Monthly rainfall hind-cast using machine learning algorithms for Coimbatore, Tamil Nadu

Due to current world climate change, the accuracy of predicting rainfall is critical. This paper presents an approach using four different machine learning algorithms, viz., Decision Tree Regression (DTR), Gradient Boosting (GB), Ada Boost (AB) and Random Forest Regression (RFR) techniques to improve the rainfall forecast performance. When historical events are entered into the model and get validated to realise how well the output suits the known results referred as Hind-cast. Historical monthly weather parameters over a period of 42 years (1976 to 2017) were collected from Agro Climate Research Centre, Tamil Nadu Agricultural University. The global climate driver’s viz., Southern Oscillation Index and Indian Ocean Dipole indices were retrieved from Bureau of Meteorology, Australia. K- means algorithm was employed for centroid identification (which select the rows with unique distinguished features) at 90 per cent of the original data for the period of 42 years by eliminating the redundancy nature of the datawhich were used as training set. The result indicated the supremacy and notable strength of RFR over the other algorithms in terms of performance with 89.2 per cent. The Co-efficient of Determination (R2) for the predicted and observed values was found to be 0.8 for the monthly rainfall from 2015 to 2017.

A novel risk score for disease control prediction of Chronic rhinosinusitis

Objectives: To assess the impact of risk factors on the disease control among CRS patients, following 1 year of functional endoscopic sinus surgery (FESS), and combining the risk factors to formulate a convenient, visualized prediction model. Design: A retrospective and nonconcurrent cohort study Setting and Participants: A total of 325 patients with Chronic rhinosinusitis (CRS) from June 2018 to July 2020 at the First Affiliated Hospital, the Third Affiliated Hospital, and the Seventh Affiliated Hospital of Sun Yat-sen University. Main Outcomes Measures: Outcomes were time to event measures: the disease control of CRS after surgery 1 year. The presence of nasal polyps, smoking habits, allergic rhinitis (AR), the ratio of tissue eosinophil (TER), and peripheral blood eosinophil count (PBEC)and asthma was assessed. The logistic regression models were used to conduct multivariate and univariate analyses. Asthma, TER, AR, PBEC were also included in the nomogram. The calibration curve and AUC (Area Under Curve) were used to evaluate the forecast performance of the model. Results: In univariate analyses, most of the covariates had significant associations with the endpoints, except for age, gender, and smoking. The nomogram showed the highest accuracy with an AUC of 0.760 (95% CI, 0.688-0.830) in the training cohort. Conclusions: In this cohort study that included the asthma, AR, TER, PBEC had significantly affected the disease control of CRS after surgery. The model provided relatively accurate prediction in the disease control of CRS after FESS and served as a visualized reference for daily diagnosis and treatment.

Comparing seasonal streamflow forecast systems for management of a fresh water reservoir in the Netherlands

For efficient management of the Dutch surface water reservoir Lake IJssel, (sub)seasonal forecasts of the water volumes going in and out of the reservoir are potentially of great interest. Here, streamflow forecasts were analyzed for the river Rhine at Lobith, which is partly routed through the river IJssel, the main influx into the reservoir. We analyzed multiple seasonal forecast data sets derived from EFAS, E-HYPE and HTESSEL, which differ in their underlying hydrological formulation, but are all forced with similar input from the ECMWF SEAS5 meteorological forecasts. We post-processed the streamflow forecasts using quantile matching (QM) and analyzed several forecast quality metrics. Forecast performance was assessed based on the available reforecast period, as well as on individual summer seasons. QM increased forecast skill for nearly all metrics evaluated. Particularly HTESSEL, a land surface scheme that is not optimized for hydrology, needed the largest correction. Averaged over the reforecast period, forecasts were skillful for the longest lead times in spring and early summer. For this period, E-HYPE showed the highest skill; Later in summer, however, skill deteriorated after 1–2 months. When investigating specific years with either low or high flow conditions, forecast skill increased with the extremity of the event. Although raw forecasts for both E-HYPE and EFAS were more skilful than HTESSEL, bias correction based on QM can significantly reduce the difference. In operational mode, the three forecast systems show comparable skill. In general, dry conditions can be forecasted with high success rates up to three months ahead, which is very promising for successful use of Rhine streamflow forecasts in downstream reservoir management.

High-resolution NWP forecast precipitation comparison over complex terrain of the Sierras de Córdoba during RELAMPAGO-CACTI.

Sierras de Córdoba (Argentina) is characterized by the occurrence of extreme precipitation events during the austral warm season. Heavy precipitation in the region has a large societal impact, causing flash floods. This motivates the forecast performance evaluation of 24-hour accumulated precipitation and vertical profiles of atmospheric variables from different numerical weather prediction (NWP) models with the final aim of helping water management in the region. The NWP models evaluated include the Global Forecast System (GFS) which parameterizes convection, and convection-permitting simulations of the Weather Research and Forecasting Model (WRF) configured by three institutions: University of Illinois at Urbana–Champaign (UIUC), Colorado State University (CSU) and National Meteorological Service of Argentina (SMN). These models were verified with daily accumulated precipitation data from rain gauges and soundings during the RELAMPAGO-CACTI field campaign. Generally all configurations of the higher-resolution WRFs outperformed the lower-resolution GFS based on multiple metrics. Among the convection-permitting WRF models, results varied with respect to rainfall threshold and forecast lead time, but the WRFUIUC mostly performed the best. However, elevation dependent biases existed among the models that may impact the use of the data for different applications. There is a dry (moist) bias in lower (upper) pressure levels which is most pronounced in the GFS. For Córdoba an overestimation of the northern flow forecasted by the NWP configurations at lower levels was encountered. These results show the importance of convection-permitting forecasts in this region, which should be complementary to the coarser-resolution global model forecasts to help various users and decision makers.

The Value of Meteorological Data in Optimizing the Pattern of Physical Load—A Forecast Model of Rowing Pacing Strategy

Mastering the information of arena environment is the premise for athletes to optimize their patterns of physical load. Therefore, improving the forecast accuracy of the arena conditions is an urgent task in competitive sports. This paper excavates the meteorological features that have great influence on outdoor events such as rowing and their influence on the pacing strategy. We selected the meteorological data of Tokyo from 1979 to 2020 to forecast the meteorology during the Tokyo 2021 Olympic Games, analyzed the athletes’ pacing choice under different temperatures, humidity and sports levels, and then recommend the best pacing strategy for rowing teams of China. The model proposed in this paper complements the absence of meteorological features in the arena environment assessment and provides an algorithm basis for improving the forecast performance of pacing strategies in outdoor sports.

Evaluation of Forecast Performance for Four Meteorological Models in Summer Over Northwestern China

At present, numerical models, which have been used for forecasting services in northwestern China, have not been extensively evaluated. We used national automatic ground station data from summer 2016 to test and assess the forecast performance of the high-resolution global European Centre for Medium-Range Weather Forecast (ECMWF) model, the mesoscale Northwestern Mesoscale Numerical Prediction System (NW-MNPS), the global China Meteorological Administration T639 model, and the mesoscale Global/Regional Assimilation and Prediction System (GRAPES) model over northwestern China. The root mean square error (RMSE) of the 2-m temperature forecast by ECMWF was the lowest, while that by T639 was the highest. The distribution of RMSE for each model forecast was similar to that of the difference between the modeled and observed terrain. The RMSE of the 10-m wind speed forecast was lower for the global ECMWF and T639 models and higher for the regional NW-MNPS and GRAPES models. The 24-h precipitation forecast was generally higher than observed for each model, with NW-MNPS having the highest score for light rain and heavy storm rain, ECMWF for medium and heavy rain, and T639 for storm rain. None of the models could forecast small-scale and high-intensity precipitation, but they could forecast large-scale precipitation. Overall, ECMWF had the best stability and smallest prediction errors, followed by NW-MNPS, T639, and GRAPES.

DSAEF_LTP Model Experiment to Forecast the Accumulated Precipitation of Landfalling Northward-Moving Typhoons in China

We designed two groups of experiments to test the forecast performance of the Dynamical-Statistical-Analog Ensemble Forecast (DSAEF_LTP) model for precipitation caused by landfalling northward-moving typhoons. The first group DSAEF_LTP-1 had the generalized initial value containing three factors (tropical cyclone track, landfall season and tropical cyclone intensity) while the second group DSAEF_LTP-2 added multiple choices of similarity regions. We selected 33 typhoons that brought about maximum daily precipitation ≥100 mm to the area north of the Yangtze River from 2004–2019. We used 22 tropical cyclones from 2004–2015 as training samples to identify the best scheme, which was then used to conduct independent sample forecasting experiments for 11 tropical cyclones from 2016–2019. The results were compared with those of four numerical models (ECMWF, GFS, GRAPES and SMS-WARMS). The simulation ability of the DSAEF_LTP model was significantly improved after adding the similarity regions. The TSsum (TS250 + TS100) for accumulated precipitation ≥250 and ≥100 mm increased from 0.1239 (0 + 0.1239) to 0.1883 (0.0526 + 0.1357). The forecast performance of the DSAEF_LTP for TS100 was 0.1355 for DSAEF_LTP-1 and 0.099 for DSAEF_LTP-2 . Both exceeded the scores for two of the operational Numerical Models, GRAPES (0.0798) and SMS-WARMS (0.0943). The DSAEF_LTP model can capture the distribution patterns of the observed precipitation in most cases. The forecasting performance was good over the southern coast of China but was limited in the north. The development of vortex identification technology for residual vortices and the introduction of new environmental factors into the generalized initial value are required to improve the DSAEF_LTP model.

Comparing the Forecast Performance of Advanced Statistical and Machine Learning Techniques Using Huge Big Data: Evidence from Monte Carlo Experiments

This research compares factor models based on principal component analysis (PCA) and partial least squares (PLS) with Autometrics, elastic smoothly clipped absolute deviation (E-SCAD), and minimax concave penalty (MCP) under different simulated schemes like multicollinearity, heteroscedasticity, and autocorrelation. The comparison is made with varying sample size and covariates. We found that in the presence of low and moderate multicollinearity, MCP often produces superior forecasts in contrast to small sample case, whereas E-SCAD remains better. In the case of high multicollinearity, the PLS-based factor model remained dominant, but asymptotically the prediction accuracy of E-SCAD significantly enhances compared to other methods. Under heteroscedasticity, MCP performs very well and most of the time beats the rival methods. In some circumstances under large samples, Autometrics provides a similar forecast as MCP. In the presence of low and moderate autocorrelation, MCP shows outstanding forecasting performance except for the small sample case, whereas E-SCAD produces a remarkable forecast. In the case of extreme autocorrelation, E-SCAD outperforms the rival techniques under both the small and medium samples, but further augmentation in sample size enables MCP forecast more accurate comparatively. To compare the predictive ability of all methods, we split the data into two halves (i.e., data over 1973–2007 as training data and data over 2008–2020 as testing data). Based on the root mean square error and mean absolute error, the PLS-based factor model outperforms the competitor models in terms of forecasting performance.