Downscaling Seasonal Precipitation Forecasts over East Africa with Deep Convolutional Neural Networks

This study assesses the suitability of convolutional neural networks(CNNs) for downscaling precipitation over East Africa in the context of seasonal forecasting. To achieve this, we design a set of experiments that compare different CNN configurations and deployed the best-performing architecture to downscale one-month lead seasonal forecasts of June–July–August–September(JJAS) precipitation from the Nanjing University of Information Science and Technology Climate Forecast System version 1.0(NUIST-CFS1.0) for 1982–2020. We also perform hyper-parameter optimization and introduce predictors over a larger area to include information about the main large-scale circulations that drive precipitation over the East Africa region, which improves the downscaling results. Finally, we validate the raw model and downscaled forecasts in terms of both deterministic and probabilistic verification metrics, as well as their ability to reproduce the observed precipitation extreme and spell indicator indices. The results show that the CNN-based downscaling consistently improves the raw model forecasts, with lower bias and more accurate representations of the observed mean and extreme precipitation spatial patterns. Besides, CNN-based downscaling yields a much more accurate forecast of extreme and spell indicators and reduces the significant relative biases exhibited by the raw model predictions. Moreover, our results show that CNN-based downscaling yields better skill scores than the raw model forecasts over most portions of East Africa. The results demonstrate the potential usefulness of CNN in downscaling seasonal precipitation predictions over East Africa,particularly in providing improved forecast products which are essential for end users.

Improving the Short-Range Precipitation Forecast of Numerical Weather Prediction through a Deep Learning-Based Mask Approach

Due to various technical issues,existing numerical weather prediction(NWP)models often perform poorly at forecasting rainfall in the first several hours.To correct the bias of an NWP model and improve the accuracy of short-range precipitation forecasting,we propose a deep learning-based approach called UNet Mask,which combines NWP forecasts with the output of a convolutional neural network called UNet.The UNet Mask involves training the UNet on historical data from the NWP model and gridded rainfall observations for 6-hour precipitation forecasting.The overlap of the UNet output and the NWP forecasts at the same rainfall threshold yields a mask.The UNet Mask blends the UNet output and the NWP forecasts by taking the maximum between them and passing through the mask,which provides the corrected 6-hour rainfall forecasts.We evaluated UNet Mask on a test set and in real-time verification.The results showed that UNet Mask outperforms the NWP model in 6-hour precipitation prediction by reducing the FAR and improving CSI scores.Sensitivity tests also showed that different small rainfall thresholds applied to the UNet and the NWP model have different effects on UNet Mask's forecast performance.This study shows that UNet Mask is a promising approach for improving rainfall forecasting of NWP models.

Synoptic Verification of Precipitation Forecast of Three NWP Models from May to August of 2008 in Liaoning Province

In order to evaluate the precipitation forecast performance of mesoscale numerical model in Northeast China,mesoscale model in Liaoning Province and T213 model,and improve the ability to use their forecast products for forecasters,the synoptic verifications of their 12 h accumulated precipitation forecasts of 3 numerical modes from May to August in 2008 were made on the basis of different systems impacting weather in Liaoning Province.The time limitations were 24,36,48 and 60 h.The verified contents included 6 aspects such as intensity and position of precipitation center,intensity,location,scope and moving velocity of precipitation main body.The results showed that the three models had good forecasting capability for precipitation in Liaoning Province,but the cupacity of each model was obviously different.

Quantitative Precipitation Forecast Experiment Based on Basic NWP Variables Using Deep Learning

The quantitative precipitation forecast(QPF)performance by numerical weather prediction(NWP)methods depends fundamentally on the adopted physical parameterization schemes(PS).However,due to the complexity of the physical mechanisms of precipitation processes,the uncertainties of PSs result in a lower QPF performance than their prediction of the basic meteorological variables such as air temperature,wind,geopotential height,and humidity.This study proposes a deep learning model named QPFNet,which uses basic meteorological variables in the ERA5 dataset by fitting a non-linear mapping relationship between the basic variables and precipitation.Basic variables forecasted by the highest-resolution model(HRES)of the European Centre for Medium-Range Weather Forecasts(ECMWF)were fed into QPFNet to forecast precipitation.Evaluation results show that QPFNet achieved better QPF performance than ECMWF HRES itself.The threat score for 3-h accumulated precipitation with depths of 0.1,3,10,and 20 mm increased by 19.7%,15.2%,43.2%,and 87.1%,respectively,indicating the proposed performance QPFNet improved with increasing levels of precipitation.The sensitivities of these meteorological variables for QPF in different pressure layers were analyzed based on the output of the QPFNet,and its performance limitations are also discussed.Using DL to extract features from basic meteorological variables can provide an important reference for QPF,and avoid some uncertainties of PSs.

A Short-Range Quantitative Precipitation Forecast Algorithm Using Back-Propagation Neural Network Approach

A back-propagation neural network （BPNN） was used to establish relationships between the shortrange （0-3-h） rainfall and the predictors ranging from extrapolative forecasts of radar reflectivity, satelliteestimated cloud-top temperature, lightning strike rates, and Nested Grid Model （NGM） outputs. Quan- titative precipitation forecasts （QPF） and the probabilities of categorical precipitation were obtained. Results of the BPNN algorithm were compared to the results obtained from the multiple linear regression algorithm for an independent dataset from the 1999 warm season over the continental United States. A sample forecast was made over the southeastern United States. Results showed that the BPNN categorical rainfall forecasts agreed well with Stage Ⅲ observations in terms of the size and shape of the area of rainfall. The BPNN tended to over-forecast the spatial extent of heavier rainfall amounts, but the positioning of the areas with rainfall ≥25.4 mm was still generally accurate. It appeared that the BPNN and linear regression approaches produce forecasts of very similar quality, although in some respects BPNN slightly outperformed the regression.

A Case Study of Impact of FY-2C Satellite Data in Cloud Analysis to Improve Short-Range Precipitation Forecast

Chinese FengYun-2C(FY-2C) satellite data were combined into the Local Analysis and Prediction System(LAPS) model to obtain three-dimensional cloud parameters and rain content. These parameters analyzed by LAPS were used to initialize the Global/Regional Assimilation and Prediction System model(GRAPES) in China to predict precipitation in a rainstorm case in the country. Three prediction experiments were conducted and were used to investigate the impacts of FY-2C satellite data on cloud analysis of LAPS and on short range precipitation forecasts. In the first experiment, the initial cloud fields was zero value. In the second, the initial cloud fields were cloud liquid water, cloud ice, and rain content derived from LAPS without combining the satellite data. In the third experiment, the initial cloud fields were cloud liquid water, cloud ice, and rain content derived from LAPS including satellite data. The results indicated that the FY-2C satellite data combination in LAPS can show more realistic cloud distributions, and the model simulation for precipitation in 1–6 h had certain improvements over that when satellite data and complex cloud analysis were not applied.

Application of the Physical Quantity Field Evolution under Numerical Model in Precipitation Forecast of Yantai

[Objective] The research aimed to understand role of the forecast data about physical quantity field in precipitation forecast.[Method] By contrasting forecast and actual situation of the precipitation in Yantai during 2-3 July and 12-15 September,2011,advantages and disadvantages of the different numerical forecast models (Japan fax chart,European center,MM5,Grapes and T639) were analyzed.[Result] MICAPS system could provide live situation of the physical quantity field,but couldn't provide the future evolution situation.Japan fax chart,European center,MM5,Grapes and T639 could provide future evolution situation of the physical quantity field.[Conclusion] The contrasts and analyses on forecast situations of the physical quantity fields in many precipitation processes showed that evolutions of the vertical velocity,temperature dew point difference,relative humidity and wind field at the different heights could improve forecast accuracy of the precipitation in Yantai.

Test and Evaluation of ECMWF Model on Precipitation Forecast in Shaoyang Area

Using actual precipitation in Shaoyang of 2018-2020,precipitation forecast of ECMWF model was tested.The results showed that winter accuracy rate was the highest,followed by autumn,and summer accuracy rate was the lowest.24-h TS scoring results showed that the shorter the cumulative time,the lower the TS.Forecasters had a strong ability to predict summer rainstorms.

Research and Analysis on Refined Precipitation Forecast Method of Highway during Flood Season in Gansu Province

In this paper, T639 global spectral model numerical forecast products issued by numerical prediction center of China Meteorological Administration and precipitation data at 676 regional stations of Gansu during flood season were used firstly for interpretation. Then, through the establishment of multiple linear stepwise regression equation, daily precipitation forecast with the time interval of 6 h during 0 -72 h in May, 2013 in Gansu Province was obtained. By precipitation forecast verification analysis, it was found that the precipitation forecast in Hexi region and central Gansu Province had best effect; forecast in some places of east Hedong region had relatively good effect; part of the city or state forecast effect in the plateau slope zone was relatively lower. The refined precipitation forecast had certain reference and use value in the post-production Gansu Province highway forecasting during flood season.

Precipitation Forecast Test Based on MODE

Abstract By testing and analyzing BJ-RUC forecast of one precipitation process, MODE was introduced. MODE could give objective comparison from position of precipitation falling zone, shape and direction, and reflect intensity difference between forecast and actual situation, which comprehensively reflected precipitation forecast performance of the model, and was close to subjective judgment thinking of forecaster.

Application of the Probability Matching Correction Method in Precipitation Forecast

Based on the observation data of 24-hour cumulative precipitation from 92 ground meteorological observation stations in Jiangxi province from March to July during 2015-2016 and the high-resolution numerical forecast data of precipitation predicted within 24-72 h by the European Centre for Medium-Range Weather Forecasts（ ECMWF）,the Gamma function was used as the fitting function of probability distribution of cumulative precipitation to match the probability of predicted and observed precipitation. Moreover,the change of forecast score before and after the correction was tested. The results showed that the predicted values of heavy precipitation based on ECMWF model were smaller than the observed values,while the predicted values of light precipitation were larger than the observed values. The probability matching correction method could be used to effectively correct systematic errors of model forecast,and the correction effect of all grades of precipitation（ especially for rainstorm） was good.The shorter the period of validity was,the better the correction effect was. The correction method has a good application effect in the interpretation of model precipitation products,and can provide better security services for agricultural production.

Long-range precipitation forecasts using paleoclimate reconstructions in the western United States

Long-range precipitation forecasts are useful when managing water supplies.Oceanicatmospheric oscillations have been shown to influence precipitation.Due to a longer cycle of some of the oscillations,a short instrumental record is a limitation in using them for long-range precipitation forecasts.The influence of oscillations over precipitation is observable within paleoclimate reconstructions;however,there have been no attempts to utilize these reconstructions in precipitation forecasting.A data-driven model,KStar,is used for obtaining long-range precipitation forecasts by extending the period of record through the use of reconstructions of oscillations.KStar is a nearest neighbor algorithm with an entropy-based distance function.Oceanic-atmospheric oscillation reconstructions include the El Nino-Southern Oscillation(ENSO),the Pacific Decadal Oscillation(PDO),the North Atlantic Oscillation(NAO),and the Atlantic Multi-decadal Oscillation(AMO).Precipitation is forecasted for 20 climate divisions in the western United States.A 10-year moving average is applied to aid in the identification of oscillation phases.A lead time approach is used to simulate a one-year forecast,with a 10-fold cross-validation technique to test the models.Reconstructions are used from 1658-1899,while the observed record is used from 1900-2007.The model is evaluated using mean absolute error(MAE),root mean squared error(RMSE),RMSE-observations standard deviation ratio(RSR),Pearson's correlation coefficient(R),NashSutcliffe coefficient of efficiency(NSE),and linear error in probability space(LEPS) skill score(SK).The role of individual and coupled oscillations is evaluated by dropping oscillations in the model.The results indicate 'good' precipitation estimates using the KStar model.This modeling technique is expected to be useful for long-term water resources planning and management.

Seasonal Forecasts of Precipitation during the First Rainy Season in South China Based on NUIST-CFS1.0

Current dynamical models experience great difficulties providing reliable seasonal forecasts of regional/local rainfall in South China.This study evaluates seasonal forecast skill for precipitation in the first rainy season(FRS,i.e.,April–June)over South China from 1982 to 2020 based on the global real-time Climate Forecast System of Nanjing University of Information Science and Technology(NUIST-CFS1.0,previously known as SINTEX-F).The potential predictability and the practical forecast skill of NUIST-CFS1.0 for FRS precipitation remain low in general.But NUIST-CFS1.0 still performs better than the average of nine international models in terms of correlation coefficient skill in predicting the interannual precipitation anomaly and its related circulation index.NUIST-CFS1.0 captures the anomalous Philippines anticyclone,which transports moisture and heat northward to South China,favoring more precipitation in South China during the FRS.By examining the correlations between sea surface temperature(SST)and FRS precipitation and the Philippines anticyclone,we find that the model reasonably captures SST-associated precipitation and circulation anomalies,which partly explains the predictability of FRS precipitation.A dynamical downscaling model with 30-km resolution forced by the large-scale circulations of the NUIST-CFS1.0 predictions could improve forecasts of the climatological states and extreme precipitation events.Our results also reveal interesting interdecadal changes in the predictive skill for FRS precipitation in South China based on the NUIST-CFS1.0 hindcasts.These results help improve the understanding and forecasts for FRS precipitation in South China.

12 HOUR PRECIPITATION FORECASTS FOR NAIROBI,KENYA

Radiosonde profiles of temperature and dewpoint temperature from one station are used to forecast 12-h precipita- tion over Nairobi,Kenya.The forecast scheme is based on statistical regression modelling.Four predictors are derived from data to use in a prognostic equation to get 12-h precipitation forecast.Observed and predicted rainfall values are plotted on a graph against time.Forecast verification shows that the forecasts are positively correlated with observations.

Probabilistic Precipitation Forecasting Based on Ensemble Output Using Generalized Additive Models and Bayesian Model Averaging

A probabilistic precipitation forecasting model using generalized additive models （GAMs） and Bayesian model averaging （BMA） was proposed in this paper. GAMs were used to fit the spatial-temporal precipitation models to individual ensemble member forecasts. The distributions of the precipitation occurrence and the cumulative precipitation amount were represented simultaneously by a single Tweedie distribution. BMA was then used as a post-processing method to combine the individual models to form a more skillful probabilistic forecasting model. The mixing weights were estimated using the expectation-maximization algorithm. The residual diagnostics was used to examine if the fitted BMA forecasting model had fully captured the spatial and temporal variations of precipitation. The proposed method was applied to daily observations at the Yishusi River basin for July 2007 using the National Centers for Environmental Prediction ensemble forecasts. By applying scoring rules, the BMA forecasts were verified and showed better performances compared with the empirical probabilistic ensemble forecasts, particularly for extreme precipitation. Finally, possible improvements and a^plication of this method to the downscaling of climate change scenarios were discussed.

An application of the LTP＿DSEF model to heavy precipitation forecasts of landfalling tropical cyclones over China in 2018

Recently, a track-similarity-based Dynamical-Statistical Ensemble Forecast(LTP_DSEF) model has been developed in an attempt to predict heavy rainfall from Landfalling Tropical cyclones(LTCs). In this study, the LTP_DSEF model is applied to predicting heavy precipitation associated with 10 LTCs occurring over China in 2018. The best forecast scheme of the model with optimized parameters is obtained after testing 3452 different schemes for the 10 LTCs. Then, its performance is compared to that of three operational dynamical models. Results show that the LTP_DSEF model has advantages over the three dynamical models in predicting heavy precipitation accumulated after landfall, especially for rainfall amounts greater than 250 mm. The model also provides superior or slightly inferior heavy rainfall forecast performance for individual LTCs compared to the three dynamical models. In particular, the LTP_DSEF model can predict heavy rainfall with valuable threat scores associated with certain LTCs, which is not possible with the three dynamical models. Moreover, the model can reasonably capture the distribution of heavier accumulated rainfall, albeit with widespread coverage compared to observations. The preliminary results suggest that the LTP_DSEF model can provide useful forecast guidance for heavy accumulated rainfall of LTCs despite its limited variables included in the model.

Improving the Seasonal Forecast of Summer Precipitation in China Using a Dynamical-Statistical Approach

A dynamical-statistical post-processing approach is applied to seasonal precipitation forecasts in China during the summer.The data are ensemble-mean seasonal forecasts in summer (June August) from four atmospheric general circulation models (GCMs) in the second phase of the Canadian Historical Forecasting Project (HFP2) from 1969 to 2001.This dynamical-statistical approach is designed based on the relationship between the 500 geopotential height (Z500) forecast and the observed sea surface temperature (SST) to calibrate the precipitation forecasts.The results show that the post-processing can improve summer precipitation forecasts for many areas in China.Further examination shows that this post-processing approach is very effective in reducing the model-dependent part of the errors,which are associated with GCMs.The possible mechanisms behind the forecast's improvements are investigated.

BMA probability quantitative precipitation forecasting of land-falling typhoons in south-east China

The probability of quantitative precipitation forecast(PQPF)of three Bayesian Model Averaging(BMA)models based on three raw super ensemble prediction schemes(i.e.,A,B,and C)are established,which through calibration of their parameters using 1-3 day precipitation ensemble prediction systems(EPSs)from the China Meteorological Administration(CMA),the European Centre for Medium-Range Weather Forecasts(ECMWF)and the National Centers for Environmental Prediction(NCEP)and observation during land-falling of three typhoons in south-east China in 2013.The comparison of PQPF shows that the performance is better in the BMA than that in raw ensemble forecasts.On average,the mean absolute error(MAE)of 1 day lead time forecast is reduced by 12.4%,and its continuous ranked probability score(CRPS)of 1-3 day lead time forecast is reduced by 26.2%,respectively.Although the amount of precipitation prediction by the BMA tends to be underestimated,but in view of the perspective of probability prediction,the probability of covering the observed precipitation by the effective forecast ranges of the BMA are increased,which is of great significance for the early warning of torrential rain and secondary disasters induced by it.

An Objective Approach to Generating Multi-Physics Ensemble Precipitation Forecasts Based on the WRF Model

Selecting proper parameterization scheme combinations for a particular application is of great interest to the Weather Research and Forecasting(WRF)model users.This study aims to develop an objective method for identifying a set of scheme combinations to form a multi-physics ensemble suitable for short-range precipitation forecasting in the Greater Beijing area.The ensemble is created by using statistical techniques and some heuristics.An initial sample of 90 scheme combinations was first generated by using Latin hypercube sampling(LHS).Then,after several rounds of screening,a final ensemble of 40 combinations were chosen.The ensemble forecasts generated for both the training and verification cases using these combinations were evaluated based on several verification metrics,including threat score(TS),Brier score(BS),relative operating characteristics(ROC),and ranked probability score(RPS).The results show that TS of the final ensemble improved by 9%-33%over that of the initial ensemble.The reliability was improved for rain≤10 mm day^-1,but decreased slightly for rain>10 mm day^-1 due to insufficient samples.The resolution remained about the same.The final ensemble forecasts were better than that generated from randomly sampled scheme combinations.These results suggest that the proposed approach is an effective way to select a multi-physics ensemble for generating accurate and reliable forecasts.

ON THE SENSITIVITY OF PRECIPITATION FORECASTS TO THE MOIST PHYSICS AND THE HORIZONTAL RESOLUTION OF NUMERICAL MODEL

The impacts of the enhanced model's moist physics and horizontal resolution upon the QPFs (quantitative precipitation forecasts)are investigated by applying the HIRLAM(high resolution limited area model)to the summer heavy-rain cases in China.The performance of the control run, for which a 0.5°×0.5°grid spacing and a traditional“grid-box supersaturation removal+Kuo type convective paramerization”are used as the moist physics,is compared with that of the sensitivity runs with an enhanced model's moist physics(Sundqvist scheme)and an increased horizontal resolution(0.25°×0.25°),respectively.The results show: (1)The enhanced moist physics scheme(Sundqvist scheme),by introducing the cloud water content as an additional prognostic variable and taking into account briefly of the microphysics involved in the cloud-rain conversion,does bring improvements in the model's QPFs.Although the deteriorated QPFs also occur occasionally,the improvements are found in the majority of the cases,indicating the great potential for the improvement of QPFs by enhancing the model's moist physics. (2)By increasing the model's horizontal resolution from 0.5°×0.5°,which is already quite high compared with that of the conventional atmospheric soundings,to 0.25°×0.25°without the simultaneous enhancement in model physics and objective analysis,the improvements in QPFs are very limited.With higher resolution,although slight amelioration in locating the rainfall centers and in resolving some finer structures of precipitation pattern are made,the number of the mis- predicted fine structures in rainfall field increases with the enhanced model resolution as well.