Sensitivity of microwave radiances at 85–183 GHz to precipitating ice particles



[1] The response of passive microwave observations to ice particle scattering and surface emissivity has been studied for frequencies at 85, 150, and 183 GHz. Using two- and three-dimensional simulations of different precipitation events, we found that a channel around 150 GHz generally exhibits the strongest scattering signature due to precipitation-sized ice particles. This channel is only moderately affected by variations in surface emissivity. The sensitivity of the 150 GHz channel to variations in cloud water is about 25 (Kg−1 m3). In comparison, the sensitivity to changes in surface emissivity is about 5 K/(10%) for cold, dry atmospheres and less than 1 K/(10%) for typical midlatitude atmospheres. Channels at around 85 GHz are much stronger affected by variable surface emissivity (sensitivities are up to 15 K/(10%) and show on average a 2 to 2.5 times smaller scattering signature (10 Kg−1 m3). The sensitivity of a water vapor sounding channel at 183–7 GHz to ice particle scattering strongly depends on the environmental conditions but is in general about a factor of 1.5 to 2.5 smaller than at 150 GHz. This is due to the contribution of water vapor in and above the cloud to total emission, masking the scattering signal. The weighting functions of channels at 183–3 and 183–1 GHz peak too high up in the atmosphere and only show a weak or negligible response to the precipitation events under investigation. Based on these results, frequencies around 150 GHz, possibly in combination with channels at 85 and 183–7 GHz seem most appropriate for the detection and retrieval of precipitation properties at middle and high latitudes.

1. Introduction

[2] With the success of the Tropical Rainfall Measuring Mission (TRMM) passive microwave estimates of precipitation have become a standard tool to retrieve precipitation properties. Future satellite missions such as the Global Precipitation Measurement (GPM) are likely to extend TRMM-like observation capabilities to higher latitudes. Middle and high-latitude precipitation poses additional constraints on the design of passive microwave sensors. In particular, the frequent occurrence of frozen precipitation makes it necessary to study the possible benefits of high-frequency channels (larger than 100 GHz), which are more sensitive to scattering by precipitation-sized ice. Currently only the cross-track scanning Advanced Microwave Sounding Unit-B (AMSU-B) onboard NOAA-15 and NOAA-16 is equipped with such channels. In the near future, the Special Sensor Microwave Imager Sounder (SSMIS) will also contribute global high-resolution measurements at various high frequencies.

[3] For nonprecipitating liquid clouds, Muller et al. [1994] quantified the relative surface contribution with less than 20 % at 157 GHz for a surface emissivity of 0.9 while at 89 GHz this sensitivity increases by a factor of 1.5–2. In case of precipitation occurrence, the information content of observations at 85 GHz and above depends crucially on the assumptions made for particle size distributions and particle density, and mixing rule [Gasiewski, 1992; Burns et al., 1997; Skofronick-Jackson and Wang, 2000; Bennartz and Petty, 2001]. In particular both 89 and 150 GHz observations together allow a better retrieval of precipitating ice microphysical properties using the differential scattering signature [Skofronick-Jackson and Wang, 2000; Weng and Grody, 2000].

[4] In the current investigation we therefore examine the sensitivity of different high-frequency channels to ice particle scattering and to various environmental parameters that might affect the observed signal. This is carried out using both model simulations and observations to distinguish between effects of water vapor, cloud water, precipitating ice, and surface emissivity. Even though we limit our investigation to higher latitudes the conclusions are rather general and also apply to tropical cloud types.

[5] In section 2, three-dimensional radar observations over Northern Europe serve as input to radiative transfer simulations. A wide range of atmospheric conditions is analyzed to differentiate between atmospheric, cloud, and surface contribution, respectively. The specific response of precipitating vs. nonprecipitating hydrometeors is demonstrated by a simulation study in section 3 based on two-dimensional spectral cloud model-radiative transfer output. All results are summarized and discussed in section 4.

2. Sensitivity to Surface Emissivity and Environmental Parameters

2.1. Input Data

[6] The first set of model simulations are based on coincident radar volume scans and SSM/I data that were taken over the island of Gotland in the center of the Baltic Sea in 1995 [Bennartz and Michelson, 2001] during the Pilot Study on Intensive Data Collection (PIDCAP) within the framework of the Baltic Sea Experiment (BALTEX). The input data, the radiative transfer model, and the approach of simulating a realistic response to ice particle scattering have been described in detail by Bennartz and Petty [2001]. Therefore we will only briefly summarize the relevant information on the three different precipitation events which have been used

  1. The first case is a frontal precipitation event covered by the F-10 SSM/I on 29 August 1995. The frontal system was associated with a low-pressure system with its center over Poland which was slowly moving to the northeast. The location of the front itself was at that time about 100 km to the east of Gotland. The water vapor fields derived from the SSM/I show a strong west-east gradient reaching from 17 kg/m2 in the western parts of the Baltic Sea to 30 kg/m2 in the eastern parts. Little or no liquid water was retrieved from the SSM/I in the western parts and more than 1 kg/m2 near the Polish coast. Observed brightness temperature depressions due to scattering at 85 GHz were of the order of 35 K.
  2. The second case has been observed on 29 September 1995, when the Baltic region was influenced by advection of an exceptionally cold arctic air mass. The synoptic situation was triggered by an extensive trough which had formed over western Europe and which persisted over 5 days from 27 September to 3 October 1995. As a consequence, extensive convection cells were formed over the North Sea and the Baltic sea as well as during daytime over Scandinavia and northern Germany and Poland. The average SSM/I retrieved water vapor path was 14 kg/m2. According to surface station reports the convective cells were associated with intensive graupel and rain showers, accumulated rainfall rates for some precipitation stations exceeded 10 mm/day.
  3. The third case is an intensive thunderstorm centered north of Gotland. This convective event was part of a cluster of cells associated with a low-pressure system which was moving northward from Germany toward Finland. The precipitation station at Visby (on Gotland) which is located at the southern edge of the thunderstorm reported 12 mm precipitation for 24 August 1995. In Stockholm, more toward the northwest, rain totals were only about 4 mm. Stations that were more directly affected by other cells of the cluster reported up to 30 mm precipitation. The observed radar reflectivities exceed 50 dBZ and the maximum vertical extent of the convective system was about 10 km with an average water vapor path of 34 kg/m2. The 85 GHz SSM/I brightness temperatures (TB) in the core of the convective cell reached minimum values below 200 K. Background values in the precipitation-free regions were between 260 K and 265 K.

[7] The radar volume scans for these three cases were taken as input for three-dimensional Monte-Carlo radiative transfer simulations. The surface emissivity was varied systematically between 0.45 and 0.95, thus covering the range between very moist and dry as well as snow covered surface types. All other simulation parameters, such as water vapor path, cloud liquid water, precipitation type and intensity where chosen as described by Bennartz and Petty [2001].

2.2. Simulation Results

[8] Figure 1 shows the results of the sensitivity studies for the three different cases. The left panels represent the average simulated brightness temperature depression for all areas with significant precipitation (radar derived rain rate larger than 0.2 mm/h (using Marshall-Palmer Z-R conversion)) as a function of surface emissivity. The depression has been calculated as the difference between the brightness temperatures of precipitation-free backgrounds and those obtained in the precipitation areas.

Figure 1.

The left panels show the sensitivity of the scattering signal simulated for three different typical high-latitude precipitation events as a function of surface emissivity. The right panels show the sensitivity of the same channels with respect to a 10% variation in surface emissivity. Simulations were performed for frequencies 85.5, 150, 183–7, 183–3, 183–1 GHz.

[9] It can be seen that for more transparent atmospheres, especially for the cold air outbreak with graupel showers (Figure 1, middle panel) the two window channels at 85 GHz and 150 GHz exhibit a moderate to strong dependence on surface emissivity. As the surface emissivity increases the precipitation-free background gets warmer so that the scattering of precipitation-sized ice particles leads to a stronger depression. As the atmosphere becomes more optically thick, either through increased water vapor (third case, intensive convection) or through a combination of increased cloud liquid water and water vapor (first case, front) the contribution of radiation emitted from the surface decreases. For the case of intensive convection, only the 85 GHz TBs show a dependence on the surface emissivity, whereas for the frontal case the observed depression is virtually independent from surface contributions in all channels.

[10] The channels around the 183.31 GHz water vapor absorption line are not sensitive to surface emissivity at all. However, for almost all cases the observed scattering signal at 183.31–7 GHz is about a factor of two smaller than for the 150 GHz channel. This is due to the increased impact of water vapor in and above the clouds that tends to mask the scattering signal. At 183.31-3 GHz only a very small scattering signature can be observed for very dry atmospheres (graupel shower) and at 183.31–1 GHz the precipitation events can not be detected at all.

[11] The sensitivity of the different channels to scattering by precipitation-sized ice has to be related to the impact of uncertainties and variations in the surface emissivity itself. The right panels of Figure 1 show the change of the TBs resulting from an increase of surface emissivity by 0.1. The three absorption channels around 183.31 GHz do not exhibit any variation due to changes in surface emissivity since the atmosphere is already opaque at those frequencies. The channel at 150 GHz only exhibits a moderate response of about 5 K/(0.1) for the dry atmosphere (graupel showers) and about 1 K/(0.1) for the two other cases. The 85 GHz channel is most sensitive to the variations in surface emissivity.

[12] These results suggest that a channel around 150 GHz might indeed be best suited to identify and retrieve precipitation at high latitudes. While the use of sounding channels (183–X GHz) might be advantageous because of even lesser sensitivity to surface emission, the relatively shallow precipitation at high latitudes reduces the use of those channels whose weighting functions peak at middle and higher levels in the troposphere. Only for very dry atmospheres and low surface emissivity the 183–7 GHz channel shows an equally high sensitivity as the channel at 150 GHz.

3. Sensitivity to Variations in Hydrometeor Concentration

[13] Another sensitivity study was carried out to investigate the relative response of the above frequencies to changes in liquid and ice concentrations based upon cloud model simulations since these provide consistent hydrometeor distributions under controlled conditions. The cloud model was developed at the Hebrew University of Jerusalem [Khain and Sednev, 1995] and the simulations were kindly provided by Prof. A. Khain.

[14] The model is nonhydrostatic and computes prognostic fields for seven hydrometeor classes, namely cloud and rain water, plates, columns, dendrites, snow, graupel, and hail. An advantage is the calculation of explicit size spectra with 33 mass classes per hydrometeor thus approximations of the radiative effects of cloud evolution by parameterized spectra are avoided. The investigated case represents a simulation over the eastern Mediterranean with convection initiated by advection of moist maritime air across the coastline followed by an extended stratiform tail [Khain and Sednev, 1996; Bauer et al., 2000]. This simulation was processed at a horizontal resolution of 3 km and at a vertical resolution of 0.4 km.

[15] Examples of hydrometeor water contents are shown in Figure 2 for time step t = 180 min. Please note that surface rain rates do not exceed 11 mm/h in the convective core (x = 80–90 km) while in the stratiform area the rain rates remain below 5 mm/h. Sensitivity of brightness temperatures to perturbations in hydrometeor contents, w, are expressed as Jacobians, i.e.,

equation image

where ϵ is the surface emissivity and w is the hydrometeor density in g/m3. The Jacobian is then given in units Kg−1m3. The radiative transfer simulations were carried out for a fixed surface temperature with an ocean surface for x < 75km and a land surface elsewhere. For the latter a constant emissivity of 0.95 was assumed while for sea water an explicit calculation of sea water emissivity was implemented. All simulations refer to a constant zenith angle of 53°.

Figure 2.

Cross section of contents [in g/m3] of (a)cloud and rain liquid water, (b) ice crystals, (c) snow, and (d) graupel+hail from two-dimensional spectral cloud model simulation.

[16] Figure 3 shows the Jacobians for 89.0, 150.0, 183−7, and 183−3 GHz with respect to cloud and rain liquid water. Interesting is the dipole between cloud droplets and rain drops in Figure 3a. An increase in cloud water results in an increase in TB at all frequencies where this increase is almost twice as high at 150 GHz compared to 89 GHz. At 183−7 the effect is reduced by the strong relative absorption contribution of water vapor as mentioned previously. However, the simulations show that this channel is still very sensitive to cloud water at lower levels (here 4–6 km altitude). The sensitivity to rain is expressed as negative ΔTB in Figure 3a due to scattering at raindrops. Even though this effect is more pronounced at higher frequencies than 89 GHz, the opacity of the cloud layers above make changes in rain intensity invisible. The 183−3 GHz channel (Figure 3d) is only weakly responding to changes in cloud water in the convective core.

Figure 3.

Jacobians for cloud and rain liquid water at (a) 89.0, (b) 150.0, (c) 183 − 7, and (d) 183 − 3 GHz.

[17] The differential scattering effects of snow is illustrated in Figures 4. Here, the reduction in TBs at 150 GHz is about 3 times as strong as at 89 GHz. Also at 183−7 the scattering is more efficient than at 89 GHz but less efficient than at 150 GHz as described earlier. Comparing Figure 4 and Figure 2 shows that the entire snow layer is well identified using both 89 and 150 GHz signatures. Therefore both channels together present an excellent tool for snow parameter retrieval. Cloud liquid water effects can be corrected through the positive relative response of TBs with frequency. That scattering is more efficient compared to cloud water absorption is indicated by the response at 183−3 GHz where the brightness temperature perturbation still reaches −10 K per g/m3. The results for variations in graupel (not shown here) were similar to the results for snow. Please note that in all cases the contribution of surface emissivity has little influence on the detectability of snow and graupel as well as cloud liquid water variations. Therefore a combination of the 89–183 GHz channels is applicable to all surfaces.

Figure 4.

As Figure 3 but for snow.

4. Conclusions

[18] The sensitivity of brightness temperatures at around 85.0, 150.0, 183−7, and 183−3 GHz to variations in hydrometeor content, different precipitation events, and surface emissivity were examined. The results were derived on the basis of three-dimensional Monte-Carlo radiative transfer simulations with combined radar and passive microwave observation of higher latitude precipitation as well as on spectral cloud model simulations of a midlatitude system representing shallow convection with extensive trailing stratiform evolution. In summary the following conclusions were derived:

  1. The results suggest that a channel around 150 GHz contains significant information to identify and retrieve frozen precipitation at middle and high latitudes. While highly sensitive to variations in ice scattering, it is only moderately affected by variations in surface emissivity. These results suggest that in addition to the typical channels between 18 GHz and 90 GHz a 150 GHz window channel should be included in forthcoming missions that also aim at snowfall/high-latitude precipitation.
  2. In principle, the use of water vapor sounding channels (183-X GHz) gives additional information because of its little sensitivity to surface emission. However, the relatively shallow precipitation at high latitudes limits the usefulness of those channels. Only for very dry atmospheres and low surface emissivity the channel at 183–7 GHz shows an equally high sensitivity to ice particle scattering as the channel at 150 GHz. Since for many surface types the emissivity can be derived with a sufficient accuracy (water surfaces, nonscattering land surfaces), a 150 GHz might be considered superior to (183–X GHz) if resources do not allow for a combination of both.
  3. At 150 GHz the sensitivity to cloud water is about 2–2.5 higher than at 85–89 GHz with equation image At 183−7 GHz, the stronger sensitivity to cloud water is compensated by water vapor absorption but still 1.5 times higher than at 85–89 GHz. ΔTBw is positive since the response is due to absorption by droplets.
  4. ΔTBw is negative for rain intensity variations because at these frequencies raindrops scatter the incoming radiation rather efficiently. At 89 GHz the magnitude is similar to that for cloud water. At 150 GHz and above the rain layer is not directly visible therefore ΔTBw ≈ 0.
  5. The response to snow and graupel is negative and increases strongly with frequency. The brightness temperature reduction is about 2–3 times higher at 150 GHz than at 89 GHz and almost as strong at 183−7 GHz. At 183−3 GHz less than half the scattering effect is left compared to 183−7 GHz since the weighting function peaks at higher altitudes.
  6. From the simulations, all Jacobians show sufficient sensitivity to hydrometeor content variations over land surfaces to be used in surface-independent retrieval schemes.


[19] The authors are indebted to A. Khain and A. Prokrovsky for providing the cloud model simulations.