The core device of the observation system is a dual-frequency (18.7 and 36.5 GHz), dual-polarized (horizontal and vertical) microwave radiometer (RPG-4CH-DP) produced by Radiometer Physics GmbH, Germany. The RPG-4CH-DP radiometer is a high-performance instrument with a direct detection receiver and a completely automatic calibration system. The radiometer is mounted on an accurate elevation/azimuth positioner so that the whole system can perform scans in any direction from the sky to the ground, thereby realizing complex scanning schemes, such as all-sky monitoring and all-round monitoring of the ground. The RPG-4CH-DP can distinguish cloud/raindrop particles during precipitation and monitor soil moisture and vegetation parameters by using signals with different polarizations. Both frequencies of 18.7 and 36.5 GHz have been widely combined to detect snow depth and snow water content and are frequently used in most spaceborne microwave imagers, such as the SSM/I, AMSR-E (Advanced Microwave Scanning Radiometer for EOS), and GMI (GPM Microwave Imager) sensors. The directly observed surface emissivities at these two frequencies can provide highly accurate references for the verification and assimilation of spaceborne microwave observations.

The RPG-4CH-DP radiometer has a comparable half-power beam width of approximately 6∘ and a calibration accuracy of ± 1 K. Currently, the height of the instrument above the ground is 2.5 m, which results in a half-power footprint width of 0.22 m on average. More details regarding the instrument specifications for the RPG-4CH-DP are shown in Table 1.

Currently, the RPG-4CH-DP provides only the basic brightness temperature (Tb) data in four channels without other related products. By incorporating the auxiliary observations from the observation system, we broadened the application of the instrument, denoted RPG-XCH-DP, thereby providing not only the basic microwave radiance, but also the complex surface emissivity.

The multitarget mobile system comprises a track, a mobile platform, a driving system, and a control unit. As the sketch of the mobile system in Fig. 2 shows, the 25 m track is parallel to the test plots with an observation interval of 0.3 m. The mobile platform placed on the track is a metal box 4 m in length, 0.8 m in height, and 1.0 m in depth. The driving system includes a stepper motor, transmission mechanism, and communication cable connected to the mobile platform and power supply. The control unit consists of a single-chip microcomputer, timer, and stepper motor driver, which can set the moving time and control the operation of the driving device. The control device is installed on the mobile platform and connects both driving devices.

In this experiment, to obtain the microwave emissivity over different surfaces in near-simultaneous time, the RPG-4CH-DP is mounted on the mobile platform and moves back and forth along the track. The communication system for receiving the data and the power supply are placed in the metal box. According to the commands from the single-chip microcomputer and the driving force from the stepper motor, the mobile platform moves along the track similar to a small train, and the onboard radiometer scans the five test plots at fixed times every day.

Figure 3 shows a sketch of the observation field, including the five test plots distributed along the 25 m track. Currently, five surface types are considered in the observation field, namely, water, cement, sand, soil, and grass. For the water body, a plastic pool 6 m long and 2.4 m wide is used to hold the water. The adjacent cement surface consists of a 2 m wide footpath. The remaining three plots of sand, bare soil, and grass are the same size (approximately 6 m long by 4 m wide) and are separated by a distance of approximately 2 m.

To scan each plot at the same place at a fixed time, five touching switches corresponding to the center of each plot are fixed on the track to stop the moving platform so that the radiometer can scan the same place for a couple of minutes. By using this mobile platform, the ground-based radiometer can scan multiple surfaces almost simultaneously (i.e., within 1 h), thereby providing valuable measurements for understanding the variation in surface emissivity over different land surfaces with different characteristics.

The auxiliary data include mainly the surface temperature, soil temperature, and soil moisture. Five thermometers with a PT100 temperature sensor made by Honeywell are placed separately on each test plot to measure their surface temperature. In addition, an SI-111 precision infrared radiometer developed by Apogee Instruments Inc. is fixed on a stand of the RPG-4CH-DP radiometer to obtain the surface temperature of each plot while the microwave radiometer is moving. Furthermore, a set of soil temperature and moisture sensors is fixed at three soil depths, 5, 10, and 20 cm, to detect the subsurface soil temperature and humidity. To monitor the real-time working situation of the whole observation system, a digital video camera is installed near the field to record the states of the mobile platform and radiometer as well as changes in the weather, such as the presence of cloud cover, rain, or snow.

To directly obtain the surface emissivity, a combined mode of ground observations at multiple elevation angles and zenith observations is designed, in which the former monitors mainly the surface radiance, while the latter monitors the sky radiance in the same 1 h period.

The ground observation mode is illustrated in Fig. 4. The mobile platform is triggered every hour, and the microwave radiometer operates using the ground scanning mode at this time. The scan is performed from the horizon (0∘) to the ground, and the elevation angle is defined as the angle between the scanning direction and the horizontal. A negative value indicates an angle below the horizon, which is equivalent to 90∘-θ, where θ is the incident angle, an important parameter for describing spaceborne radiometer scanning. The radiometer is 2.5 m above the ground, so it can scan each test plot with a length of 6 m when the elevation angle is between -21 and -72∘, as shown in Fig. 4. The valid elevation angle range for water is different due to the different length of the pool. To determine the surface emissivity variation with the elevation angle, the radiometer is set to scan each test plot with an angle interval of 3∘ from -21 to -45∘, an angle interval of 5∘ from -45 to -70∘, and then back to -21∘ to scan the test plot repeatedly during the ground observation mode. To acquire ground observations over all five test plots within 1 h, each plot is given 9 min; in other words, the mobile platform will move to the cement plot at 9 min, the sand plot at 18 min, the bare soil plot at 27 min, and finally the grass plot at 36 min. After finishing the ground observations in all five test plots, the mobile platform will begin to move back at 45 min and reach the beginning location after approximately 6 min. During the return trip, the scan mode changes to the zenith observation mode so that the radiometer scans from the ground to the sky. When the elevation angle is raised to 90∘, the radiometer will continually acquire zenith observations for approximately 5 min to obtain the sky radiance. After obtaining these zenith observations, the elevation angle changes from the zenith observation mode to the ground observation mode at -21∘ so that the radiometer is already in the ground observation mode when the next measurement cycle arrives. In this way, the radiometer on the mobile platform can obtain not only the ground radiance over five test plots, but also the sky radiance within a 1 h period. Here, we assume that 1 h is short enough to neglect the minute-scale differences in the surface and sky radiance, and thus, the mobile system can obtain the microwave emissivity over different surfaces nearly simultaneously.

Since a scanning angle of 36∘ is equivalent to an incident angle of 54∘ used for many spaceborne microwave imagers, such as AMSR-E (55∘) or SSM/I (53∘), we first compare the variation in the observed Tb over different surfaces at an elevation angle of 36∘. As Fig. 5a shows, the changes in the observed Tb at 36.5 GHz in horizontal (Tb36h) and vertical (Tb36v) polarization over the four land surfaces within 24 h (Beijing time, BJT) are quite similar, with smaller values at night and larger values in the afternoon. Less variation in the radiance is noted at Tb36v (not shown), but more significant variations are detected at Tb36h over the four surfaces (Fig. 5a): the observed Tb36h from grass is approximately 270–285 K but varies within 240–270 K over sand and bare soil and reaches only 200–230 K for cement. The observed Tb at 18.75 GHz within 24 h shows similar variations, with only slight changes among the different land surfaces. Likewise, the corresponding polarization differences (V-H) of Tb within 24 h are very similar to one another, so both DTb18vh and DTb36vh at 02:00 (BJT) are shown in Fig. 5b, revealing a slight difference (close to zero) for grass but considerably larger differences for water and cement (almost up to 70 K for water) and smaller differences over sand and soil (below 30 K). In addition, the values of DTb18vh are larger than those of DTb36vh. The Tb polarization difference is more significant over water than over land and is closely related to the roughness of the land surface. In addition, the roughness of grass is obviously larger than that of the other three land surfaces and thus scatters more surface radiance and reduces the polarization difference. Therefore, the observed Tb polarization differences over the different surfaces shown in Fig. 5b appear reasonable, and the given quantitative polarization differences for certain surfaces can serve as a valid reference for identifying land surfaces and water bodies.

To study the variations in Tb at more than a single angle, Fig. 6a shows the changes in the observed Tb with the elevation angle ranging from 24 to 65∘ over the four land surfaces. The horizontally polarized Tb is clearly more sensitive to the land surface type than the vertically polarized Tb with increasing elevation angle; in particular, Tb36h rises rapidly from 180 to 240 K over cement but slowly increases from 240 to 260 K over sand and bare soil and remains almost constant over grass. In contrast, the variations in the vertically polarized Tb with increasing elevation angle are similar among the land surfaces and are smaller than those in the horizontally polarized Tb, showing a decreasing trend from 280 to 260 K over different surfaces. The variations in the observed Tb over water are presented in Fig. 6b. Different from the above observations over land surfaces, the vertically polarized Tb over water obviously reduces from 200 to 140 K with increasing elevation angle, while the horizontally polarized Tb slowly rises from 100 to 120 K, almost opposite to the Tb polarization variations over land surfaces. The corresponding changes in the polarization difference of Tb at 18.75 GHz (DTb18vh = Tb18v-Tb18h) over all five classes of surfaces are further plotted in Fig. 6c. In general, the Tb polarization difference decreases with increasing elevation angle, and the varied ranges with the elevation angle over the five classes surfaces in Fig. 6c are similar to those in Fig. 5b; thus, the decreasing trend is most obvious over water and cement and less evident over grass with increasing elevation angle. The variations of the Tb polarization difference at 36.5 GHz with the elevation angle are similar to those at 18.75 GHz over all five test plots.

Combining the surface and sky Tb radiance with the surface temperature derived from the infrared sensor, the surface emissivity (ε) is derived from Eq. (2). Since the diurnal variation of ε is more constant and less significant than that of the Tb radiance, the surface emissivity observed at 02:00 (BJT) is chosen for the following investigation. First, the polarized ε at both 18.75 and 36.5 GHz and their polarization differences at an elevation angle of 36∘ are compared in Fig. 7a. The vertically polarized ε(εv) is clearly much larger than the horizontally polarized ε(εh), and the ε values at the same frequencies are close, but the ε values over water are smaller than those over the four land surfaces due to quite different dielectric constants. The εh values obviously differ among the four land surfaces, although their corresponding εv values are relatively similar, exceeding 0.95, which indicates that εh is more sensitive to land surface variability than εv. The εh is lower than 0.75 over cement but increases to 0.90 over sand and bare soil and up to 0.97 over grass. Thus, the emissivity polarization difference (εv-εh) shown in Fig. 7b is obvious over water (>0.3) and cement (approximately 0.25) but reduces to 0.1 over sand and 0.05 over bare soil and almost 0.01 or close to zero over grass; this trend is similar to that of the Tb polarization difference shown in Fig. 5b. Emissivity polarization difference is more significant over water than over land due to different surface reflectivity and dielectric constant properties. Among four land surfaces, εv-εh over cement is most obvious and over grass is slight, which is closely related to land surface roughness. The polarized difference of both Tb and emissivity demonstrated that surface roughness over grass is obviously larger than that over the other three land surfaces, especially smooth cement surfaces, thus generating more volume scattering by vegetation and weakening the polarization difference over grass.

In addition to investigating the variations at a fixed angle, the variations in ε at multiple elevation angles over the four land surfaces are compared in Fig. 8a. Because εh is more sensitive to surface type than to water, when the elevation angle changes from -24 to 65∘, εh clearly rises from 0.65 to 0.85 over cement, followed by sand and bare soil, with εh increasing from 0.85 to 0.95, and εh is constant at 0.95 over grass. The corresponding εv values over the four land surfaces are closer and exhibit a slightly decreasing trend within the range of 0.9–1.0 with increasing elevation angle.

Different from land surfaces, the ε values over water in Fig. 8b show considerably different variation trends with the elevation angle: when the elevation angle changes from -33 to -65∘, εv clearly reduces from 0.7 to 0.5, while εh slightly increases near to 0.4. The corresponding ε polarization differences (εv-εh) over five surfaces (Fig. 8c) present a decreasing trend with increasing elevation angle, and the larger the ε polarization difference in Fig. 7b is, the greater the variation with the elevation angle in Fig. 8c is; i.e., the decreasing trend is most obvious over water and smooth cement but slightly changes over grass with increasing elevation angle. The variation in the ε polarization difference at 36.5 GHz with the elevation angle is similar to that at 18.75 GHz over all five test plots (results not shown).