A long dialogue summary generation model with integrated salience discourse context window sampling method (SDCWS) is proposed according to the characteristics of dialogue corpus. The model is divided into two modules. 1) The salience discourse context window sampling module (CWS) evaluates the dialogue discourse for salience, uses the salient discourse as the sampling anchor point, and then sets the sampling window to extract the discourse adjacent to the left and right of the sampling anchor point together as fragments, containing richer discourse relations. 2) The inter-fragment information fusion summary generation module (IF) uses the transformer block to fuse information from mutually independent fragments, enhancing the semantic relationships between fragments and assigning blended weights to fragments during summary generation. The loss-of-consistency mechanism is used to encourage the salience discourse context window sampling module to determine better sampling anchors. Experimental results on the publicly available query-based long conversation summary dataset QMSum show that scores of the proposed model are significantly higher than the best existing model on the ROUGE evaluation metric.

Based on the observational data of a self-developed fast-response air pressure sensor at the Atmospheric Boundary Layer and Atmospheric Environment Comprehensive Experimental Station in the Horqin area, Inner Mongolia in the summer of 2019, the characteristic parameters of the pressure fluctuations were calculated, and the spectral characteristics of the pressure fluctuations and the characteristics of the pressure standard deviation were studied. The results show that the self-developed fast-response air pressure sensor can reflect the rapid fluctuations of pressure, and the frequency response is close to 1 Hz. The variance spectra of the pressure fluctuations satisfy the n^-2 scaling law in the frequency range from 0.0006 to 0.5 Hz, and the peak frequency is lower than that of the wind speed and temperature. The normalized variance spectra of pressure fluctuations under different atmospheric stabilities merge into a single line in the high-frequency range and distribute around the stability parameter in the low-frequency range. The contribution of pressure fluctuations to turbulent energy is mainly at larger scales, while that of the wind speed and temperature is mainly at smaller scales. The standard deviation and fluctuation intensity of the pressure have obvious diurnal variation characteristics, which is strong during the daytime and weak during the nighttime.

Based on three laser optical Parsivel disdrometers installed at different heights on meteorological tower, the vertical distribution characteristics of raindrop size distribution (DSD) and the influences on rainfall measurement by radar in the lower atmospheric boundary layer were investigated, during 31 precipitation episodes in Beijing from June to August 2020. The results show that the break-up process is dominant from 280 m to 140 m, which leads to the decrease of large scale raindrops and the increase of small scale raindrops. Whereas, from 140 m to ground, the collision-coalescence process is dominant on the falling path of raindrops, which leads to the decrease of large and small scale raindrops and the increase of medium scale raindrops. Because of the great influence on DSD by break up, large scale raindrops in convection rain decrease rapidly with height. The DSD of convective rain in Beijing is different from oceanic-like cluster and continental-like cluster. Due to the change of DSD with height, characteristic parameters of rainfall vary nonlinerly. A Z-R mismatch error would be caused when using ground precipitation data for correction of the meteorological radar rainfall estimation, which is heavier in stratiform and light rain.

Aiming at the problem that the measurement results of OTT-Parsivel in elevated environment are prone to error due to the influence of ambient air flow, the authors propose the detection test of OTT-Parsivel on unstable platform, proposes the quality control method under different levels of wind speed, gives the measurement speed correction model, and gives simulation and error analysis. Based on the summer precipitation process in Beijing in 2020, several precipitation cases are analyzed by using the measured data of OTT-Parsivel, three-dimensional ultrasonic wind thermometer and dynamic inclination sensor at 280 meters. The results show that the proposed model has a good correction effect on the vertical falling velocity of raindrops, the velocity distribution of each dimension tends to be normal distribution, and the number of effective particles can increased by 6%. When the horizontal wind speed exceeds 8 m/s, the availability of OTT-Parsivel raindrop spectrometer measurement data is improved.

To analyze the data detected by optical disdrometer under the influence of environment wind, this paper uses data measured by OTT-Parsivel and ultrasonic anemometer during several precipitation processed in 2018 in Beijing, and studies how horizontal and vertical wind affect the measurement of disdrometer respectively. When horizontal wind speed is less than 10 m/s and vertical wind speed is less than 0.4 m/s, more than 80% vertical wind speed estimated by velocity of hydrometeors from OTT-Parsivel are consistent with that given by anemometer. While the correction of velocity of hydrometeors brings significant change to raindrop whose diameter is about 1 mm, thereby affects other parameters. Moreover, this paper defines two feature parameters calculated from size and velocity of hydrometeors: the average of standard deviation and the ratio of outlier, which can distinguish more than 80% samples with horizontal wind speed more than 10 m/s.