Field experiment in South China was undertaken to improve understanding of cloud and precipitation properties. Measurements of the vertical structures of non-precipitating and precipitating clouds were obtained using passive and active remote sensing equipment: a Ka-band cloud radar (CR) system, a C-band frequency modulated continuous wave vertical pointing radar (CVPR), a microwave radiometer and a laser ceilometer (CEIL). CR plays a key role in high-level cloud observation, whereas CVPR is important for observing low- and mid-level clouds and heavy precipitation. CEIL helps us diminish the effects of “clear-sky” in the planetary boundary layer. The experiment took place in Longmen, Guangdong Province, China from May to September of 2016. This study focuses on evaluating the ability of the two radars to deliver consistent observation data and develops an algorithm to merge the CR, CVPR and CEIL data. Cloud echo base, thickness, frequency of observed cloud types and reflectivity vertical distributions are analyzed in the radar data. Comparisons between the collocated data sets show that reflectivity biases between the CR three operating modes are less than 2 dB. The averaged difference between CR and CVPR reflectivity can be reduced with attenuation correction to 3.57 dB from the original 4.82 dB. No systemic biases were observed between velocity data collected in the three CR modes and CVPR. The corrected CR reflectivity and velocity data were then merged with the CVPR data and CEIL data to fill in the gaps during the heavy precipitation periods and reduce the effects of Bragg scattering and fog on cloud observations in the boundary layer. Meanwhile, the merging of velocity data with different Nyquist velocities and resolutions diminishes velocity folding to provide fine-grain information about cloud and precipitation dynamics. The three daily periods in which low-level clouds tended to occur were at sunrise, noon and sunset and large differences in the average reflectivity values were observed. Mid- and high-level clouds tended to occur at 1400 and 1800 BT. Few clouds were found between a height of 3 and 5 km.

In the summertime, convective cells occur frequently over the Tibetan Plateau (TP) because of the large dynamic and thermal effects of the landmass. Measurements of vertical air velocity in convective cloud are useful for advancing our understanding of the dynamic and microphysical mechanisms of clouds and can be used to improve the parameterization of current numerical models. This paper presents a technique for retrieving high-resolution vertical air velocities in convective clouds over the TP through the use of Doppler spectra from vertically pointing Ka-band cloud radar. The method was based on the development of a “small-particle-traced” idea and its associated data processing, and it used three modes of radar. Spectral broadening corrections, uncertainty estimations, and results merging were used to ensure accurate results. Qualitative analysis of two typical convective cases showed that the retrievals were reliable and agreed with the expected results inferred from other radar measurements. A quantitative retrieval of vertical air motion from a ground-based optical disdrometer was used to compare with the radar-derived result. This comparison illustrated that, while the data trends from the two methods of retrieval were in agreement while identifying the updrafts and downdrafts, the cloud radar had a much higher resolution and was able to reveal the small-scale variations in vertical air motion.

The millimeter-wave cloud radar, ceilometer, and disdrometer have been widely used to observe clouds and precipitation. However, there are some drawbacks when those three instruments are solely employed due to their own limitations, such as the fact that radars usually suffer from signal attenuation and ceilometers/disdrometers cannot provide measurements of the hydrometeors of aloft clouds and precipitation. Thus, in this paper, we developed an integrated technology by combining and utilizing the advantages of three instruments together to investigate the vertical structure and diurnal variation of warm clouds and precipitation, and the raindrop size distribution. Specifically, the technology consists of appropriate data processing, quality control, and retrieval methods. It was implemented to study the warm clouds and precipitation in South China during the pre-flood season of 2016. The results showed that the hydrometeors of warm clouds and precipitation were mainly distributed below 2.5 km and most of the rainfall events were very light with a rain rate less than 1 mm h⁻¹, however, the stronger precipitation primarily contributed the accumulated rain amount. Furthermore, a rising trend of cloud base height from 1000 to 1900 BJT was found. The cloud top height and cloud thickness gradually increased from 1200 BJT to reach a maximum at 1600 BJT (Beijing Standard Time, UTC+8), and then decreased until 2000 BJT. Also, three periods of the apparent rainfall on the ground of the day, namely, 0400–0700 BJT, 1400–1800 BJT, and 2300–2400 BJT were observed. During three periods, the raindrops had wider size spectra, higher number concentrations, larger rain rates, and higher water contents than at other times. The hydrometeor type, size, and concentration were gradually changed in the vertical orientation. The raindrop size distributions of warm precipitation in the air and on the ground were different, which can be expressed by $\gamma$ distributions N(D) = 1.49 × 10⁴D^−0.9484exp(−6.79D) in the air and N(D) = 1.875 × 10³D^0.862exp(−2.444D) on the ground, where D and N(D) denote the diameter and number concentration of the raindrops, respectively.

Day-ahead forecasting of solar radiation is essential for grid balancing, real-time unit dispatching, scheduling and trading in the solar energy utilization system. In order to provide reliable forecasts of solar radiation, a novel hybrid model is proposed in this study. The hybrid model consists of two modules: a mesoscale numerical weather prediction model (WRF: Weather Research and Forecasting) and Kalman filter. However, the Kalman filter is less likely to predict sudden changes in the forecasting errors. To address this shortcoming, we develop a new framework to implement a Kalman filter based on the clearness index. The performance of this hybrid model is evaluated using a one-year dataset of solar radiation taken from a photovoltaic plant located at Maizuru, Japan and Qinghai, China, respectively. The numerical results reveal that the proposed hybrid model performs much better in comparison with the WRF-alone forecasts under different sky conditions. In particular, in the case of clear sky conditions, the hybrid model can improve the forecasting accuracy by 95.7% and 90.9% in mean bias error (MBE), and 42.2% and 26.8% in root mean square error (RMSE) for Maizuru and Qinghai sites, respectively.

The Chinese Ka-band solid-state transmitter cloud radar (CR) can operate in three different work modes with different pulse widths and coherent integration and non-coherent integration numbers to meet the requirement for long-term cloud measurements. The CR was used to observe cloud and precipitation data in southern China in 2016. In order to resolve the data quality problems caused by coherent integration and pulse compression, which are used to detect weak cloud in the cloud radar, this study focuses on analyzing the consistencies of reflectivity spectra using the three modes and the influence of coherent integration and pulse compression, developing an algorithm for Doppler spectral density data quality control (QC) and merging based on multiple-mode observation data. After dealiasing Doppler velocity and artefact removal, the three types of Doppler spectral density data were merged. Then, Doppler moments such as reflectivity, radial velocity, and spectral width were recalculated from the merged reflectivity spectra. Performance of the merging algorithm was evaluated. Three conclusions were drawn. Firstly, four rounds of coherent integration with a pulse repetition frequency (PRF) of 8333 Hz underestimated the reflectivity spectra for Doppler velocities exceeding 2 m·s⁻¹, causing a large negative bias in the reflectivity and radial velocity when large drops were present. In contrast, two rounds of coherent integration affected the reflectivity spectra to a lesser extent. The reflectivity spectra were underestimated for low signal-to-noise ratios in the low-sensitivity mode. Secondly, pulse compression improved the radar sensitivity and air vertical speed observation, whereas the precipitation mode and coherent integration led to an underestimation of the number concentration of big raindrops and an overestimation of the number concentration of small drops. Thirdly, a comparison of the individual spectra with the merged reflectivity spectra showed that the Doppler moments filled in the gaps in the individual spectra during weak cloud periods, reduced the effects of coherent integration and pulse compression in liquid precipitation, mitigated the aliasing of Doppler velocity, and removed the artefacts, yielding a comprehensive and accurate depiction of most of the clouds and precipitation in the vertical column above the radar. The recalculated moments of the Doppler spectra had better quality than those merged from raw data.

Radar beam blockage is an important error source that affects the quality of weather radar data. An echo-filling network (EFnet) is proposed based on a deep learning algorithm to correct the echo intensity under the occlusion area in the Nanjing S-band new-generation weather radar (CINRAD/SA). The training dataset is constructed by the labels, which are the echo intensity at the 0.5° elevation in the unblocked area, and by the input features, which are the intensity in the cube including multiple elevations and gates corresponding to the location of bottom labels. Two loss functions are applied to compile the network: one is the common mean square error (MSE), and the other is a self-defined loss function that increases the weight of strong echoes. Considering that the radar beam broadens with distance and height, the 0.5° elevation scan is divided into six range bands every 25 km to train different models. The models are evaluated by three indicators: explained variance (EVar), mean absolute error (MAE), and correlation coefficient (CC). Two cases are demonstrated to compare the effect of the echo-filling model by different loss functions. The results suggest that EFnet can effectively correct the echo reflectivity and improve the data quality in the occlusion area, and there are better results for strong echoes when the self-defined loss function is used.