China 
Africa 
Explore chapters and articles related to this topic
Sustainability Analytics Applications
Published in Ram Ramanan, Introduction to Sustainability Analytics, 2018
The collected data is ingested by IBM’s machine learning technologies to determine the highest probability forecast. Errors in model forecasting often result from uncertainties in initial conditions. The tool leverages data assimilation to combine the range of different data sources, including surface monitoring data, weather data, emissions data, satellite data, and geographical data, to estimate the initial state of a model (i.e., initial condition) and to set the stage to provide a high-accuracy air quality outlook. Once the optimal initial conditions are set, the data enters the model blending process, where different physics and chemistry predictive models are combined. As each model performs best in different conditions (i.e., temperature, wind speed/strength, geographic and topographic factors, season of the year, etc.), IBM’s machine learning techniques play a key role in finding just the right blending of the different models to deliver the best forecast given the actual parameters being provided by the recorded data set at any given point of time. The system uses adaptive machine learning mechanisms to train those models and adaptively adjust the parameters for each model, and selects the optimized one with best performance for each specific situation, based on the combination of the historical and current data recorded.736
Simulation of sediment hyper concentration in the lower Yellow River using variational data assimilation method
Published in Silke Wieprecht, Stefan Haun, Karolin Weber, Markus Noack, Kristina Terheiden, River Sedimentation, 2016
R. Lai, M. Wang, M. Wang, H. Wang
Data assimilation is a method which blends observational information and a short term model forecast to get a higher accurate prediction (Talagrand 1997). Variational based data assimilation is one of the most popular data assimilation algorithms. Variational based algorithm implements the minimum of the distance between observation and prediction, while, at the same time, taking an explicit dynamic system as a constraint condition. The dynamic system can be described by partial differential equations which control the transport of sediment in a river. The obvious theoretical advantage of variational data assimilation is that it can provide exact consistency between the forecasts and the dynamics (Le Dimet et al., 1986). Recently, the data assimilation methods have received considerable attention in a wide area including ocean circulation, numerical weather prediction, hydrology, coastal area and atmosphere (e.g. Wunsch 1996, Kalnay 2003, Moradkhani et al. 2005, Lewis et al. 2006, Stroud et al. 2009, Thornhill et al. 2012, Zhang et al. 2012, Margvelashvili et al. 2013, Lai et al. 2013). However, cases that use variational data assimilation to improve the accuracy of simulated sediment concentration are rare.
Conclusions and Recommendations
Published in Maurizio Mazzoleni, Improving Flood Prediction Assimilating Uncertain Crowdsourced Data into Hydrological and Hydraulic Models, 2017
Catastrophic floods cause significant socio-economic losses. Therefore, accurate realtime forecasting of streamflow and water level is crucial for a proper evaluation of the flood risk and subsequent damages. A large number of hydrological and hydraulic models of varying complexity, have been proposed in the last few decades to accurately estimate streamflow and water level along the rivers. Nowadays, model updating techniques, in particular data assimilation methods, have been used to improve flood forecasts by integrating static ground observations, and in some cases also remote sensing observations, within hydrological and hydrodynamic models. In recent years, continued technological improvement has stimulated the spread of low-cost sensors that allow for employing crowdsourced data and obtaining observations of hydrological variables in a more distributed way than the classical static physical sensors. However, current hydrologic and hydraulic research typically considers assimilation of observations coming from traditional static sensors. One reason for this is that crowdsourced measurements have random arrival frequency and varying accuracy. This PhD research aims to develop and test the methods for assimilating CS observations having variable spatio-temporal coverage and provided by citizens by means of different low-cost sensors, and to demonstrate that such data can be useful for improving flood forecasts. The proposed methods have been successfully implemented in the AMICO EWS in the Bacchiglione catchment, an official case study of the WeSenseIt EU project which is funding this PhD research.
Statistical analysis of simulated oceanic dispersion of dissolved radionuclide hypothetically released from the Fukushima Dai-ichi Nuclear Power Plant using long-term oceanographic reanalysis data
Published in Journal of Nuclear Science and Technology, 2023
Tsubasa Ikenoue, Hideyuki Kawamura, Yuki Kamidaira
The multivariate ocean variational estimation (MOVE) system was developed as an ocean data assimilation system in JMA. MOVE/MRI.COM developed by the Meteorological Research Institute (MRI) of JMA, is based on the Meteorological Research Institute Community Ocean Model (MRI.COM) [22]. This study used oceanographic reanalysis data of the western North Pacific version (MOVE/MRI.COM-WNP) [23] of MOVE/MRI.COM. The domain of MOVE/MRI.COM-WNP extends from 14.85°N to 49.75°N in latitude and from 116.85°E to 159.75°E in longitude for a horizontal resolution of 0.1° and 54 vertical layers. Figure 1 shows the MOVE/MRI.COM-WNP domain and sea-bottom topography [24]. MOVE/MRI.COM-WNP provides daily averaged oceanographic data, including sea surface height, temperature, salinity, and horizontal current velocities. Data assimilation is the combination of numerical models and observed data to produce high accuracy data. Numerous satellite and in situ data, including advanced research and global observation floats, are assimilated in MOVE/MRI.COM-WNP using a three-dimensional variational adjoint method. Therefore, MOVE/MRI.COM-WNP can reproduce the oceanographic condition accurately in the model domain. For example, the simulated variations in the Kuroshio transports crossing the affiliated surveys of the Kuroshio off Cape Ashizuri line and the simulated mean volume transport crossing the pollution Nagasaki line agreed well with the observations [23].
A Review on Energy Forecasting Algorithms Crucial for Energy Industry Development and Policy Design
Published in Energy Sources, Part A: Recovery, Utilization, and Environmental Effects, 2021
Smart grids enable the flow of information and energy in both directions. Reliable confidence interims are more likely to participate in an offering procedure or bid process in the electricity markets when planning for sustaining power plants. Because the induced suspicions must be related to innovation, technology, or application, hybrid techniques or confidence intervals are essential. Much work remains to be done in the forecasting industry to provide accurate information for these networks and encourage integrating renewable energy systems into conventional grids. In India, there is an ever-increasing demand for electricity. In such circumstances, renewable energy can play a significant role in meeting electrical demand. With the access to enormous data sets and the addition of additional data sources, our future study path in the field of renewable energy forecasting has the potential to widen the types and quantity of data utilized in both statistical and physical modeling. Larger datasets will be used to power innovative statistical and machine learning methods, while novel data assimilation processes will make additional data sources available to NWP, improving forecast skills. Advanced forecasting abilities for multiple energy sources, including “behind-the-meter” forecasting and novel representations for prediction uncertainty, will be applied to manage power networks with substantial amounts of decentralized renewable energy generation.
Data assimilation for flow forecasting in urban drainage systems by updating a hydrodynamic model of Damhusåen Catchment, Copenhagen
Published in Urban Water Journal, 2020
Mukand Babel, Husnain Tansar, Ole Mark, Sutat Weesakul, Henrik Madsen
Data assimilation has been widely applied, in the last three decades, in the field of meteorology, hydrology, and oceanography. The Kalman Filter (KF) and its extended forms have been successfully applied in hydrological modeling by assimilating groundwater head (e.g. Rasmussen et al. 2015; Zhang et al. 2016), soil moisture (Fairbairn et al. 2015; Ridler et al. 2014), and river water level measurements (Schneider et al. 2018; Hartnack, Madsen, and Sørensen 2005). Data assimilation has also received significant attention and recognition in research due to the abundance of in-situ and remotely sensed observations, which are used to improve hydrological forecasts (Velzen et al. 2014; Liu et al. 2012; McMillan et al. 2013). Cañizares et al. (2001) applied an efficient filtering scheme based on a predefined, time-invariant Kalman gain (steady KF). By using this method, the update simulation would be a little more expensive than the normal model simulation. The steady Kalman gain was calculated based on the average of the time-varying Kalman gains using an off-line data assimilation simulation. Madsen and Skotner (2005) introduced a combined filtering and error forecasting procedure to update states in a real-time flood forecasting model. The states of the system were updated by distribution of the model errors at the measurement locations, which were based on predefined, time-invariant weighting functions. The parameters of the error forecast model were estimated based on observed model errors prior to time of forecast in order to update the initial conditions of the error forecast model for the forecast period.