Advanced Research in Precipitation Measurements

Accurate high-spatial-resolution precipitation is significantly important in hydrological and meteorological modelling, especially in rain-gauge-sparse areas. Some methods and strategies have been applied for satellite-based precipitation downscaling, residual correction and precipitation calibration. However, which downscaling scheme can provide reliable high-resolution precipitation efficiently remains unanswered. To address this issue, this study aimed to present a framework combining the machine learning downscaling algorithm and post-process procedures. Firstly, four ML-based models, namely support vector regression, random forest, spatial random forest (SRF) and eXtreme gradient boosting (XGBoost), were tested for downscaling and compared with conventional downscaling methods. Then, the effectiveness of the residual correction process using ordinary Kriging and the calibration process using the geographical difference analysis (GDA) method was investigated. The results showed that the ML-based methods had better performance than the conventional regression and interpolation approaches. The SRF and XGBoost outperformed others in generating accurate precipitation estimation with a high resolution. The GDA calibration process significantly improved the downscaled results. However, the residual correction process decreased the downscaling performance of the ML-based models. Combining the SRF or XGBoost downscaling algorithm with the GDA calibration method could be a promising downscaling scheme for precipitation data. The scheme could be used to generate high-resolution precipitation, especially in areas urgently requiring data, which would benefit regional water resource management and hydrological disaster prevention. Full article

The Integrated Multisatellite Retrievals for GPM (IMERG) is designed to derive precipitation by merging data from all the passive microwave (PMW) and infrared (IR) sensors. While the input source errors originating from the PMW and IR sensors are important, their structure, characteristics, and algorithm improvement remain unclear. Our study utilized a four-component error decomposition (4CED) method and a systematic and random error decomposition method to evaluate the detectability of IMERG dataset and identify the precipitation errors based on the multi-sensors. The 30 min data from 30 precipitation stations in the Tunxi Watershed were used to evaluate the IMERG data from 2018 to 2020. The input source includes five types of PMW sensors and IR instruments. The results show that the sample ratio for IR (Morph, IR + Morph, and IR only) is much higher than that for PMW (AMSR2, SSMIS, GMI, MHS, and ATMS), with a ratio of 72.8% for IR sources and a ratio of 27.2% for PMW sources. The high false ratio of the IR sensor leads to poor detectability performance of the false alarm ratio (FAR, 0.5854), critical success index (CSI, 0.3014), and Brier score (BS, 0.1126). As for the 4CED, Morph and Morph + IR have a large magnitude of high total bias (TB), hit overestimate bias (HOB), hit underestimate bias (HUB), false bias (FB), and miss bias (MB), which is related to the prediction ability and sample size. In addition, systematic error is the prominent component for AMSR2, SSMIS, GMI, and Morph + IR, indicating some inherent error (retrieval algorithm) that needs to be removed. These findings can support improving the retrieval algorithm and reducing errors in the IMERG dataset. Full article

Natural environmental factors and human activity intensity factors, the two main factors that affect the spatial and temporal distribution of PM_2.5 concentration near the surface, have different mechanisms of action on PM_2.5 concentration. In this paper, a GTWR-XGBoost two-stage sequential hybrid model is proposed aiming at detecting the expression of spatiotemporal heterogeneity in the traditional machine learning retrieval model of PM_2.5 concentration and the difficulty of expressing the complex nonlinear relationship in the statistical regression model. In the first stage, the natural environmental factors are used to predict PM_2.5 concentration with spatiotemporal characteristics by collinearity diagnosis method and Geographically and Temporally Weighted Regression method (GTWR). In the second stage, the simulation results in the first stage and the natural factors eliminated through LUR stepwise regression in the first stage are into the XGBoost model together with the human activity intensity factors in the buffer zone with the best correlation coefficient of PM_2.5, and finally the temporal and spatial distribution of PM_2.5 concentration. Taking the Chengdu Chongqing Economic Circle as an example, the proposed model is used to retrieve PM_2.5 concentration and compared with the single GTWR, XGBoost, and coupling model published recently. The experimental results show that the R², RMSE, and MAE of the GTWR-XGBoost two-stage model cross-validation are 0.92, 5.44 ug·m⁻³, and 4.12 ug·m⁻³, respectively. Compared with the above single models, R² increased by 0.01 and 0.12, and MAE decreased by more than 0.11 and 3.1, respectively. Compared with the coupling model published recently, R² is increased by 0.02, and MAE is reduced by more than 0.4. In addition, the PM_2.5 concentration in Chengdu Chongqing showed obvious seasonal temporal and spatial changes, and the influence ratios of natural environmental factors and human activity intensity activities factors on PM_2.5 were 0.66 and 0.34. The results show that the GTWR-XGBoost two-stage Model can not only describe the heterogeneity and objectively reflect the complex nonlinear relationship between the phenomenon and the influencing factors, but also enhance the interpretability of the phenomenon when simulating the spatiotemporal distribution characteristics of PM_2.5 concentration. Full article

Satellite-based and reanalysis precipitation products have experienced increasing popularity in agricultural, hydrological and meteorological applications, but their accuracy is still uncertain in different areas. In this study, six frequently used high-resolution daily precipitation products, including Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS), Global Satellite Mapping of Precipitation (GSMaP), Integrated Multi-satellitE Retrievals for Global Precipitation Measurement (IMERG), Multi-Source Weighted-Ensemble Precipitation (MSWEP), Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks-Cloud Classification System-Climate Data Record (PERSIANN-CCS-CDR) and European Center for Medium-Range Weather Forecasts Reanalysis V5-Land (ERA5-Land), were comprehensively evaluated and compared in nine regions of mainland China between 2015 and 2019. The results reveal that, in general, GSMaP is the best precipitation product in different agricultural regions, especially based on the Pearson correlation coefficient (CC) and critical success index (CSI). ERA5-Land and MSWEP tend to have the highest probability of detection (POD) values, and MSWEP tends to have the smallest relative root mean squared error (RRMSE) values. GSMaP performs better at almost all precipitation levels and in most agricultural regions in each season, while MSWEP has the best performance for capturing the time series of mean daily precipitation. In addition, all precipitation products perform better in summer and worse in winter, and they are more accurate in the eastern region. The findings of this study will contribute to understanding the uncertainties of precipitation products, improving product quality and guiding product selection. Full article

Located in the interior of Eurasia, the Mongolian Plateau (MP) is extremely sensitive to global warming and become a critical area for studying precipitation patterns. Based on the monthly data of 135 meteorological stations during 1976–2017, we analyze the spatiotemporal change in precipitation and discuss its response to atmospheric circulation. The results show that: (1) Precipitation shows increasing trends in spring, autumn, and winter, but a decreasing trend at a rate of 5.3 mm/decade in summer. The annual precipitation also shows an overall slight decreasing trend. (2) The spatial distribution is uneven, the annual precipitation in the northern Great Khingan Mountains is more, but it gradually decreases at the rate of 10–30 mm/decade, showing a trend of “wet gets dry”; while there is less in the southwest Gobi Desert region, but it gradually increases with the rate of 10–20 mm/decade, showing a trend of “dry gets wet”. (3) Over decades, the East Asian summer monsoon (EASM) and westerly circulation show a seesaw change in MP. Affected by the weakening of the EASM, the area of arid regions has gradually expanded. The results also demonstrate that the EASM has a higher impact on the annual precipitation change pattern, particularly in the southeastern MP. The conclusion indicated that the variation in the position and orientation between EASM and the westerly circulation may be an explanation for the spatiotemporal precipitation pattern, providing a new viewpoint to the question of circulation mechanisms behind climate change in MP in recent 40 years. Full article

Long-term and high-resolution reanalysis precipitation datasets provide important support for research on climate change, hydrological forecasting, etc. The comprehensive evaluation of the error performances of the newly released ERA5-Land and CRA40-Land reanalysis precipitation datasets over the Yongding River Basin in North China was based on the two error decomposition schemes, namely, decomposition of the total mean square error into systematic and random errors and decomposition of the total precipitation bias into hit bias, missed precipitation, and false precipitation. Then, the error features of the two datasets and precipitation intensity and terrain effects against error features were analyzed in this study. The results indicated the following: (1) Based on the decomposition approach of systematic and random errors, the total error of ERA5-Land is generally greater than that of CRA40-Land. Additionally, the proportion of random errors was higher in summer and over mountainous areas, specifically, the ERA5-Land accounts for more than 75%, while the other was less than 70%; (2) Considering the decomposition method of hit, missed, and false bias, the total precipitation bias of ERA5-Land and CRA40-Land was consistent with the hit bias. The magnitude of missed precipitation and false precipitation was less than the hit bias. (3) When the precipitation intensity is less than 38 mm/d, the random errors of ERA5-Land and CRA40-Land are larger than the systematic error. The relationship between precipitation intensity and hit, missed, and false precipitation is complicated, for the hit bias of ERA5-L is always smaller than that of CRA40-L, and the missed precipitation and false precipitation are larger than those ofCRA40-L when the precipitation is small. The error of ERA5-Land and CRA40-Land was significantly correlated with elevation. A comprehensive understanding of the error features of the two reanalysis precipitation datasets is valuable for error correction and the construction of a multi-source fusion model with gauge-based and satellite-based precipitation datasets. Full article

Rainfall data have a profound significance for meteorology, climatology, hydrology, and environmental sciences. However, existing rainfall observation methods (including ground-based rain gauges and radar-/satellite-based remote sensing) are not efficient in terms of spatiotemporal resolution and cannot meet the needs of high-resolution application scenarios (urban waterlogging, emergency rescue, etc.). Widespread surveillance cameras have been regarded as alternative rain gauges in existing studies. Surveillance audio, through exploiting their nonstop use to record rainfall acoustic signals, should be considered a type of data source to obtain high-resolution and all-weather data. In this study, a method named parallel neural network based on attention mechanisms and multi-scale fusion (PNNAMMS) is proposed for automatically classifying rainfall levels by surveillance audio. The proposed model employs a parallel dual-channel network with spatial channel extracting the frequency domain correlation, and temporal channel capturing the time-domain continuity of the rainfall sound. Additionally, attention mechanisms are used on the two channels to obtain significant spatiotemporal elements. A multi-scale fusion method was adopted to fuse different scale features in the spatial channel for more robust performance in complex surveillance scenarios. In experiments showed that our method achieved an estimation accuracy of 84.64% for rainfall levels and outperformed previously proposed methods. Full article

This study aims to provide a brief overview of four regional climate model (RCM) estimations for (Daily, Monthly, Seasonal, and Annual) averaged precipitation over the Wadi Chemora Basin in northeastern Algeria for the historical period (1970–2005) and future forecasts (2006–2100). Data from seven ground stations were compared to data from four RCMs: RCA4 driven by ICHEC-EC-EARTH and NOAA-GFDL-GFDL-ESM2M from MENA-CORDEX domain with intermediate resolution (25 km, 0.22°) and ALADIN and RegCM4 from MED-CORDEX domain with high resolution (12 km, 0.11°). In most time steps (Annual, Seasonal, Monthly, and Daily), the raw RCMs overestimated precipitation, but their performance improved significantly after applying gamma quantile mapping (GQM) bias correction method. The bias-corrected projections indicate decreases of seasonal rainfall for the near future (2010–2039), mid-century (2040–2069), and end of century (2070–2100) periods. Overall decreases in all seasons lead to the projected decrease in annual rainfall of an average of 66% by the end of the 21st century. Full article

Precipitation data from ground-based observatories in the Dongting Lake basin are often missing, resulting in large errors in surface precipitation data obtained by interpolation, which affects the accuracy of hydro-meteorological studies. Integrated Multisatellite Retrievals for Global Precipitation Measurement (IMERG) is the main high-resolution precipitation product, which is available to supplement measured missing data. To evaluate the applicability of this product in the Dongting Lake basin at multiple spatial and temporal scales, this paper analyzes daily, monthly, seasonal, annual, and extreme precipitation events of the three latest IMERG precipitation products (IPPs) (IMERG-F, IMERG-E, and IMERG-L) using eight statistical evaluation metrics. We find that the spatial and temporal performance of IMERG precipitation products varies over different time scales and topographic conditions. However, all three metrics (CC, RMSE, and RB) of the IMERG-F precipitation products outperform the IMERG-E and IMERG-L precipitation products for the same period. In the comparison of IMERG and TRMM (Tropical Rainfall Measuring Mission) precipitation products on monthly and seasonal scales, IMERG-F performed the best. IPPs can capture precipitation more accurately on seasonal scales and perform better in winter, indicating good detection of trace precipitation. Both high and low altitudes are not favorable for the satellite detection of extreme precipitation in both general and extreme precipitation events. Overall, the accuracy of IMERG-F with correction delay is slightly better than that of IMERG-E and IMERG-L without correction under near-real-time conditions, which is applicable in the Dongting Lake basin. However, the correction process also exacerbates overestimation of the precipitation extent. Full article

Rain type classification into convective and stratiform is an essential step required to improve quantitative precipitation estimations by remote sensing instruments. Previous studies with Micro Rain Radar (MRR) measurements and subjective rules have been performed to classify rain events. However, automating this process by using machine learning (ML) models provides the advantages of fast and reliable classification with the possibility to classify rain minute by minute. A total of 20,979 min of rain data measured by an MRR at Das in northeast Spain were used to build seven types of ML models for stratiform and convective rain type classification. The proposed classification models use a set of 22 parameters that summarize the reflectivity, the Doppler velocity, and the spectral width (SW) above and below the so-called separation level (SL). This level is defined as the level with the highest increase in Doppler velocity and corresponds with the bright band in stratiform rain. A pre-classification of the rain type for each minute based on the rain microstructure provided by the collocated disdrometer was performed. Our results indicate that complex ML models, particularly tree-based ensembles such as xgboost and random forest which capture the interactions of different features, perform better than simpler models. Applying methods from the field of interpretable ML, we identified reflectivity at the lowest layer and the average spectral width in the layers below SL as the most important features. High reflectivity and low SW values indicate a higher probability of convective rain. Full article

The quality of satellite-based precipitation products including TMPA 3B42, IMERG-early, IMERG-final, and CMORPH-CRT, is evaluated by comparing with gauge observations in Hunan province of China between 2017 and 2019. By using the outputs of the Dominant River Routing Integrated with VIC Environment (DRIVE) model, the hydrological applications of gauge- and satellite-based precipitation products are analyzed by comparing them with streamflow observations. Furthermore, we conduct a case study considering Typhoon Bailu. It is found that IMERG-final can produce better results compared to the other three satellite-based products against gauge-based precipitation. In terms of discharge simulations, the gauge-based precipitation provides the most accurate results, followed by IMERG-final. During Typhoon Bailu, the peak of the mean gauge-based precipitation in the rainfall center (24.5°N–26°N, 111°E–114°E) occurred on 25 August 2019, whereas the daily streamflow reached its peak one day later, suggesting the lagged impact of precipitation on streamflow. From the Taylor diagram, the gauge-based precipitation is the most accurate for estimating the streamflow during Typhoon Bailu, followed by IMERG-final, IMERG-early, TMPA 3B42, and CMORPH-CRT, respectively. Overall, gauge-based precipitation has the best performance in terms of hydrological application, whereas IMERG-final performs the best among four satellite-based precipitation products. Full article

Reliable near-real-time precipitation estimation is crucial for scientific research and resistance to natural disasters such as floods. Compared with ground-based precipitation measurements, satellite-based precipitation measurements have great advantages, but precipitation estimation based on satellite is still a challenging issue. In this paper, we propose a deep learning model named Attention-Unet for precipitation estimation. The model utilizes the high temporal, spatial and spectral resolution data of the FY4A satellite to improve the accuracy of precipitation estimation. To evaluate the effectiveness of the proposed model, we compare it with operational near-real-time satellite-based precipitation products and deep learning models which proved to be effective in precipitation estimation. We use classification metrics such as Probability of detection (POD), False Alarm Ratio (FAR), Critical success index (CSI), and regression metrics including Root Mean Square Error (RMSE) and Pearson correlation coefficient (CC) to evaluate the performance of precipitation identification and precipitation amounts estimation, respectively. Furthermore, we select an extreme precipitation event to validate the generalization ability of our proposed model. Statistics and visualizations of the experimental results show the proposed model has better performance than operational precipitation products and baseline deep learning models in both precipitation identification and precipitation amounts estimation. Therefore, the proposed model has the potential to serve as a more accurate and reliable satellite-based precipitation estimation product. This study suggests that applying an appropriate deep learning algorithm may provide an opportunity to improve the quality of satellite-based precipitation products. Full article

Various flood resilience enhancement measures have been proposed to deal with the growing problem of urban flooding. However, there is a lack of evaluation about the applicability of these measures at a community scale. This paper investigates the effects of two types of flood resilience enhancement measures: engineering measures and adaptive measures, in order to explore their effectiveness in different flood-prone communities. A community-scale oriented flood resilience assessment method is used to assess the impact of different types of measures. A case study is applied in three communities that suffer from waterlogging problems in Jingdezhen city, China. Results show that there are spatial differences of flood resilience in three flood-prone communities. Future scenarios present a poorer performance in flood resilience compared to current scenarios due to the effects of urbanization and human activities. Engineering measures are suitable for the old communities with high-density residential areas when sitting alongside the river, for example the communities of Fuliang and Zhushan. On the other hand, adaptive measures exhibit more efficiency in improving flood resilience in all communities, especially effective for the new city town Changjiang where engineering measures are nearly saturated. The findings can help local governments develop appropriate flood resilience enhancement strategies for different types of communities. Full article

On the basis of hourly rain-gauge data from 735 stations over the Qilian Mountains in Northwest China, the rainfall diurnal variation represented in ERA5 reanalysis data from the European Centre for Medium-Range Weather Forecasts (ECMWF) was evaluated from May to October during 2012–2019. Results show that rainfall with intensities below 4 mm h⁻¹ was mostly overestimated, while intensities above 4 mm h⁻¹ were underestimated in ERA5. The most severe overestimation of weak precipitation occurs in the late afternoon, while heavy precipitation is mostly underestimated at night. Deviation in both heavy and weak precipitation is more evident in mountainous areas. The diurnal peak was reasonably reproduced for the rainfall events with durations shorter than 4 h, while the peak hour of events with longer duration showed evident bias. The positive (negative) deviations of short (long) duration rainfall events mainly appear in the late afternoon (night). Around the Qilian Mountains, where deviation is pronounced, the bias of afternoon short-duration events is influenced by higher-frequency precipitation, while the bias of long-duration events is related to the lower frequency of precipitation at night. In terms of the spatial distribution of precipitation with varied elevation, ERA5 fails to represent variation in weak and heavy precipitation with increasing elevation, which may be related to the deviation of surface-specific humidity in reanalysis. The results of this study imply the uncertainty of rainfall products by ERA5 over regions with complex topographic effects and provide metrics to evaluate rainfall products or forecasts over complex terrain area. Full article

The spatial and temporal characteristics of precipitation and floods during typhoons in Guangdong province were examined by using TRMM TMPA 3B42 precipitation data and the Dominant River Routing Integrated with VIC Environment (DRIVE) model outputs for the period 1998–2019. The evaluations based on gauge-measured and model-simulated streamflow show the reliability of the DRIVE model. The typhoon tracks are divided into five categories for those that landed on or influenced Guangdong province. Generally, the spatial distribution of precipitation and floods differ for different typhoon tracks. Precipitation has a similar spatial distribution to flood duration (FD) but is substantially different from flood intensity (FI). The average precipitation over Guangdong province usually reaches its peak at the landing time of typhoons, while the average FD and FI reach their peaks several hours later than precipitation peak. The lagged correlations between precipitation and FD/FI are hence always higher than their simultaneous correlations. Full article

Gridded precipitation data with a high spatiotemporal resolution are of great importance for studies in hydrology, meteorology, and agronomy. Observational data from meteorological stations cannot accurately reflect the spatiotemporal distribution and variations of precipitation over a large area. Meanwhile, radar-derived precipitation data are restricted by low accuracy in areas of complex terrain and satellite-based precipitation data by low spatial resolution. Therefore, hourly precipitation models were employed to merge data from meteorological stations, Radar, and satellites; the models used five machine learning algorithms (XGBoost, gradient boosting decision tree, random forests (RF), LightGBM, and multiple linear regression (MLR)), as well as the CoKriging method. In the north of Guangdong Province, data of four heavy rainfall events in 2018 were processed with geographic data to obtain merged hourly precipitation data. The CoKriging method secured the best prediction of spatial distribution of accumulated precipitation, followed by the tree-based machine learning (ML) algorithms, and significantly, the prediction of MLR deviated from the actual pattern. All machine learning methods showed poor performances for timepoints with little precipitation during the heavy rainfall events. The tree-based ML method showed poor performance at some timepoints when precipitation was over-related to latitude, longitude, and distance from the coast. Full article

This article describes the development of a machine learning (ML)-based algorithm for snowfall retrieval (Snow retrievaL ALgorithm fOr gpM–Cross Track, SLALOM-CT), exploiting ATMS radiometer measurements and using the CloudSat CPR snowfall products as references. During a preliminary analysis, different ML techniques (tree-based algorithms, shallow and convolutional neural networks—NNs) were intercompared. A large dataset (three years) of coincident observations from CPR and ATMS was used for training and testing the different techniques. The SLALOM-CT algorithm is based on four independent modules for the detection of snowfall and supercooled droplets, and for the estimation of snow water path and snowfall rate. Each module was designed by choosing the best-performing ML approach through model selection and optimization. While a convolutional NN was the most accurate for the snowfall detection module, a shallow NN was selected for all other modules. SLALOM-CT showed a high degree of consistency with CPR. Moreover, the results were almost independent of the background surface categorization and the observation angle. The reliability of the SLALOM-CT estimates was also highlighted by the good results obtained from a direct comparison with a reference algorithm (GPROF). Full article

The acquisition of the precise spatial distribution of precipitation is of great importance and necessity in many fields, such as hydrology, meteorology and ecological environments. However, in the arid and semiarid regions of Northwest China, especially over mountainous areas such as the Heihe River basin (HRB), the scarcity and uneven distribution of rainfall stations have created certain challenges in gathering information that accurately describes the spatial distribution of precipitation for use in applications. In this study, the accuracy of precipitation estimates from eight high-resolution gridded precipitation products (CMORPHv1-CRT, CRU TSv.4.05, ERA5, GSMaP_NRT, IMERG V06B-Final, MSWEPv2.0, PERSIANN-CDR and TRMM 3B42v7) are comprehensively evaluated by referring to the precipitation observations from 23 stations over the HRB using six indices (root mean square error, standard deviation, Pearson correlation coefficient, relative deviation, mean error and Kling–Gupta efficiency) from different spatial and temporal scales. The results show that at an annual scale, MSWEP has the highest accuracy over the entire basin, while PERSIANN, CRU and ERA5 show the most accurate results in the upper, middle and lower reaches of the HRB, respectively. At a seasonal scale, the performance of IMERG, CRU and ERA5 is superior to that of the other products in all seasons in the upper, middle and lower reaches, respectively. Over the entire HRB, PERSIANN displays the smallest deviation in all seasons except for spring. TRMM shows the highest accuracy in spring and autumn, while MSWEP and CRU show the highest accuracy in summer and winter, respectively. At a monthly scale, TRMM is superior to the other products, with a relatively stronger correlation almost every month, while GSMaP is inferior to the other products. Moreover, MSWEP and PERSIANN perform relatively best, with favorable statistical results around almost every station, while GSMaP shows the worse performance. In addition, ERA5 tends to overestimate higher values, while GSMaP tends to overestimate lower values over the entire basin. Moreover, the overestimation of ERA5 tends to appear in the upper reach area, while that of GSMaP tends to appear in the lower reach area. Only CRU and PERSIANN yield underestimations of precipitation, with the strongest tendency appearing in the upper reach area. The results of this study display some findings on the uncertainties of several frequently used precipitation datasets in the high mountains and poorly gauged regions in the HRB and will be helpful to researchers in various fields who need high-resolution gridded precipitation datasets over the HRB, as well as to data producers who want to improve their products. Full article

Studies of snow virga precipitation are rare. In this study, we investigated data from a vertically pointing Doppler Micro Rain Radar (MRR) in Bern, Switzerland, from 2008 to 2013 for snow virga precipitation events. The MRR data were reprocessed using the radar data processing algorithm of Garcia-Benardi et al., which allows the reliable determination of the snow virga precipitation rate. We focus on a long-lasting snow virga event from 17 March 2013, supported by atmospheric reanalysis data and atmospheric back trajectories. The snow virga was associated with a wind shear carrying moist air and snow precipitation in the upper air layers and dry air in the lower air layers. The lowest altitudes reached by the precipitation varied between 300 m and 1500 m above the ground over the course of the event. The duration of the snow virga was 22 h. In disagreement with the MRR observations, ERA5 reanalysis indicated drizzle at the ground over a time segment of 4 h during the snow virga event. Full article

Precipitation nowcasting has been gaining importance in the operational weather forecast, being essential for economic and social development. Conventional methods of precipitation nowcasting are mainly focused on the task of radar echo extrapolation. In recent years, deep learning methods have been used in this task. Nevertheless, raising the accuracy and extending the lead time of prediction remains as a challenging problem. To address the problem, we proposed a Multi-Scale Criss-Cross Attention Context Sensing Long Short-Term Memory (MCCS-LSTM). In this model, Context Sensing framework (CS framework) focuses on contextual correlations, and Multi-Scale Spatiotemporal block (MS block) with criss-cross attention is designed to extract multi-scale spatiotemporal feature and full-image dependency. To validate the effectiveness of our model, we conduct experiments on CIKM AnalytiCup 2017 data sets and Guangdong Province of China radar data sets. By comparing with existing deep learning models, the results demonstrate that the MCCS-LSTM has the best prediction performance, especially for predicting accuracy with longer lead times. Full article