Potential of support vector regression for prediction of monthly streamflow using endogenous property

In the recent past, a variety of statistical and other modelling approaches have been developed to capture the properties of hydrological time series for their reliable prediction. However, the extent of complexity hinders the applicability of such traditional models in many cases. Kernel‐based machine learning approaches have been found to be more popular due to their inherent advantages over traditional modelling techniques including artificial neural networks(ANNs ). In this paper, a kernel‐based learning approach is investigated for its suitability to capture the monthly variation of streamflow time series. Its performance is compared with that of the traditional approaches. Support vector machines (SVMs) are one such kernel‐based algorithm that has given promising results in hydrology and associated areas. In this paper, the application of SVMs to regression problems, known as support vector regression (SVR), is presented to predict the monthly streamflow of the Mahanadi River in the state of Orissa, India. The results obtained are compared against the results derived from the traditional Box–Jenkins approach. While the correlation coefficient between the observed and predicted streamflows was found to be 0·77 in case of SVR, the same for different auto‐regressive integrated moving average (ARIMA) models ranges between 0·67 and 0·69. The superiority of SVR as compared to traditional Box‐Jenkins approach is also explained through the feature space representation. Copyright © 2009 John Wiley & Sons, Ltd.     

Satellite-based technique for nowcasting of thunderstorms over Indian region

India experiences severe thunderstorms during the months, March–June. But these systems are not predicted well, mainly due to the absence of mesoscale observational network over Indian region and the expert system. As these are short lived systems, the nowcast is attempted worldwide based on satellite and radar observations. Due to inadequate radar network, satellite plays the dominant role for nowcast of these thunderstorms. In this study, a nowcast based algorithm ForTracc developed by Vila et al. (Weather Forecast 23:233–245, 2008) has been examined over the Indian region using Infrared Channel \((10.8~\upmu \hbox {m})\) of INSAT-3D for prediction of Mesoscale Convective Systems (MCS). In this technique, the current location and intensity in terms of Cloud Top Brightness Temperature (CTBT) of the MCS are extrapolated. The purpose of this study is to validate this satellite-based nowcasting technique for Convective Cloud Clusters that helps in optimum utilization of satellite data and improve the nowcasting. The model could predict reasonably the minimum CTBT of the convective cell with average absolute error (AAE) of \({<}7\hbox { K}\) for different lead periods (30–180 min). However, it was underestimated for all the lead periods of forecasts. The AAE in the forecasts of size of the cluster varies from about \(3\times 10^{4}\hbox { km}^{2}\) for 30-min forecast to \(7\times 10^{4}\hbox { km}^{2}\) for 120-min forecast. The mean absolute error in prediction of size is above 31–38% of actual size for different lead periods of forecasts from 30 to 180 min. There is over estimation in prediction of size for 30 and 60 min forecasts (17% and 2.6% of actual size of the cluster, respectively) and underestimation in 90 to 180-min forecasts (–2.4% to –28%). The direct position error (DPE) based on the location of minimum CTBT ranges from 70 to 144 km for 30–180-min forecast respectively.     

Evaluation of TRMM multi-satellite precipitation analysis (TMPA) against terrestrial measurement over a humid sub-tropical basin,India

To support the GPM mission which is homologous to its predecessor, the Tropical Rainfall Measuring Mission (TRMM), this study has been undertaken to evaluate the accuracy of Tropical Rainfall Measuring Mission multi-satellite precipitation analysis (TMPA) daily-accumulated precipitation products for 5 years (2008–2012) using the statistical methods and contingency table method. The analysis was performed on daily, monthly, seasonal and yearly basis. The TMPA precipitation estimates were also evaluated for each grid point i.e. 0.25° × 0.25° and for 18 rain gauge stations of the Betwa River basin, India. Results indicated that TMPA precipitation overestimates the daily and monthly precipitation in general, particularly for the middle sub-basin in the non-monsoon season. Furthermore, precision of TMPA precipitation estimates declines with the decrease of altitude at both grid and sub-basin scale. The study also revealed that TMPA precipitation estimates provide better accuracy in the upstream of the basin compared to downstream basin. Nevertheless, the detection capability of daily TMPA precipitation improves with increase in altitude for drizzle rain events. However, the detection capability decreases during non-monsoon and monsoon seasons when capturing moderate and heavy rain events, respectively. The veracity of TMPA precipitation estimates was improved during the rainy season than during the dry season at all scenarios investigated. The analyses suggest that there is a need for better precipitation estimation algorithm and extensive accuracy verification against terrestrial precipitation measurement to capture the different types of rain events more reliably over the sub-humid tropical regions of India.