Search Results

You are looking at 1 - 10 of 10 items for :

  • Data quality control x
  • The 1st NOAA Workshop on Leveraging AI in the Exploitation of Satellite Earth Observations & Numerical Weather Prediction x
  • All content x
Clear All
Ryan Lagerquist, Amy McGovern, Cameron R. Homeyer, David John Gagne II, and Travis Smith

. a. Data description MYRORSS 5 contains quality-controlled, merged data from all WSR-88D ( Crum and Alberty 1993 ) sites in the contiguous United States (CONUS). Each radar scans a different part of the atmosphere, and where multiple radars scan the same point, they generally have differing resolution and errors. Merging data from all radars allows the data to be represented on a common grid, and the merging algorithm includes quality-control measures that cannot be applied to single-radar data

Restricted access
Kyle A. Hilburn, Imme Ebert-Uphoff, and Steven D. Miller

5-min accumulation periods but found that this finer temporal granularity produced stratiform areas that flicker on and off from frame to frame. The lighting data units are given as groups per 5 min per kilometer squared. c. MRMS dataset The target dataset to which we are training is the quality-controlled composite reflectivity from the MRMS product ( Smith et al. 2016 ). The vertical coverage of MRMS as a function of location is given in Fig. 1 , which was created using the 3D reflectivity

Open access
Sid-Ahmed Boukabara, Vladimir Krasnopolsky, Jebb Q. Stewart, Eric S. Maddy, Narges Shahroudi, and Ross N. Hoffman

ML to augment or replace many components of the NWP processing chain—the processing chain that adds value at each of the series of steps from collecting and preprocessing Earth observations to the postprocessing and issuance of forecasts and warnings. As will be shown in the examples provided below, ML can 1) speed up and improve the processing of satellite data (quality control, gap filling, retrievals, etc.), 2) facilitate data assimilation and initialization of numerical weather and climate

Free access
Dan Lu, Goutam Konapala, Scott L. Painter, Shih-Chieh Kao, and Sudershan Gangrade

overfitting can be an issue and where the long lag time between precipitation and discharge needs to be simulated with UQ. d. Physics-informed hybrid LSTM model The performance of a data-driven LSTM model is mainly controlled by the quantity and quality of the training data. On the other hand, the performance of a physics-based hydrologic model is controlled by the reasonableness of the model structure and its parameterization. With limited historic observations, the purely data-driven LSTM cannot

Restricted access
Hanoi Medina, Di Tian, Fabio R. Marin, and Giovanni B. Chirico

global NWP models ( Bauer et al. 2015 ). The representation of these processes is especially challenging over continental areas from the Southern Hemisphere where the abundant vegetation and the sparse observations for evaluation and data assimilation have limited the models’ accuracy. Recent progress in forecasting tropical convection ( Bechtold et al. 2014 ; Subramanian et al. 2017 ) and the increasing quantity and quality of global information encourage the use of NWP for tropical precipitation

Full access
Yaling Liu, Dongdong Chen, Soukayna Mouatadid, Xiaoliang Lu, Min Chen, Yu Cheng, Zhenghui Xie, Binghao Jia, Huan Wu, and Pierre Gentine

, spatiotemporally explicit cropping pattern data are currently not available, we thus assume them to be static along time. Table 3. Major cropping pattern year round in the eight regions of China (e.g., fallow + maize stands for fallow in the winter followed by maize in the summer). c. Experiment design 1) Control experiments at station level To tease apart the contribution of the individual factors in determining SM variations, we conduct a series of control experiments with the station level observations

Restricted access
Eric D. Loken, Adam J. Clark, Amy McGovern, Montgomery Flora, and Kent Knopfmeier

-time predictions is cheap. For example, real-time RFFPs are currently being generated from 0000 UTC HREFv2 data. Including the preprocessing step, the RFFPs can be made in 30 min or less on a single processor. Nevertheless, ML-based postprocessing has several important drawbacks. Most notably, since ML-based techniques “learn” based on past results, they require quality historical datasets of sufficient length for both the forecast and observations. When modifications are made to the ensemble forecast system

Full access
Ricardo Martins Campos, Vladimir Krasnopolsky, Jose-Henrique G. M. Alves, and Stephen G. Penny

complexity of the ANN should be carefully controlled and kept to the minimum level sufficient for the desired accuracy of the approximation to avoid overfitting. Furthermore, the training set must represent the mapping for the ANN, with a sufficient sample size of properly distributed data points that adequately resolve the functional complexity of the target mapping. For environmental variables, at least one year is necessary in order to properly cover distinct conditions and a full seasonal cycle

Full access
Anthony Wimmers, Christopher Velden, and Joshua H. Cossuth

observations from the past and future (CMORPH has a 24-h latency). Because of this latency and use of future data it is unfortunately not a near-real-time estimation like those in Table 7 . Regardless, it shows great promise in the combination of passive microwave imagery with other modes of satellite observation. 7. Conclusions By following a fairly standard recipe for CNN modeling, this project has produced operational-quality TC intensity estimates using satellite bands that have traditionally not

Full access
Imme Ebert-Uphoff and Kyle Hilburn

intermediate layers to the output layer. The ultimate goal of the NN is to learn from data samples, which are provided in the form of input–output pairs, how to map the inputs to the outputs. Each neuron is connected to one or more neurons in the preceding layer, and each neuron’s state is represented by its activation value , a scalar variable that takes continuous values. A neuron’s activation value is calculated from the activation values of its directly preceding neurons through simple regression

Full access