Evaluation of CloudSat snowfall rate profiles by a comparison with in-situ micro rain radars

The Antarctic continent is a vast desert, the coldest and the most unknown area on Earth. It contains the Antarctic ice sheet, the largest continental water reservoir on Earth that could be affected by the current global warming, leading to sea level rise. The only significant supply of ice is through precipitation, which can be observed from the surface and from space. Remote sensing observations of the coastal regions and the inner continent using CloudSat radar give an estimated rate of snowfall but 5 with uncertainties twice as large as each single measured value, whereas climate models give a range from half to twice the time and spatial average ::::::::: space-time ::::::: averaged : observations. The aim of this study is the evaluation of the vertical precipitation rate profiles of CloudSat radar by comparison with two surface-based Micro-Rain Radars (MRR), located at the coastal French Dumont d’Urville station and at the Belgian Princess Elisabeth station, located in the Dronning Maud Land escarpment zone, respectively. This in turn leads to a better understanding and reassessment of CloudSat uncertainties. We compared a total of 10 four precipitation events, two per station, when CloudSat overpassed within 10 km of the stations and we compared these two different data sets :::::: datasets : at each vertical level. The correlation between both datasets is near-perfect, even though climatic


Introduction
In the context of global warming, predicting the evolution of the Antarctic ice sheet is a major challenge. Snowfall is the principal :::: main input of the ice sheet mass balance, but it is difficult to estimate its amount. Indeed precipitation characteristics depend on the region of Antarctica. In coastal areas, precipitation is influenced by cyclones and fronts (Bromwich, 1988) and a few times a year, these fronts intrude on the high continental plateau, likely bringing most of the snow accumulation (Genthon 10 et al., 2016), the remaining annual precipitation rate being in the form of "Diamond Dust" (thin ice crystals) under clear-sky conditions (Bromwich, 1988;Fujita and Abe, 2006).
Some field campaigns with in-situ observations were conducted to estimate local snow accumulations (Arthern et al., 2006;Eisen et al., 2008), but ground-based measurements are difficult in Antarctica and the size of this continent (twice the size of Australia) does not permit one to cover and study the whole occurrence, rate and distribution of precipitation. Moreover, 15 accumulation observed from stake measurements is a poor proxy for snowfall as it is strongly affected by synoptic upstream conditions :::: local ::::: winds (Souverijns et al., 2018a).
In January 2010, a first micro rain radar (MRR) used for precipitation studies was installed in Antarctica at the Belgian Princess Elisabeth station in the escarpment zone of Dronning Maud Land (PE, 71 o 57'S,23 o 21'E at 1392 above mean sea level) in the context of the Belgian project HYDRANT (The Atmospheric branch of the hYDRological :::::::::::: HYDRological : cycle 5 in ANTarctica) (Gorodetskaya et al., 2015). The PE station is located in the escarpment zone of Dronning Maud Land with Sør Rondande mountains to the south of it (for detailed description of the station meteorological conditions see Gorodetskaya et al. (2013) and Souverijns et al. (2018a) radar observations of precipitation by a scanning X-band polarimetric radar and a K-band vertically profiling micro-rain radar (Grazioli et al., 2017a). A comparison of MRR and CloudSat derived surface snowfall product showed that CloudSat is able to accurately represent the snowfall climatology with biases smaller than 15%, outperforming ERA-Interim (Souverijns et al., 2018b). Moreover, CloudSat's blind zone (lowest measurement available at about 1200 m above the surface) leads to :::::: surface precipitation amounts being underestimated by about 10 % on average although differences during specific events can be much 15 larger (Maahn et al., 2014). This paper focuses on the vertical structure of the precipitation.
As a first step, we characterize the general weather conditions of the four cases (section 3.1). Then, a comparison is done 25 between CloudSat and the vertical MRRs precipitation profiles (section 4.1 and 4.2). From this comparison we highlight a systematic difference (section 4.3), then from a statistical study described in Appendix A, a nearly-perfect correlation between MRR and CloudSat datasets is derived (section 4.4). To conclude, we assess a new range of CloudSat uncertainties at seconds-short-time and 10km-space scales based in this study :::: short :::: time ::::: scale :: (a ::: few :::::::: seconds) ::: and ::::: 10km ::::: space ::::: scale (section 4.4).

CloudSat cloud-profiling radar
The CloudSat cloud-profiling radar is a nadir-looking 94 GHz radar which measures the signal backscattered by hydrometeors.

Radiosondes
A radiosonde is a meteorological device containing a set of sensors to measure the characteristics of the atmosphere from ground level to an altitude ranging from 25 up to 30 km. Parameters measured are temperature, relative humidity, wind speed, wind direction, pressure.

20
At DDU, the used radiosonde system is a METEOMODEM M10. The relative humidity accuracy is 3% and its temporal resolution is 2 s. The temperature measurement is realized every 1s with an accuracy of 0.3°C. At PE, the ground receiving system used are GRAW-GS-E and GRAW radiosondes DFM-09-QRE. Relative humidity is measured with an accuracy of 3% and a temporal resolution of 4 s. The accuracy and the temporal resolution of the temperature measurements are 0.2°C and 3-4 s.

25
3 Meteorological conditions of the four recorded snowfall events

Event characteristics
We summarize in table 1 the characteristics of the only four recorded precipitation cases, when both CloudSat and ground-based MRRs simultaneously record a snowfall event, and when the satellite is in the vicinity of the stations. Due to the CloudSat phase :::: delay :: of :::::: revisit, satellite overflights near the DDU station are located either less than 10 km and then more than 80 km away. CloudSat tracks passing through a radius of 10 km around each station (figure 1) were selected. Each CloudSat flyby over a station takes less than 10 sec and corresponds to ::::: covers a distance between 11.90 km and 17.33 km. We consider that the four associated weather systems are static in regards with CloudSat satellite overfly. However, MRRs are stationary and local precipitation patterns are typically associated with transient large-and meso-scale weather systems. We therefore analyzed the synoptic conditions by using radiosonde data and reanalysis (ERA-Interim) from the European Centre for Medium-Range

5
Weather Forecasts (ECMWF) in order to determine the adequate MRR time-series corresponding to CloudSat observations.
We estimated a duration for which MRR observing conditions agree most with :::: those ::: of CloudSat using the following equation : ∆t avg = ∆x sat V cld ∆x sat V wind ::::: where ∆t avg represents the temporal range of the MRR observations wrapping CloudSat overflight date, ∆x sat is the length of 10 the track inside the 10 km radius area over stations and V cld ::::: V wind : is the vertically averaged wind velocity. All characteristics are shown in table 1.

Events at DDU
The February 17 th 2016 precipitation event at DDU was overflown by CloudSat in the local afternoon. It occurred on the edge of a low pressure system which was approaching the station, in agreement with the radiosounding launched in the morning at 15 09:00 LT. Indeed on figure 2b, 2c, above 1.5 km, a westerly wind brings moisture and a warmer air mass. The radiosounding also shows wind with a continental origin below 1 km which brings a relatively dry air. The recorded precipitation profile (figure 3a) presents a low-level evaporation ::::::::: sublimation : below 1 km and thus suggests that this layer might be dried by continental winds, according to wind direction, relative humidity and temperature profiles.
Located between two low pressure systems, the March 20 th 2016 radiosounding is characterized by a shear between continental and oceanic winds below 500 m, marked by an inversion of relative humidity ( figure 2e, figure 2f). Being at the rear margin of the first passing low pressure system, it explains the easterly origin of oceanic winds. It is followed by a strong 5 event recorded in the afternoon by the radars, with katabatic winds blowing down the ice cap and sublimating precipitation at low altitude below 1000 m (figure 3b). This kind of dry air leading to significant low-level sublimation of snowfall is well documented by Grazioli et al. (2017b).

Events at PE
To analyze the vertical meteorological profiles at the Princess Elisabeth station we used ERA-Interim reanalysis, due to the The fourth observed radiosounding, released 3 hours before the January 13 th 2015 event in the afternoon, is characterized :::::::: explained by a low pressure system located north-west of PE and a strong, constant in altitude, easterly wind (figure 2k). The temperature and relative humidity suggest a cloudy weather with a dryer and hotter boundary layer (figure 2l). The observed 20 precipitation profile suggests in-cloud snowfall and virga (figure 3d). This is confirmed with a backscatter profile measured by a ceilometer installed at PE station (see figure 5 in appendix) observing a passing cloud over the station during the record of the precipitation event by the CloudSat and MRR radars.

Estimation of the confidence in CloudSat reports
All CloudSat measurements were selected within a 10 km-radius from each station and averaged for each vertical bin. A 25 variance on the CloudSat retrievals is computed for the duration of each overpass (see figure 7 in appendix).
4 Results and discussion 5

Precipitation profiles
Focusing on the Dumont d'Urville station, figure 3a shows a good agreement between CloudSat and MRRs snowfall rates for each vertical level. Indeed, averaged satellite precipitation rate at all levels is included within the 95 % MRR confidence interval. The MRR profile presents a maximum of the snowfall rate of 0.75 mm/h at 750 m and an inversion of the precipitation rate likely due to low-level sublimation processes, whereas the ground clutter prevents CloudSat from seeing the inversion.
For each CloudSat vertical bin, we calculated the distance of satellite measurement to the corresponding interpolated MRR observation. We averaged these values by weighting them with the MRR confidence intervals and we found a range of CloudSat 25 uncertainties from -13 % up to +22 %. By applying to CloudSat profiles the calibration difference estimated in the previous section we assessed a new range of uncertainties from -24 % up to +21 %. The applied correction of the difference in calibration shifts the CloudSat distribution of the precipitation rates on the baseline MRR values.

Conclusion
CloudSat remote sensing observations were compared with two in-situ Micro-Rain radars at the coastal French Dumont difference in the CloudSat data set :::::: dataset with respect to the MRR zero snowfall rate signal observations. This difference is statistically estimated at + 0.039 mm/h ± 0.004 mm/h and is presumed to be a MRR uncertainty with CloudSat through a difference in sensitivity between onboard and ground radars, according to their respective frequencies :: for :::: very :::: light ::::::::::: precipitation. :::: This ::::: might :: be :::::::::: precipitable :::::: cloud ::::: water ::::::: recorded ::: by :::::::: CloudSat :: or :: a ::::: MRR :::::::: limitation :::: due :: to :: a ::::: strong :::::::::: attenuation :: of ::: the :::::: signal :::::: through :::::::: important ::::::::::: precipitation. From our correlation and statistical studies based on the quantification of the CloudSat devia-5 tion to the MRR values, and with the correction of this shift, we assessed new CloudSat precipitation uncertainties ranging to -24 ::: -13 % / +21 :: 22 % based on this short-time and small-space scale studythrough MRR. This new assessment of the CloudSat uncertainties, in spite of the limited number of events, provides confidence in the retrieval given the different climatic and geographical conditions of the two stations. It also justifies further analysis of this dataset in this region of the globe, where snowfall is critical and poorly known. Subsequent studies using weak precipitation rates profiles over other Antarctic regions, 10 particularly in the interior of the continent, will strengthen the robustness of this new range of uncertainties and corroborate the difference recorded by both CPR and MRRs. Moreover, the EarthCare spaceborne radar, with a much better vertical resolution, should be even more instructive and improve our understanding of clouds and snowfall in the polar regions, where field observations are so hard to perform.

Calculation of the correlation factor between CloudSat and MRRs
In order to compute the correlation between both datasets, we assume that both the MRRs and CloudSat deviations from the average follow a Gaussian-shaped distribution. MRR data is a Gaussian-shaped distribution, according to its interval confidence ::::::::: confidence :::::: interval : calculation. CloudSat deviation from the mean measurements follows also a Gaussian-shaped distribution, as shown on figure 7. The figure :::::: Figure 4 shows an evident linear fit between both dataset :::::: datasets.