Search Results

You are looking at 1 - 8 of 8 items for :

  • The 1st NOAA Workshop on Leveraging AI in the Exploitation of Satellite Earth Observations & Numerical Weather Prediction x
  • Refine by Access: All Content x
Clear All
John L. Cintineo, Michael J. Pavolonis, Justin M. Sieglaff, Anthony Wimmers, Jason Brunner, and Willard Bellon

. The model of this paper uses ReLU for the activations of the first two fully connected layers and uses the sigmoid function for the activation of the final fully connected layer with a single output, forcing the final prediction to be a probability between [0, 1]. We used the Keras Python API with TensorFlow backend to perform the training and evaluation of CNNs ( Chollet 2015 ). This is a binary classification problem (“intense” or “ordinary” convection are the classes), so the loss function

Restricted access
Andrew E. Mercer, Alexandria D. Grimes, and Kimberly M. Wood

Zhang (2013) and Tao and Jiang (2015) suggesting that shallow to moderate depth convection could help to trigger RI, while deep convection was a response to RI. These recent studies still do not fully encompass all RI situations as evidenced by the ongoing challenge of improving RI forecasts. These limitations force investigators to identify relevant TC environmental features associated with RI using an event-centric approach [e.g., Hurricane Opal 1995 ( Bosart et al. 2000) ; Hurricane Felix 2007

Restricted access
Ryan Lagerquist, Amy McGovern, Cameron R. Homeyer, David John Gagne II, and Travis Smith

except the last follow this linear transformation with leaky ReLU and batch normalization, like the convolutional layers. The last dense layer uses the sigmoid activation function (section of Goodfellow et al. 2016 ), which forces the output to range over [0, 1], allowing it to be interpreted as a probability. The last dense layer does not use batch normalization, because this would force the outputs to a Gaussian distribution, which permits values outside [0, 1] and is therefore invalid for

Free access
Noah D. Brenowitz, Tom Beucler, Michael Pritchard, and Christopher S. Bretherton

( Beucler et al. 2018 ; Kuang 2018 , 2010 ; Herman and Kuang 2013 ). These linearized response functions (LRFs) are typically computed by perturbing inputs in some basis and reading the outputs (appendix B of Beucler 2019 ) or by perturbing the forcing and inverting the corresponding operator ( Kuang 2010 ). If the input/output bases are of finite dimension, then the LRF can be represented by a matrix. LRFs can also be directly computed from data, for instance, by fitting a linear regression model

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

September 2014 to 13 October 2016, a total of 774 daily observations. Pumphouse has 3 years of data from 1 October 2014 to 30 September 2017, while Rock Creek has data from 31 August 2014 to 4 October 2017, a total of 1131 daily observations. Gauge station 09112500 has over 86 years of continuous daily streamflow records from 1 October 1934 till present. Besides the observed streamflow, we have catchment aggregated meteorological forcing data such as precipitation, daily maximum and minimum temperatures

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

final section concludes with a summary and outlook. SATELLITE REMOTE SENSING AND NWP CHALLENGES AND ML. The many challenges to satellite remote sensing and NWP are grouped here by forcing mechanism—Big Data, advanced models and applications, and user demands. In the sections that follow we show that recent advances in ML in terms of efficiency, capability, and ease of implementation, can help to meet these challenges. First, NWP is failing to exploit the growing diversity and volume of observations

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

with the proper value of σ , spatially smoothing ensemble probabilities reduces sharpness (e.g., Sobash et al. 2011 , 2016 ; Loken et al. 2017 , 2019 ) and potentially sacrifices resolution if too much smoothing is required. Moreover, the “best” value of σ may vary based on geographic location and time of year (e.g., Fig. 3 ), as precipitation uncertainty is reduced where stronger and/or more predictable forcing is present, such as near high terrain (e.g., Blake et al. 2018 ) or during the

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

interference of dense plant canopy and limitations in the depth of detection within the soil column ( Entekhabi et al. 2014 ), as well as sensor limitations (especially before the use of L-band radiometers). Likewise, the fidelity of LSM simulations may be undermined by model errors (e.g., model structure error, model parameter error), forcing data uncertainties and limited usage of ground-based observations for model calibration and validation ( Xia et al. 2014 ). Thus, efforts that aim to improve large

Restricted access