Current Issue
2024 Vol. 30, No. 3
column
2024, 30(3): .
Abstract:
2024, 30(3): 211-222.
doi: 10.3724/j.1006-8775.2024.019
Abstract:
In this study, we simulated the tropical cyclone (TC) Wutip, which originated in the South China Sea in 2013, using three planetary boundary layer (PBL) parameterized schemes within the Weather Research and Forecasting model, i.e., Medium-Range Forecast (MRF), Yonsei University scheme (YSU), and Asymmetric Convective Model Version 2 (ACM2), with different vertical mixing mechanisms. We investigated the effects of different PBL mixing mechanisms on the simulation of TC track, intensity, structure, and precipitation. The results reveal that the surface flux and vertical mixing of PBL jointly influenced the TC throughout its lifecycle, and the simulated TC intensity was closely correlated with the eyewall structure. These three schemes were all first-order and nonlocal closure schemes. However, the MRF scheme was over-mixed, which led to a relatively dryer and warmer near-surface layer, a wetter and colder upper PBL, and thus a simulated eyewall with the smallest wet static energy and weaker convection. Moreover, the MRF scheme produced the smallest 10-m wind speed, which was closest to the observation, and the weakest TC warm-core structure and intensity. The YSU scheme was similar to the MRF scheme, yet it distinguished itself by incorporating an explicit treatment of the entrainment process at the top of the PBL and developing thermal-free convection above the PBL of the eyewall, which significantly increased the wet static energy over the TC eyewall. Thus, the simulated eyewall was more contracted and steeper with stronger upward motion while the eye area became even warmer, finally leading to the strongest TC. The precipitation distribution simulated by the YSU scheme was the most consistent with the observation. The ACM2 scheme used the nonlocal upward and local downward mixed asymmetric convection modes, which reduced the excessive development of thermal-free convection at the eyewall, and avoided restricting the dynamically forced turbulent motion outside the eyewall, leading to a larger radius of the maximum wind speed, and thus more reasonable structural characteristics of PBL and TC intensity. In summary, compared with the YSU scheme and the MRF scheme, the ACM2 scheme demonstrated superior performance in capturing the structure, track, and intensity of Typhoon Wutip. It is important to note that this analysis was based on a specific case study, which might have inherent limitations due to its modest focus.
In this study, we simulated the tropical cyclone (TC) Wutip, which originated in the South China Sea in 2013, using three planetary boundary layer (PBL) parameterized schemes within the Weather Research and Forecasting model, i.e., Medium-Range Forecast (MRF), Yonsei University scheme (YSU), and Asymmetric Convective Model Version 2 (ACM2), with different vertical mixing mechanisms. We investigated the effects of different PBL mixing mechanisms on the simulation of TC track, intensity, structure, and precipitation. The results reveal that the surface flux and vertical mixing of PBL jointly influenced the TC throughout its lifecycle, and the simulated TC intensity was closely correlated with the eyewall structure. These three schemes were all first-order and nonlocal closure schemes. However, the MRF scheme was over-mixed, which led to a relatively dryer and warmer near-surface layer, a wetter and colder upper PBL, and thus a simulated eyewall with the smallest wet static energy and weaker convection. Moreover, the MRF scheme produced the smallest 10-m wind speed, which was closest to the observation, and the weakest TC warm-core structure and intensity. The YSU scheme was similar to the MRF scheme, yet it distinguished itself by incorporating an explicit treatment of the entrainment process at the top of the PBL and developing thermal-free convection above the PBL of the eyewall, which significantly increased the wet static energy over the TC eyewall. Thus, the simulated eyewall was more contracted and steeper with stronger upward motion while the eye area became even warmer, finally leading to the strongest TC. The precipitation distribution simulated by the YSU scheme was the most consistent with the observation. The ACM2 scheme used the nonlocal upward and local downward mixed asymmetric convection modes, which reduced the excessive development of thermal-free convection at the eyewall, and avoided restricting the dynamically forced turbulent motion outside the eyewall, leading to a larger radius of the maximum wind speed, and thus more reasonable structural characteristics of PBL and TC intensity. In summary, compared with the YSU scheme and the MRF scheme, the ACM2 scheme demonstrated superior performance in capturing the structure, track, and intensity of Typhoon Wutip. It is important to note that this analysis was based on a specific case study, which might have inherent limitations due to its modest focus.
2024, 30(3): 223-229.
doi: 10.3724/j.1006-8775.2024.020
Abstract:
In this paper, we utilized the deep convolutional neural network D-LinkNet, a model for semantic segmentation, to analyze the Himawari-8 satellite data captured from 16 channels at a spatial resolution of 0.5 km, with a focus on the area over the Yellow Sea and the Bohai Sea (32°–42°N, 117°–127°E). The objective was to develop an algorithm for fusing and segmenting multi-channel images from geostationary meteorological satellites, specifically for monitoring sea fog in this region. Firstly, the extreme gradient boosting algorithm was adopted to evaluate the data from the 16 channels of the Himawari-8 satellite for sea fog detection, and we found that the top three channels in order of importance were channels 3, 4, and 14, which were fused into false color daytime images, while channels 7, 13, and 15 were fused into false color nighttime images. Secondly, the simple linear iterative super-pixel clustering algorithm was used for the pixel-level segmentation of false color images, and based on super-pixel blocks, manual sea-fog annotation was performed to obtain fine-grained annotation labels. The deep convolutional neural network D-LinkNet was built on the ResNet backbone and the dilated convolutional layers with direct connections were added in the central part to form a string-and-combine structure with five branches having different depths and receptive fields. Results show that the accuracy rate of fog area (proportion of detected real fog to detected fog) was 66.5%, the recognition rate of fog zone (proportion of detected real fog to real fog or cloud cover) was 51.9%, and the detection accuracy rate (proportion of samples detected correctly to total samples) was 93.2%.
In this paper, we utilized the deep convolutional neural network D-LinkNet, a model for semantic segmentation, to analyze the Himawari-8 satellite data captured from 16 channels at a spatial resolution of 0.5 km, with a focus on the area over the Yellow Sea and the Bohai Sea (32°–42°N, 117°–127°E). The objective was to develop an algorithm for fusing and segmenting multi-channel images from geostationary meteorological satellites, specifically for monitoring sea fog in this region. Firstly, the extreme gradient boosting algorithm was adopted to evaluate the data from the 16 channels of the Himawari-8 satellite for sea fog detection, and we found that the top three channels in order of importance were channels 3, 4, and 14, which were fused into false color daytime images, while channels 7, 13, and 15 were fused into false color nighttime images. Secondly, the simple linear iterative super-pixel clustering algorithm was used for the pixel-level segmentation of false color images, and based on super-pixel blocks, manual sea-fog annotation was performed to obtain fine-grained annotation labels. The deep convolutional neural network D-LinkNet was built on the ResNet backbone and the dilated convolutional layers with direct connections were added in the central part to form a string-and-combine structure with five branches having different depths and receptive fields. Results show that the accuracy rate of fog area (proportion of detected real fog to detected fog) was 66.5%, the recognition rate of fog zone (proportion of detected real fog to real fog or cloud cover) was 51.9%, and the detection accuracy rate (proportion of samples detected correctly to total samples) was 93.2%.
2024, 30(3): 230-240.
doi: 10.3724/j.1006-8775.2024.021
Abstract:
Sea surface temperature (SST) is a crucial physical parameter in meteorology and oceanography. This study demonstrates that the influence of earth incidence angle (EIA) on the SST retrieved from the microwave radiation imager (MWRI) onboard FengYun-3 (FY-3) meteorological satellites should not be ignored. Compared with algorithms that do not consider the influence of EIA in the regression, those that integrate the EIA into the regression can enhance the accuracy of SST retrievals. Subsequently, based on the recalibrated Level 1B data from the FY-3/MWRI, a long-term SST dataset was reprocessed by employing the algorithm that integrates the EIA into the regression. The reprocessed SST data, including FY-3B/MWRI SST during 2010–2019, FY-3C/MWRI SST during 2013–2019, and FY-3D/MWRI SST during 2018–2020, were compared with the in-situ SST and the SST dataset from the Operational Sea Surface Temperature and Ice Analysis (OSTIA). The results show that the FY-3/MWRI SST data were consistent with both the in-situ SST and the OSTIA SST dataset. Compared with the Copernicus Climate Change Service V2.0 SST, the absolute deviation of the reprocessed SST, with a quality flag of 50, was less than 1.5℃. The root mean square errors of the FY-3/MWRI orbital, daily, and monthly SSTs, with a quality flag of 50, were approximately 0.82℃, 0.69℃, and 0.37℃, respectively. The primary discrepancies between the FY-3/MWRI SST and the OSTIA SST were found mainly in the regions of the western boundary current and the Antarctic Circumpolar Current. Overall, this reprocessed SST product is recommended for El Ni?o and La Ni?a events monitoring.
Sea surface temperature (SST) is a crucial physical parameter in meteorology and oceanography. This study demonstrates that the influence of earth incidence angle (EIA) on the SST retrieved from the microwave radiation imager (MWRI) onboard FengYun-3 (FY-3) meteorological satellites should not be ignored. Compared with algorithms that do not consider the influence of EIA in the regression, those that integrate the EIA into the regression can enhance the accuracy of SST retrievals. Subsequently, based on the recalibrated Level 1B data from the FY-3/MWRI, a long-term SST dataset was reprocessed by employing the algorithm that integrates the EIA into the regression. The reprocessed SST data, including FY-3B/MWRI SST during 2010–2019, FY-3C/MWRI SST during 2013–2019, and FY-3D/MWRI SST during 2018–2020, were compared with the in-situ SST and the SST dataset from the Operational Sea Surface Temperature and Ice Analysis (OSTIA). The results show that the FY-3/MWRI SST data were consistent with both the in-situ SST and the OSTIA SST dataset. Compared with the Copernicus Climate Change Service V2.0 SST, the absolute deviation of the reprocessed SST, with a quality flag of 50, was less than 1.5℃. The root mean square errors of the FY-3/MWRI orbital, daily, and monthly SSTs, with a quality flag of 50, were approximately 0.82℃, 0.69℃, and 0.37℃, respectively. The primary discrepancies between the FY-3/MWRI SST and the OSTIA SST were found mainly in the regions of the western boundary current and the Antarctic Circumpolar Current. Overall, this reprocessed SST product is recommended for El Ni?o and La Ni?a events monitoring.
2024, 30(3): 241-256.
doi: 10.3724/j.1006-8775.2024.022
Abstract:
Short-duration heavy rainfall (SHR), as delineated by the National Meteorological Center of the China Meteorological Administration, is characterized by hourly rainfall amounts no less than 20.0 mm. SHR is one of the most common convective weather phenomena that can cause severe damage. Short-range forecasting of SHR is an important part of operational severe weather prediction. In the present study, an improved objective SHR forecasting scheme was developed by adopting the ingredients-based methodology and using the fuzzy logic approach. The 1.0°×1.0° National Centers for Environmental Prediction (NCEP) final analysis data and the ordinary rainfall (0.1–19.9 mm h–1) and SHR observational data from 411 stations were used in the improved scheme. The best lifted index, the total precipitable water, the 925 hPa specific humidity (Q925), and the 925 hPa divergence (DIV925) were selected as predictors based on objective analysis. Continuously distributed membership functions of predictors were obtained based on relative frequency analysis. The weights of predictors were also objectively determined. Experiments with a typhoon SHR case and a spring SHR case show that the main possible areas could be captured by the improved scheme. Verification of SHR forecasts within 96 hours with NCEP global forecasts 1.0°×1.0° data initiated at 08:00 Beijing Time during the warm seasons in 2015 show the results were improved from both deterministic and probabilistic perspectives. This study provides an objectively feasible choice for short-range guidance forecasts of SHR. The scheme can be applied to other convective phenomena.
Short-duration heavy rainfall (SHR), as delineated by the National Meteorological Center of the China Meteorological Administration, is characterized by hourly rainfall amounts no less than 20.0 mm. SHR is one of the most common convective weather phenomena that can cause severe damage. Short-range forecasting of SHR is an important part of operational severe weather prediction. In the present study, an improved objective SHR forecasting scheme was developed by adopting the ingredients-based methodology and using the fuzzy logic approach. The 1.0°×1.0° National Centers for Environmental Prediction (NCEP) final analysis data and the ordinary rainfall (0.1–19.9 mm h–1) and SHR observational data from 411 stations were used in the improved scheme. The best lifted index, the total precipitable water, the 925 hPa specific humidity (Q925), and the 925 hPa divergence (DIV925) were selected as predictors based on objective analysis. Continuously distributed membership functions of predictors were obtained based on relative frequency analysis. The weights of predictors were also objectively determined. Experiments with a typhoon SHR case and a spring SHR case show that the main possible areas could be captured by the improved scheme. Verification of SHR forecasts within 96 hours with NCEP global forecasts 1.0°×1.0° data initiated at 08:00 Beijing Time during the warm seasons in 2015 show the results were improved from both deterministic and probabilistic perspectives. This study provides an objectively feasible choice for short-range guidance forecasts of SHR. The scheme can be applied to other convective phenomena.
2024, 30(3): 257-274.
doi: 10.3724/j.1006-8775.2024.023
Abstract:
This study used the China Meteorological Administration (CMA) three-source fusion gridded precipitation analysis data as a reference to evaluate the precipitation forecast performance of the European Centre for Medium-Range Weather Forecasts (ECMWF) model for China from 2017 to 2022. The main conclusions are as follows. The precipitation forecast capability of the ECMWF model for China has gradually improved from 2017 to 2022. Various scores such as bias, equitable threat score (ETS), and Fractions Skill Score (FSS) showed improvements for different categories of precipitation. The bias of light rain forecasts overall adjusted towards smaller values, and the increase in forecast scores was greater in the warm season than in the cold season. The ETS for torrential rain more intense categories significantly increased, although there were large fluctuations in bias across different months. The model exhibited higher precipitation bias in most areas of North China, indicating overprediction, while it showed lower bias in South China, indicating underprediction. The ETSs indicate that the model performed better in forecasting precipitation in the northeastern part of China without the influence of climatic background conditions. Comparison of the differences between the first period and the second period of the forecast shows that the precipitation amplitude in the ECMWF forecast shifted from slight underestimation to overestimation compared to that of CMPAS05, reducing the likelihood of missing extreme precipitation events. The improvement in ETS is mainly due to the reduction in bias and false alarm rates and, more importantly, an increase in the hit rate. From 2017 to 2022, the area coverage error of model precipitation forecast relative to observations showed a decreasing trend at different scales, while the FSS showed an increasing trend, with the highest FSS observed in 2021. The ETS followed a parabolic trend with increasing neighborhood radius, with the better ETS neighborhood radius generally being larger for moderate rain and heavy rain compared with light rain and torrential rain events.
This study used the China Meteorological Administration (CMA) three-source fusion gridded precipitation analysis data as a reference to evaluate the precipitation forecast performance of the European Centre for Medium-Range Weather Forecasts (ECMWF) model for China from 2017 to 2022. The main conclusions are as follows. The precipitation forecast capability of the ECMWF model for China has gradually improved from 2017 to 2022. Various scores such as bias, equitable threat score (ETS), and Fractions Skill Score (FSS) showed improvements for different categories of precipitation. The bias of light rain forecasts overall adjusted towards smaller values, and the increase in forecast scores was greater in the warm season than in the cold season. The ETS for torrential rain more intense categories significantly increased, although there were large fluctuations in bias across different months. The model exhibited higher precipitation bias in most areas of North China, indicating overprediction, while it showed lower bias in South China, indicating underprediction. The ETSs indicate that the model performed better in forecasting precipitation in the northeastern part of China without the influence of climatic background conditions. Comparison of the differences between the first period and the second period of the forecast shows that the precipitation amplitude in the ECMWF forecast shifted from slight underestimation to overestimation compared to that of CMPAS05, reducing the likelihood of missing extreme precipitation events. The improvement in ETS is mainly due to the reduction in bias and false alarm rates and, more importantly, an increase in the hit rate. From 2017 to 2022, the area coverage error of model precipitation forecast relative to observations showed a decreasing trend at different scales, while the FSS showed an increasing trend, with the highest FSS observed in 2021. The ETS followed a parabolic trend with increasing neighborhood radius, with the better ETS neighborhood radius generally being larger for moderate rain and heavy rain compared with light rain and torrential rain events.
2024, 30(3): 275-288.
doi: 10.3724/j.1006-8775.2024.024
Abstract:
Satellite-based and reanalysis precipitation products provide valuable information for various applications. However, their performance varies widely across regions due to different data sources and production processes. This paper evaluated the daily performance of four precipitation products (MSWEP, ERA5, PERSIANN, and TRMM) for seven regions of the Chinese mainland, using observations from 2462 ground stations across the country as a benchmark. We used four statistical and four classification indicators to describe their spatial and temporal accuracy, and capability to detect precipitation events while analyzing their applicability. The results show that according to the precipitation characteristics and accuracy of different types of precipitation products over the Chinese mainland, MSWEP was the most suitable product over the Chinese mainland, having the lowest root mean square error and mean absolute error, along with the highest coefficient of determination. It was followed by TRMM and ERA5, whereas PERSIANN lagged behind in terms of performance. In terms of different regions, MSWEP still performed well, especially in North China and East China. The accuracy of the four precipitation products was relatively low in the summer months, and they all overestimated in the northwest region. In other months, MSWEP and TRMM were better than PERSIANN and ERA5. The four precipitation products had good detection performance over the Chinese mainland, with probability of detection above 0.5. However, with the increase of precipitation threshold, the detection capability of the four products decreased, and MSWEP and ERA5 had good detection capability for moderate rain. TRMM's detection capability for heavy rain and rainstorms was better than that of the other three products, and PERSIANN's detection capability for moderate rain, heavy rain and rainstorms was relatively poor, with a large deviation.
Satellite-based and reanalysis precipitation products provide valuable information for various applications. However, their performance varies widely across regions due to different data sources and production processes. This paper evaluated the daily performance of four precipitation products (MSWEP, ERA5, PERSIANN, and TRMM) for seven regions of the Chinese mainland, using observations from 2462 ground stations across the country as a benchmark. We used four statistical and four classification indicators to describe their spatial and temporal accuracy, and capability to detect precipitation events while analyzing their applicability. The results show that according to the precipitation characteristics and accuracy of different types of precipitation products over the Chinese mainland, MSWEP was the most suitable product over the Chinese mainland, having the lowest root mean square error and mean absolute error, along with the highest coefficient of determination. It was followed by TRMM and ERA5, whereas PERSIANN lagged behind in terms of performance. In terms of different regions, MSWEP still performed well, especially in North China and East China. The accuracy of the four precipitation products was relatively low in the summer months, and they all overestimated in the northwest region. In other months, MSWEP and TRMM were better than PERSIANN and ERA5. The four precipitation products had good detection performance over the Chinese mainland, with probability of detection above 0.5. However, with the increase of precipitation threshold, the detection capability of the four products decreased, and MSWEP and ERA5 had good detection capability for moderate rain. TRMM's detection capability for heavy rain and rainstorms was better than that of the other three products, and PERSIANN's detection capability for moderate rain, heavy rain and rainstorms was relatively poor, with a large deviation.
2024, 30(3): 289-305.
doi: 10.3724/j.1006-8775.2024.025
Abstract:
The onset, evolution, and propagation processes of convective cells can be reflected by the organizational morphology of mesoscale convective systems (MCSs), which are key factors in determining the potential for heavy precipitation. This paper proposed a method for objectively classifying and segmenting MCSs using geosynchronous satellite observations. Validation of the product relative to the classification in radar composite reflectivity imagery indicates that the algorithm offers skill for discriminating between convective and stratiform areas and matched 65% of convective area identifications in radar imagery with a false alarm rate of 39% and an accuracy of 94%. A quantitative evaluation of the similarity between the structures of 50 MCSs randomly obtained from satellite and radar observations shows that the similarity was as high as 60%. For further testing, the organizational modes of the MCS that caused the heavy precipitation in Northwest China on August 21, 2016 (hereinafter known as the "0821" rainstorm) were identified. It was found that the MCS, accompanied by the "0821" rainstorm, successively exhibited modes of the isolated cell, squall line with parallel stratiform (PS) rain, and non-linear system during its life cycle. Among them, the PS mode might have played a key role in causing this flooding. These findings are in line with previous studies.
The onset, evolution, and propagation processes of convective cells can be reflected by the organizational morphology of mesoscale convective systems (MCSs), which are key factors in determining the potential for heavy precipitation. This paper proposed a method for objectively classifying and segmenting MCSs using geosynchronous satellite observations. Validation of the product relative to the classification in radar composite reflectivity imagery indicates that the algorithm offers skill for discriminating between convective and stratiform areas and matched 65% of convective area identifications in radar imagery with a false alarm rate of 39% and an accuracy of 94%. A quantitative evaluation of the similarity between the structures of 50 MCSs randomly obtained from satellite and radar observations shows that the similarity was as high as 60%. For further testing, the organizational modes of the MCS that caused the heavy precipitation in Northwest China on August 21, 2016 (hereinafter known as the "0821" rainstorm) were identified. It was found that the MCS, accompanied by the "0821" rainstorm, successively exhibited modes of the isolated cell, squall line with parallel stratiform (PS) rain, and non-linear system during its life cycle. Among them, the PS mode might have played a key role in causing this flooding. These findings are in line with previous studies.
2024, 30(3): 306-326.
doi: 10.3724/j.1006-8775.2024.026
Abstract:
To investigate the stratosphere-troposphere exchange (STE) process induced by the gravity waves (GWs) caused by Typhoon Molave (2020) in the upper troposphere and lower stratosphere, we analyzed the ERA5 reanalysis data provided by the European Centre for Medium-Range Weather Forecasts and the CMA Tropical Cyclone Best Track Dataset. We also adopted the mesoscale forecast model Weather Research and Forecasting model V4.3 for numerical simulation. Most of the previous studies were about typhoon-induced STE and typhoon-induced GWs, while our research focused on the STE caused by typhoon-induced gravity waves. Our analysis shows that most of the time, the gravity wave signal of Typhoon Molave appeared below the tropopause. It was stronger on the east side of the typhoon center (10°–20°N, 110°–120°E) than on the west side, suggesting an eastward tilted structure with height increase. When the GWs in the upper troposphere and lower stratosphere region on the west side of the typhoon center broke up, it produced strong turbulence, resulting in stratosphere-troposphere exchange. At this time, the average potential vorticity vertical flux increased with the average ozone mass mixing ratio. The gravity wave events and STE process simulated by the WRF model were basically consistent with the results of ERA5 reanalysis data, but the time of gravity wave breaking was different. This study indicates that after the breaking of the GWs induced by typhoons, turbulent mixing will also be generated, and thus the STE.
To investigate the stratosphere-troposphere exchange (STE) process induced by the gravity waves (GWs) caused by Typhoon Molave (2020) in the upper troposphere and lower stratosphere, we analyzed the ERA5 reanalysis data provided by the European Centre for Medium-Range Weather Forecasts and the CMA Tropical Cyclone Best Track Dataset. We also adopted the mesoscale forecast model Weather Research and Forecasting model V4.3 for numerical simulation. Most of the previous studies were about typhoon-induced STE and typhoon-induced GWs, while our research focused on the STE caused by typhoon-induced gravity waves. Our analysis shows that most of the time, the gravity wave signal of Typhoon Molave appeared below the tropopause. It was stronger on the east side of the typhoon center (10°–20°N, 110°–120°E) than on the west side, suggesting an eastward tilted structure with height increase. When the GWs in the upper troposphere and lower stratosphere region on the west side of the typhoon center broke up, it produced strong turbulence, resulting in stratosphere-troposphere exchange. At this time, the average potential vorticity vertical flux increased with the average ozone mass mixing ratio. The gravity wave events and STE process simulated by the WRF model were basically consistent with the results of ERA5 reanalysis data, but the time of gravity wave breaking was different. This study indicates that after the breaking of the GWs induced by typhoons, turbulent mixing will also be generated, and thus the STE.
2024, 30(3): 327-336.
doi: 10.3724/j.1006-8775.2024.027
Abstract:
In the present study, multimodel ensemble forecast experiments of the global horizontal irradiance (GHI) were conducted using the dynamic variable weight technique. The study was based on the forecasts of four numerical models, namely, the China Meteorological Administration Wind Energy and Solar Energy Prediction System, the Mesoscale Weather Numerical Prediction System of China Meteorological Administration, the China Meteorological Administration Regional Mesoscale Numerical Prediction System-Guangdong, and the Weather Research and Forecasting Model-Solar, and observational data from four photovoltaic (PV) power stations in Yangjiang City, Guangdong Province. The results show that compared with those of the monthly optimal numerical model forecasts, the dynamic variable weight-based ensemble forecasts exhibited 0.97%–15.96% smaller values of the mean absolute error and 3.31%–18.40% lower values of the root mean square error (RMSE). However, the increase in the correlation coefficient was not obvious. Specifically, the multimodel ensemble mainly improved the performance of GHI forecasts below 700 W m–2, particularly below 400 W m–2, with RMSE reductions as high as 7.56%–28.28%. In contrast, the RMSE increased at GHI levels above 700 W m–2. As for the key period of PV power station output (02:00–07:00), the accuracy of GHI forecasts could be improved by the multimodel ensemble: the multimodel ensemble could effectively decrease the daily maximum absolute error (AEmax) of GHI forecasts. Moreover, with increasing forecasting difficulty under cloudy conditions, the multimodel ensemble, which yields data closer to the actual observations, could simulate GHI fluctuations more accurately.
In the present study, multimodel ensemble forecast experiments of the global horizontal irradiance (GHI) were conducted using the dynamic variable weight technique. The study was based on the forecasts of four numerical models, namely, the China Meteorological Administration Wind Energy and Solar Energy Prediction System, the Mesoscale Weather Numerical Prediction System of China Meteorological Administration, the China Meteorological Administration Regional Mesoscale Numerical Prediction System-Guangdong, and the Weather Research and Forecasting Model-Solar, and observational data from four photovoltaic (PV) power stations in Yangjiang City, Guangdong Province. The results show that compared with those of the monthly optimal numerical model forecasts, the dynamic variable weight-based ensemble forecasts exhibited 0.97%–15.96% smaller values of the mean absolute error and 3.31%–18.40% lower values of the root mean square error (RMSE). However, the increase in the correlation coefficient was not obvious. Specifically, the multimodel ensemble mainly improved the performance of GHI forecasts below 700 W m–2, particularly below 400 W m–2, with RMSE reductions as high as 7.56%–28.28%. In contrast, the RMSE increased at GHI levels above 700 W m–2. As for the key period of PV power station output (02:00–07:00), the accuracy of GHI forecasts could be improved by the multimodel ensemble: the multimodel ensemble could effectively decrease the daily maximum absolute error (AEmax) of GHI forecasts. Moreover, with increasing forecasting difficulty under cloudy conditions, the multimodel ensemble, which yields data closer to the actual observations, could simulate GHI fluctuations more accurately.
2024, 30(3): 337-350.
doi: 10.3724/j.1006-8775.2024.028
Abstract:
Precipitation nowcasting, as a crucial component of weather forecasting, focuses on predicting very short-range precipitation, typically within six hours. This approach relies heavily on real-time observations rather than numerical weather models. The core concept involves the spatio-temporal extrapolation of current precipitation fields derived from ground radar echoes and/or satellite images, which was generally actualized by employing computer image or vision techniques. Recently, with stirring breakthroughs in artificial intelligence (AI) techniques, deep learning (DL) methods have been used as the basis for developing novel approaches to precipitation nowcasting. Notable progress has been obtained in recent years, manifesting the strong potential of DL-based nowcasting models for their advantages in both prediction accuracy and computational cost. This paper provides an overview of these precipitation nowcasting approaches, from which two stages along the advancing in this field emerge. Classic models that were established on an elementary neural network dominated in the first stage, while large meteorological models that were based on complex network architectures prevailed in the second. In particular, the nowcasting accuracy of such data-driven models has been greatly increased by imposing suitable physical constraints. The integration of AI models and physical models seems to be a promising way to improve precipitation nowcasting techniques further.
Precipitation nowcasting, as a crucial component of weather forecasting, focuses on predicting very short-range precipitation, typically within six hours. This approach relies heavily on real-time observations rather than numerical weather models. The core concept involves the spatio-temporal extrapolation of current precipitation fields derived from ground radar echoes and/or satellite images, which was generally actualized by employing computer image or vision techniques. Recently, with stirring breakthroughs in artificial intelligence (AI) techniques, deep learning (DL) methods have been used as the basis for developing novel approaches to precipitation nowcasting. Notable progress has been obtained in recent years, manifesting the strong potential of DL-based nowcasting models for their advantages in both prediction accuracy and computational cost. This paper provides an overview of these precipitation nowcasting approaches, from which two stages along the advancing in this field emerge. Classic models that were established on an elementary neural network dominated in the first stage, while large meteorological models that were based on complex network architectures prevailed in the second. In particular, the nowcasting accuracy of such data-driven models has been greatly increased by imposing suitable physical constraints. The integration of AI models and physical models seems to be a promising way to improve precipitation nowcasting techniques further.
粤公网安备 4401069904700002号