reflectivity cross-comparison data were filtered to remove cases for which strong attenuation was expected. For the comparison, reflectivity was subtracted between the two radars using the hemispherical RHI scans along the six scans that occurred every 30
This characterization is important because the X-band is used to characterize the application of eRCA to millimeter-wave radars in the next section.
The wide deployment of Ka-band radars with ARM facilities (where every SACR has a Ka-band component) makes the automatic tracking of calibration for millimeter-wave radars a desirable goal. Millimeter-wave radars, however, can be very sensitive to small perturbations in the environment and operating conditions making tracking of calibration difficult. The application of the eRCA technique to the KaSACR radar was ultimately successful, but processing required additional care and pre-processing steps, as detailed here and in Appendix A.
Sub-daily RCA for KaSACR at the COR site during a 72
The first run of eRCA on the KaSACR data from CACTI resulted in very high variability during many time periods despite the strict attenuation filtering detailed in Appendix A. When periods with rain falling at the radar are removed to limit radome attenuation, much of the high variability persisted. The output of the individual file-level (sub-daily) RCA values shown in Fig.
Interestingly, the comparison hints at the ability for sub-daily corrections using eRCA. While these sub-daily corrections are not fully explored in this paper, the variability in RCA estimates at shorter timescales appears small enough to correct at a time resolution of less than 1 d. This behavior was verified for several different 3 d periods during the campaign.
Panel
Rather than sub-daily corrections due to environmental factors, the goal of this work is to correct systematic biases in radar calibration. Based on this analysis, a relative humidity filter was implemented in the processing for the KaSACR. Figure
The RCA daily values for the CACTI KaSACR radar are shown with a corresponding piecewise polynomial fit. A decreasing trend in power resulted in an increasing trend in the RCA value. On 18 March 2019, the waveguide had an obstruction removed, and a sharp decrease (due to corresponding increase in effective power) can be observed. As such a new polynomial fit is calculated for this time period. However, the general trend of decreasing power and increasing RCA continues. The fit is given relative to the epoch time in seconds of each day.
The scatter plot of the difference of RCA values between the Ka- and X-band radars as compared to the difference in mean filtered reflectivities for the two radars. This shows that RCA captures the change in reflectivity calibrations as well as the direct comparison.
Histograms of daily mean RCA values for COR XSACR and KaSACR between 8 November 2018 and 30 April 2019. The XSACR RCA residual is shown in
After filtering the data for attenuation and high relative humidity, the sub-daily measurements are combined into a daily measurement in Fig.
To both validate and cross-compare the eRCA technique applied at Ka-band, the KaSACR data were cross-compared with the co-aligned XSACR radar. Figure
The KaSACR initially had a large calibration offset of 6–10
The distribution of KaSACR RCA values shows a much greater variability than the C- or X-band. This can be seen in Fig.
Despite greater variability, eRCA can be used to correct systematic changes in the radar calibration at Ka-band. Indeed, eRCA shows potential to correct more transient issues in the calibration.
This study proposes an extension of the RCA technique (eRCA) for use with range–height scans and higher-frequency radars. The eRCA method was successfully applied to X-band PPI scans at the ENA ARM observatory. Increased variability in the clutter field at the ENA site prompted an extension of the clutter map development method, which utilizes only the most stable clutter points from a selection of observed days to establish a composite clutter map. The necessity of a composite clutter map in a region with a variable clutter field is proven with ENA XSAPR2 data. The proposed eRCA clutter map selection method mitigates these situations.
The eRCA technique was successfully extended for use with RHI scans, shown with a comparison of C-band PPI and RHI data from the CACTI field campaign. To accommodate RHI scans, the range of the clutter map is extended out to 40
Results from a cross comparison of CSAPR2, XSACR, and KaSACR at the ARM site during CACTI validated use of the eRCA technique for X-band RHIs. Through a cross comparison of the co-mounted X- and Ka-band radars, the eRCA technique was shown to be valid for Ka-band radars as well as long as additional constraints were applied. The variability in the Ka-band RCA is greater than for X-band and is more sensitive to environmental and operational parameters of the radar. For the unique situation at the CACTI site, attenuation and high relative humidity filters drastically improved the performance of the eRCA technique.
The eRCA technique is not intended for real-time calibration adjustments, but it is useful for a posteriori calibration. The technique is shown to capture changes in calibration due to radar performance as well as engineering changes. For this reason, the eRCA technique can be used as a tool for monitoring the health of the radar. Correcting the data when changes in calibration exceed
Future work with the eRCA technique needs to address the role of changing meteorological conditions and suitability of the locale. Most current applications have been either tropical or oceanic, where the seasonality of ground clutter is limited, or relatively short in duration, which fails to fully capture seasonal variations in the environment. At many mid- and high-latitude locations, the ground clutter signature can change seasonally. For example, snow in the Arctic changes the surface backscattering properties. This necessitates a methodology for determining when to update a clutter map to cope with changing conditions. Similarly, RCA has not to our knowledge been evaluated in environments where the presence of evolving snow and ice conditions can modulate the radar cross section of environmental targets. Finally, this work suggests that eRCA may have the potential for monitoring and correcting radome attenuation, which deserves further investigation.
Higher-frequency radars are subject to larger path attenuation. As such this attenuation needs to be accounted for before using eRCA so that path attenuation does not bias the results of using clutter on days with precipitation. To mitigate the effect of this, data were filtered using a stringent path-integrated attenuation filter. For each radar ray path-integrated attenuation is calculated, and all rays that exceed a set threshold are removed from the 95th-percentile calculation. The equation used for specific attenuation with relation to reflectivity is
Path-integrated attenuation is calculated along each ray as
The co-mounted and co-located radars at the CACTI site allowed for direct comparison of reflectivities in space and time. The scan strategies for the radars were synchronized to provide as much temporal overlap as possible. Given the varying gate spacing and beamwidths of the three radars (see Table
In order to compare reflectivities between radars, we select for light precipitation conditions, focusing on low reflectivities (below 25
Attenuation coefficients used for path-integrated attenuation filtering, where
C-band attenuation not calculated.
Thresholds for weather and data selection used in radar reflectivity cross comparisons.
After discovering a correlation between relative humidity near the surface and Ka-band RCA, a relative humidity filter was applied to remove RCA data points during periods of relative humidity greater than 90 %. Since this filter removes all relative humidity values greater than 90 %, this filter also serves to remove time periods when there was rain on-site over the radar.
The original data used in this study can all be found in the DOE ARM archive (
The code developed in this work is available and can be found on GitHub at
AH was responsible for the implementation and analysis of eRCA code and writing the paper. JCH evaluated and guided the project and wrote the paper. NB evaluated the results. AV evaluated the results and provided guidance on the COR environment. AM made cross comparisons with KAZR radar and conducted additional processing.
The authors declare that they have no conflict of interest.
The authors would like to thank Peter Argay, Vagner Castro, Tercio Silva, Juarez Viegas, Bruno Cunha, Todd Houchens, and Andrei Lindenmaier for their work in deploying and maintaining the radars during the CACTI field campaign that provided many of the data used in this study.
This research has been supported by the US Department of Energy Office of Biological and Environmental Research as part of the Atmospheric Radiation Measurement (ARM) Climate Research Facility, an Office of Science scientific user facility (grant no. 15990). Adam Varble's contributions were supported under the US Department of Energy Office of Biological and Environmental Research as part of the Atmospheric Systems Research Program. The PNNL is operated for the DOE by Battelle Memorial Institute (contract no. DE-AC05-76RL01830).
This paper was edited by Gianfranco Vulpiani and reviewed by two anonymous referees.