Current Issue

Vol.35, NO.5, 2024

Display Method:
Articles
Short-term Precipitation Correction Based on GRU Deep Learning
Zeng Xiaotuan, Zou Chenxi, Fan Jiao, Wang Qingguo, Huang Dajian, Liang Xiao, Ding Yuqin, Tan Zhao
2024, 35(5): 513-525. DOI: 10.11898/1001-7313.20240501
Abstract:
To improve the accuracy of short-term precipitation forecasts, a deep learning method is proposed to correct numerical model precipitation forecast products. This method extracts spatiotemporal features from numerical model forecasts and observations using a neural network and performs corrections based on a gated recurrent unit (GRU) framework. Additionally, an atmospheric physics adaptor module is meticulously designed to address systematic errors in the intensity and displacement of numerical model forecast by leveraging physical condition mechanisms. The module plays a crucial role within the overarching model framework, which consists of three integrated components: Feature network, recurrent-revising network, and physical adaptor. The feature network extracts precipitation intensity, distribution, motion characteristics and other related atmospheric physical features from precipitation in situ and numerical model forecast data, serving as input to the recurrent-revising network. Recurrent-revising network utilizes a recurrent neural network structure to adjust grid point forecast results on a time-step basis. Deep neural networks are used to extract spatiotemporal variation features from numerical model forecast data and historical observations, learning systematic errors in the evolution processes to correct the precipitation magnitude and distribution. The physical adaptor is an atmospheric physics adaptation module, which preprocesses numerical model forecast data using frequency distribution fitting and distribution deviation correction methods. In Guangxi convective-scale model precipitation forecast data, when there are significant differences between numerous samples and the precipitation in situ, the feature correlation is low, making it a challenge to capture systematic error characteristics during neural network training. By preprocessing the samples with the physical adaptor, differences between forecasts and observations are reduced, enhancing feature correlation between training input datasets and observations, thus facilitating better neural network training and achieving superior correction skills. This method not only adheres to but also integrates fundamental atmospheric physical laws governing precipitation evolution, thereby offering a robust and innovative approach for post-processing numerical model short-term precipitation forecasts. By incorporating these physical principles into the model framework, corrected forecasts not only reflect statistical patterns but also adhere closely to the underlying physical processes driving precipitation dynamics. Experimental results in Guangxi indicate that the model demonstrates positive threat score skills across various forecast times and precipitation intensities. Specifically, for different precipitation intensities (average of all times), threat score skills for 0.1 mm·h-1, 2 mm·h-1, 7 mm·h-1, 15 mm·h-1, 25 mm·h-1, and 40 mm·h-1 are 5.67%, 3.59%, 2.18%, 1.46%, 1.01%, and 0.46%, respectively; for different lead times, threat score skills for 0-2 h, 2-4 h, and 4-6 h are 4.77%, 1.28%, and 0.91%, respectively; and the overall average threat score skill across all precipitation intensities and times is 2.21%.
Validation and Evaluation of Ocean Calibration Accuracy of FY-3G Precipitation Measurement Radar
Yuan Mei, Yin Honggang, Shang Jian, Jiang Baisen, Yang Runfeng, Gu Songyan, Zhang Peng
2024, 35(5): 526-537. DOI: 10.11898/1001-7313.20240502
Abstract:
FY-3G precipitation satellite launched in April 2023 is the first dedicated precipitation measurement satellite in China. The dual-frequency precipitation measurement radar (PMR) is the core instrument on the satellite. Because the backscattering performance of the vast ocean area is relatively stable, the calibration accuracy of the on-orbit radar can be tested by studying the backscattering cross-section of the sea surface. FY-3G PMR level 1 data in July 2023 and GPM DPR (global precipitation measurement, dual-frequency precipitation radar) level 2A data are used to analyze the mean value and mean square error of the global sea surface backscattering cross section under no-rain conditions to evaluate the radar performance. At the same time, the theoretical model of ocean surface backscattering is studied to simulate the sea surface backscattering cross-section under the condition of no rain, and the sea surface backscattering cross-section is compared with the actual radar measurement, so as to realize the preliminary evaluation of FY-3G PMR calibration accuracy. Furthermore, the accuracy of FY-3G PMR calibration is evaluated by the ocean calibration test results of GPM DPR data. Test results of ocean calibration accuracy show that when the incidence angle of FY-3G PMR Ku-band is less than 15°, the deviation between the observed value and the model simulation value is small. The deviation of FY-3G PMR ranges from 1.65 to 2.73 dB, while the standard deviation ranges from 0.74 to 1.82 dB. The deviation of FY-3G PMR Ka-band at an 18° incidence is less than 0.27 dB, and the standard deviation of the deviation is 3.49 dB. The calibration deviation of FY-3G PMR and GPM DPR is relatively constant, with the difference is primarily attributed to the backscattering statistical characteristics of the data itself. The stability of the backscattering data of FY-3G PMR Ku- and Ka-band sea surfaces at each incidence angle is comparable to that of GPM DPR. Gas attenuation is not considered at the moment. In the future, the impact of gas attenuation on the Ku- and Ka-band ocean calibration accuracy validation will be further evaluated, and the systematic deviation of FY-3G PMR will be corrected.
Validation and Correction of FY-4B/GIIRS Temperature and Humidity Profiles Based on Radiosonde Data
Jin Ziqi, Yu Zhenshou, Hao Shifeng, Zhang Honglei, Lu Zhengqi, Zhang Shuxian
2024, 35(5): 538-550. DOI: 10.11898/1001-7313.20240503
Abstract:
In order to promote applications of FY-4B satellite data, temperature and humidity profile products of FY-4B geostationary interferometric infrared sounder (GIIRS) are verified and evaluated from February 2023 to January 2024 based on radiosonde data. Deviation characteristics are compared and analyzed under different conditions. In addition, the probability density function (PDF) matching method is employed to correct systematic errors in FY-4B/GIIRS temperature profile under cloudy condition. Results indicate that the quality of FY-4B/GIIRS temperature and humidity profiles is significantly influenced by cloud activity, leading to a notable reduction in the proportion of high-quality data when affected by the cloud. Under clear sky condition, the mean bias (MB) of temperature profiles ranges from -0.3 K to 1 K, the root mean square error (RMSE) is within 2 K, and the minimum error is approximately 1.1 K near 400 hPa height. The MB of humidity profiles ranges from 0 to 1.3 g·kg-1, and the maximum RMSE is about 2.1 g·kg-1 at the surface layer. Temperature and humidity profile errors increase under cloudy condition, while the bias of entire atmospheric layer is predominantly positive. The RMSE of temperature ranges from 2.2 K to 2.7 K, while the maximum RMSE for humidity is approximately 3 g·kg-1. The trend of errors is consistently similar at 0000 UTC and 1200 UTC. Compared with 0000 UTC, the deviation of temperature profiles at the surface layer at 1200 UTC is larger and slightly more distinct. The humidity error at 1200 UTC is greater than that at 0000 UTC at the layer below 400 hPa under clear sky condition, while the humidity error at 0000 UTC is greater than that at 1200 UTC at layer between 750 hPa and 950 hPa under cloudy condition. Significant systematic errors exist in temperature and humidity profiles under cloudy condition. Samples with quality control of 1 tend to be colder and drier compared to those with quality control of 0. The deviation distribution is more discrete, while the deviation of temperature follows an asymmetric bimodal distribution. After correction using the PDF method, systematic errors of FY-4B/GIIRS temperature profiles are effectively reduced. MBs of samples with quality control of 0 and 1 decrease from 0.74 K and 2.07 K to 0.03 K and 0.01 K, and RMSEs decrease from 1.89 K and 3.20 K to 1.73 K and 2.34 K, respectively. When the deviation is generally unbiased, the effectiveness of PDF methods is limited.
Relative Humidity Correction Method of Microwave Radiometer Combined with Cloud Radar
Zhang Ting, Jiao Zhimin, Mao Jiajia, Zhang Xuefen, Wang Yanfei, Chen Peiyu, Jin Long
2024, 35(5): 551-563. DOI: 10.11898/1001-7313.20240504
Abstract:
The microwave radiometer can detect and retrieve temperature and humidity profiles with high spatial and temporal resolution throughout the day. However, microwave radiometers have few detection frequencies in the middle and upper layers, making them easily affected by clouds. After integrating cloud information into brightness temperature data, the improvement in detection accuracy in the middle and upper layers still remains insufficient, failing to meet accuracy standards required for relative humidity. With the construction of a national ground-based remote sensing vertical profile system, the continuous observation of cloud radar and microwave radiometer at the same site has been achieved, enhancing the spatial and temporal resolution. Combined with the relationship between humidity and cloud formation, a comprehensive quality control method is proposed for relative humidity using cloud radar and microwave radiometer. It plays a crucial role in enhancing the accuracy of humidity profile of microwave radiometer under cloudy conditions.By analyzing the characteristic relationship between the cloud radar reflectivity factor and the radiosonde relative humidity, a piecewise correction method for the microwave radiometer relative humidity of the combined cloud radar is proposed. Error correction results are analyzed using the radiosonde and ERA5 data. It shows that there is a positive linear correlation between the relative humidity and the reflectivity factor, the relative humidity in the middle of the cloud region is approximately saturated, and the relative humidity variation trend with the height of the cloud exit region and the cloud entry region is approximately symmetric about a certain height. Under the condition of stratiform clouds, the root mean square error of relative humidity by microwave radiometer decreases by 7.99% and 8.91% when comparing with radiosonde and ERA5, and the absolute value of median deviation decreases by 12.62% and 13.05%, respectively. The absolute median deviation also decreases. Further investigation indicates the method is also effective under the condition of convection cloud, but the relative humidity in the cloud region after correction is larger than that of sounding and ERA5, and the median deviation changes from negative deviation to positive deviation. Therefore, the relative humidity segment correction method of combing cloud radar can realize the continuous real-time correction of the relative humidity profile of microwave radiometer, which partly improves the observation quality of microwave radiometer under cloud conditions.
S-band and X-band Radar Observation Characteristics of EF2 Tornado at Qingyuan of Baoding in 2021
Chen Xuejiao, Hua Jiajia, Pei Yujie, Wang Zhenchao, Liu Shu, Liu Shujun, Wang Fuxia
2024, 35(5): 564-576. DOI: 10.11898/1001-7313.20240505
Abstract:

Using multiple observations such as S-band radar (SPOL) in Shijiazhuang, X-band phased array radar (XPAR) in Xiong'an, and ground-based encrypted automatic stations, detection features and evolutions of EF2 tornado at Donglü Village, Qingyuan District of Baoding City Hebei Province on 21 July 2021 are studied. The tornado occurred within the center of high dew point values and in an area characterized by a significant temperature gradient. There are convergence lines within the center of high dew point temperatures and a temperature gradient zone. From perspectives of environmental conditions such as convective available potential energy (CAPE), 0-6 km vertical wind shear, and the lifting condensation level (LCL), there is a possibility for tornado occurrence. It is evident that the tornado formed within a low-vortex precipitation cloud system, showing significant divergence at high altitudes. The subsequent storm propagation leads to multiple single-cell mergers and a supercell formation. A significant reflectivity factor core moving from southeast to northwest is observed at the top of hook echo, corresponding to the tornado location. Both SPOL and XPAR detected continuous mesocyclones on average radial velocity images, with dimensions ranging from 1.4 to 4.2 km, and rotating speeds of 10-20 m·s-1, indicating weak mesocyclones with short durations (30-35 min). During tornado development, a decrease in the lower angle detection of adjacent rotational speed pairs coincides with mesocyclone downward extension to 1.2-1.4 km and its diameter shrinking to 0.8-1 km, indicating tornado formation. Tornado storm parameters show maximum rotation speed and vorticity at low levels, promoting its intensification. Compared with XPAR storm parameters, SPOL features a larger maximum reflectivity factor (noless than 55 dBZ) and a greater distribution height (8-10 km). The consistency of SPOL and XPAR in detecting the tornado location, radial velocity, and storm diameter is compared. On the radar radial velocity image, there are pairs of positive and negative velocity values arranged symmetrically along the radial direction. The echo top of XPAR radar is approximately 6 km higher than that of SPOL radar, and the peak time of XPAR echo coincides with the storm's appearance. The tornado vortex signature (TVS) reaches its strongest period from 1536 BT to 1542 BT, extending vertically up to 2-4 km.

Characteristics and Evolution of Radar Polarization During Extremely Persistent Heavy Rainfall
Feng Jinqin, Pan Jiawen, He Qingfang, Lai Qiaozhen
2024, 35(5): 577-589. DOI: 10.11898/1001-7313.20240506
Abstract:
Based on data of S-band dual-polarization Doppler radar, automatic weather station, disdrometer, 2-D lightning locator and Doppler radar wind field retrieval method, mesoscale structure and cloud microphysical characteristics of an extremely persistent heavy rainfall occurred in southwest Fujian on 27 May 2022 are analyzed. Results indicate that the event takes place under the southwest flow on the south side of the low-level shear. Sufficient water vapor, moderate unsteady convective stratification, low lifting condensation height, and convective condensation height over the rainstorm area are all favorable for producing high-efficiency heavy rainfall. Strong echoes (no less than 45 dBZ) persist over the rainstorm area during heavy rainfall. The strong echo center is concentrated on the windward side of the mountain, located at the contraction of the topography of the trumpet opening to the southwest. The wind field retrieval shows that the strong echo persists for an extended period at the convergence of wind speed and the convergence of southerly and northerly airflow. During the first two stages, a strong echo continuously moves into the rainstorm area from the west, generating the train effect of backward propagation. In the third stage, the strong echo moves southeast under the guidance of northeast winds at the middle and upper levels. This process is dominated by oceanic convective rainfall and warm rain. Heavy rainfall is primarily composed of raindrop particles with high concentration and small scale. The lower layer is primarily composed of raindrop particles with high concentration and smaller scales. Raindrop particles in the middle layer are larger than those in the lower layer. Due to strong upward motion, negative flashes occur when graupel particles above 0 ℃ layer collide with ice crystals during the second stage. Due to the ice phase process, large ice phase particles like graupel particles fall, melt, coalesce, and merge with smaller raindrops, resulting in the formation of larger raindrops and the production of highly efficient rainfall. The development of KDP above 0 ℃ layer indicates an increase in rainfall, forecasting an advance of 6-20 minutes for the strengthening of relative surface ground rainfall. A large number of raindrop particles are primarily concentrated at the confluence of air currents. Hydrometeor accumulates here. The prolonged and intense echo causes the accumulation of water condensate over an extended period, ultimately causing heavy rainfall. There are high concentrations of raindrop particles in the middle and lower layers. The high value of ZDR is mostly concentrated in the middle layer updraft area. The high-value ZDR area and the high-value KDP area do not completely overlap. The distribution of ZDR is wider than that of KDP. Large raindrops break into small raindrops as they fall, increasing the number of raindrop particles.
Comparison of Two Damaging Wind Events Caused by Strong Downbursts
Guo Feiyan, Ding Feng, Chu Yingjia, Lang Jiahe, Li Xiaodong, Luan Zaimao
2024, 35(5): 590-605. DOI: 10.11898/1001-7313.20240507
Abstract:
Multi-source observations are used to comprehensively analyze the Doppler weather radar characteristics of strong storms and the formation mechanisms of surface damaging wind induced by two strong downbursts, occurred in Shandong on 2 June 2017 and 6 August 2017. It's found that two damaging wind events occurs under strong synoptic scale weather system forcing and favorable meso-scale environmental conditions, the relatively isolated supercell storm on 2 June (6·2 supercell) and strong single storm on 6 August (8·6 strong single storm) fiercely develop into series of downbursts, which leads to the occurrence of large-scale damaging winds. As two strong downbursts induced by 6·2 supercell storm and 8·6 strong single storm evolves, the vertically integrated liquid water content radar parameter first rises and then quickly plunges down. During 6·2 supercell storm, a powerful downburst descends sharply, causing the mesocyclone's top and bottom to shoal and its thickness to decrease. The occurrence of two strong downbursts are accompanied by obvious radar characteristics including reflective factor core rapid decline, high value area of radial velocity at low elevation, strong divergence at bottom, remarkable mid altitude radial convergence and severe divergence at upper-level. 6·2 supercell storm is characterized by intense rotation. Its mesocyclone lasts for a long time and extends deeply both upwards and downwards. Additionally, there are arc-shaped inflow notches and hook echoes at low levels of 6·2 supercell storm. 8·6 strong single storm is characterized by significant low-level convergence. Besides, there is a convergence zone at or near the surface formed by the outflow of the strong downburst and the inflow in front of 8·6 strong single storm. Among all the formation mechanisms of two damaging winds induced by two strong downbursts, the negative buoyancy effect of two storms is basically equivalent, but the cold pool outflow effect is more evident for 6·2 supercell storm, and the downward transport momentum effect is more significant for 8·6 strong single storm. The Nansun Station of Weifang locating right ahead of 8·6 strong single storm's approaching direction, therefore the front divergence flow from the strong downbursts is preferable superimposed on the fast moving homodromous storm, indicating the speed of the front divergence flow better superimposes on the speed of the storm itself, which is crucial for the occurrence of 37.0 m·s-1 extreme wind.
Remote Sensing Study on Blue-sky Days in Beijing, Tianjin, and Hebei During the Period of 2000-2023
Yan Hao, Liu Guiqing, Cao Yun, Mo Jianfei, Sun Yinglong, Chen Zixuan, Cheng Lu
2024, 35(5): 606-618. DOI: 10.11898/1001-7313.20240508
Abstract:

Blue sky often represents better air quality and lower air pollution. Using satellite aerosol optical depth (AOD) data of Beijing, Tianjin and Hebei Province from 2000 to 2023, combined with the blue-sky data observed at noon time in 2023, a blue-sky grade index is established based on satellite AOD, in which the monitoring index of blue-sky grade is the AOD at 550 nm less than 0.36, and that of deep blue-sky grade is the AOD at 550 nm less than 0.2.Spatial and temporal characteristics of blue-sky grade days in Beijing-Tianjin-Hebei Region from 2000 to 2023 are investigated. Results show that the multi-year average blue-sky days in Beijing, Tianjin, and Hebei are 144.2 d·a-1, 96.3 d·a-1, and 119.6 d·a-1, respectively, with the highest number of blue-sky days for Beijing, followed by Hebei and the lowest number is recorded in Tianjin. In terms of spatial distribution, the northern part of Hebei has the highest annual average of blue-sky days, while the southern part of Hebei has the lowest number of blue-sky days. The number of blue-sky days in Beijing-Tianjin-Hebei exhibits noticeable seasonal changes, with the highest number of blue-sky days in winter and autumn, followed by spring, and the lowest in summer.From 2001 to 2023, the average annual number of clear-sky days in Beijing, Tianjin, and Hebei takes on an increasing trend, with an increase of 18.1 d, 22.3 d and 16.3 d per decade, respectively. There is no significant trend change from 2001 to 2013. However, the annual average blue-sky days in Beijing-Tianjin-Hebei from 2013 to 2023 all show increasing trends, with increments of 26.9 d, 46.5 d, and 36.4 d per decade, respectively. The annual average blue-sky days and deep blue-sky days in Beijing-Tianjin-Hebei from 2013 to 2023 are higher than those in 2001-2013, with the annual average blue-sky days in Beijing, Tianjin, and Hebei of 153.5 d, 107.5 d, and 128.5 d, respectively, which are 17.5 d, 21.8 d, and 16.9 d higher than those in 2001-2013. It may be largely due to the implementation of regional air pollution prevention and control measures, which have led to a reduction in atmospheric particulate matter concentration since 2013.

Influence of Different Sowing Dates on Yield and Quality of Corn Xianyu 335
Song Yanling, Zhou Guangsheng, Guo Jianping, Pan Yaru, Yang Mengjiao, Tian Jinfeng, Li Xiangxue, Meng Xiangyi, Lan Huiting, Jiang Weiguang, Sui Dan, Zhou Lingyu, Shi Junchen, Nie Chang, Man Yi
2024, 35(5): 619-628. DOI: 10.11898/1001-7313.20240509
Abstract:
Using data from Yushu Agricultural Meteorological Station of Jilin from 2018 to 2023, the impact of different sowing dates of corn is investigated focusing on its growth and yield composition as well as grain quality under global warming. It is also debated whether adjusting the sowing date of corn could be a measure for agriculture to adapt to climate change. Results show that the utilization efficiency of accumulated temperature during the growing season of corn varies with different sowing dates. The accumulated temperature is the highest in the first sowing date and lowest in the fourth sowing date, with an average decrease of 8.3% compared to the first sowing date. Different sowing dates of corn have an impact on the growing period. The duration of the first sowing date for corn is extended by an average of 7.5 days compared to the normal sowing date, while durations of the third and fourth sowing dates are shortened by 5.7 days and 13.8 days, respectively. Different sowing dates have an impact on the yield structure of corn. In the first sowing date, there is an increase in the weight of 100 grains of corn in 2 years during 6 years, while a decrease by 4.8% and 8.7% in the third and fourth sowing dates compared to the normal sowing date. The average number of grains per plant in the first sowing date increases by 0.2%, while decreases by 6.0% and 9.3% in the third and fourth sowing date. Overall, delaying the corn sowing date by 10 days and 20 days results in an average yield reduction of 10.9% and 17.1%. Sowing corn 10 days earlier could increase the yield in some years. The change in sowing dates has little effect on the quality of grains. Therefore, an early corn sowing date can be utilized as a strategy to adapt to climate change in certain regions of Northeast China.
Water-nitrogen Managements for Spring Maize at Tuquan, Inner Mongolia Based on APSIM
Guo Erjing, Yang Feiyun, Wu Lu, Sun Shuang, Gao Jiabao, Zhang Chaoqun, Zhang Ling
2024, 35(5): 629-640. DOI: 10.11898/1001-7313.20240510
Abstract:
Water and nitrogen are critical factors that constrain the sustainable production of dryland agriculture. With increasingly severe crisis of water and nitrogen resources and environment, exploring and optimizing water-nitrogen managements, and hence achieving coordinated and unified resource conservation, high and stable grain production, and high efficiency are of great significance for agricultural development. Key parameters of APSIM (agricultural production system simulator) are calibrated and validated based on spring maize phenology, yield, and field management data from Tuquan, Inner Mongolia Autonomous Region from 2013 to 2022. Combined with meteorological data from 1981 to 2022 at Tuquan, water-nitrogen management scenarios are designed under different water deficit levels. Optimal water-nitrogen managements for spring maize at Tuquan are proposed based on indicators including spring maize yield, irrigation amount, nitrogen application amount, water productivity, and agronomic efficiency of applied nitrogen. Furthermore, the suitable irrigation and nitrogen application amounts for spring maize under different precipitation year types are analyzed. Results show that normalized root mean squared errors of the simulated and observed days from emergence to flowering, days from emergence to maturity, and yield of spring maize are 1.3%, 1.2% and 2.8%, respectively. APSIM can quantitatively simulate the growth period and yield of spring maize. Based on the principle that yield, water productivity, and agronomic efficiency of applied nitrogen of spring maize do not significantly decrease compared to the maximum values of all scenarios, and irrigation and nitrogen application amounts do not significantly increase compared to the minimum values of all scenarios, four management measures with no significant differences can be selected, namely, starting automatic irrigation when water deficit reaches 40%, 50%, 60% at the depth of 0-100 cm, and when the water deficit reaches 60% at the depth of 0-60 cm. Among them, the optimal auto-irrigation management for spring maize at Tuquan is to apply irrigation when the water deficit reaches 60% at the depth of 0-100 cm. In this scenario, the irrigation amount is 171.0 mm, and the nitrogen application amount is 197.8 kg·hm-2. When the precipitation during the spring maize growing season is 200-400 mm, the appropriate irrigation amount is 233.0-283.5 mm, and the nitrogen application amount is 176.9-219.3 kg·hm-2. When the precipitation during the spring maize growing season is 401-600 mm, the appropriate irrigation amount is 110.5-148.4 mm, and the nitrogen application amount is 218.3-241.5 kg·hm-2, respectively. When the precipitation during the spring maize growing season is 601-800 mm, the suitable irrigation amount is 125.0-155.0 mm, and the nitrogen application amount is 211.8-249.9 kg·hm-2. This study provides a quantitative reference for utilizing crop mechanism models in real-time monitoring, diagnosis, and precise management of crop water and nitrogen.