Application of recurrent neural networks for drought projections in California

Le, J. A.; El-Askary, Hesham; Allali, Mohamed; Struppa, Daniele C.

doi:10.1016/j.atmosres.2017.01.002

Cited by 42 publications

(19 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A hyperbolic tangent is used as the activation function, and the weights are initialized randomly from a uniform distribution between −0.1 and 0.1. Gradient descent (Werbos, 1990;Le et al, 2017) is used as the optimization algorithm, with a learning rate of 0.003 and a momentum of 0.01. In-depth testing was undertaken to adjust the details of the model's settings by comparing model skill for a varied number of model setups as in Hartmann et al (2016).…”

Section: Design Of the Study And Application Of The Neural Networkmentioning

confidence: 99%

Understanding the drivers of marine liquid-water cloud occurrence and properties with global observations using neural networks

et al. 2017

View full text Add to dashboard Cite

Abstract. The role of aerosols, clouds and their interactions with radiation remain among the largest unknowns in the climate system. Even though the processes involved are complex, aerosol-cloud interactions are often analyzed by means of bivariate relationships. In this study, 15 years (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015) of monthly satellite-retrieved near-global aerosol products are combined with reanalysis data of various meteorological parameters to predict satellite-derived marine liquidwater cloud occurrence and properties by means of regionspecific artificial neural networks. The statistical models used are shown to be capable of predicting clouds, especially in regions of high cloud variability. On this monthly scale, lower-tropospheric stability is shown to be the main determinant of cloud fraction and droplet size, especially in stratocumulus regions, while boundary layer height controls the liquid-water amount and thus the optical thickness of clouds. While aerosols show the expected impact on clouds, at this scale they are less relevant than some meteorological factors. Global patterns of the derived sensitivities point to regional characteristics of aerosol and cloud processes.

show abstract

Section: Design Of the Study And Application Of The Neural Networkmentioning

confidence: 99%

Understanding the drivers of marine liquid-water cloud occurrence and properties with global observations using neural networks

et al. 2017

View full text Add to dashboard Cite

show abstract

“…However, our results shows that the very strong El Niño of 2015/2016 appears in the modeled data sets on January 2017, but does not appear for the gauge data sets. This again concurs with the findings of Le et al[71] for climate division six.The climate divisions 1, 2, 3, 5, and 7 have different findings than climate division six. They all concur with the three very strong El Niños, as they are properly mapped out within their appropriate time frame.…”

supporting

confidence: 93%

Long Term Ground Based Precipitation Data Analysis: Spatial and Temporal Variability

Rodriguez¹

View full text Add to dashboard Cite

Long Term Ground Based Precipitation Data Analysis: Spatial and Temporal Variability by Luciano Rodriguez This dissertation evaluates response variables (classifiers) on various models applied to the detection of El Niño Southern Oscillation (ENSO) on California's seven climate divisions by using modeled and gauge (in-situ/ground) precipitation measurements and various climate indices. Three scientific studies were conducted as part of this research for evaluation of spatial and temporal ENSO events from modeled and gauge data using: 1) Wavelets 2) Autoregressive-moving-average (ARMA) model / Empirical Mode Decomposition (EMD) 3) Vector Generalized Linear Model (VGLM). This dissertation aims to propose and evaluate a methodology for developing a model to measure ENSO events accurately. The hypothesis is that precipitation data (either modeled or gauge) can be used to forecast ENSO events. One can create an assimilated index from various weighted indices as opposed to solely relying on popular climate indices as SOI, PDO, or Niño 3.4. Another objective is to identify how well modeled precipitation compares to gauge precipitation (ground truth) and if the composition of the indices are the same for both. A methodology for validating the generalization performance of the model is proposed and implemented (Controlled Parameter Cross-Validation). An analysis was performed using modeled data (regeneration of observable measurements and ground measurements) and gauge from seven climate divisions in California. VI

show abstract

“…The analysis proved the DNN model can effectively optimize the results from the available AOD products. Similarly, the usage of DNN also has been proved to be effective to predict and estimate the other parameters related to human health and living conditions such as PM10 and precipitation [67,68].…”

Section: Spatial Variability and Properties Of Aerosol Over The Selecmentioning

confidence: 99%

Urban Health Related Air Quality Indicators over the Middle East and North Africa Countries Using Multiple Satellites and AERONET Data

El-Nadry

Awad

et al. 2019

Remote Sensing

Self Cite

View full text Add to dashboard Cite

Air pollution is reported as one of the most severe environmental problems in the Middle East and North Africa (MENA) region. Remotely sensed data from newly available TROPOMI - TROPOspheric Monitoring Instrument on board Sentinel-5 Precursor, shows an annual mean of high-resolution maps of selected air quality indicators (NO2, CO, O3, and UVAI) of the MENA countries for the first time. The correlation analysis among the aforementioned indicators show the coherency of the air pollutants in urban areas. Multi-year data from the Aerosol Robotic Network (AERONET) stations from nine MENA countries are utilized here to study the aerosol optical depth (AOD) and Ångström exponent (AE) with other available observations. Additionally, a total of 65 different machine learning models of four categories, namely: linear regression, ensemble, decision tree, and deep neural network (DNN), were built from multiple data sources (MODIS, MISR, OMI, and MERRA-2) to predict the best usable AOD product as compared to AERONET data. DNN validates well against AERONET data and proves to be the best model to generate optimized aerosol products when the ground observations are insufficient. This approach can improve the knowledge of air pollutant variability and intensity in the MENA region for decision makers to operate proper mitigation strategies.

show abstract

Application of recurrent neural networks for drought projections in California

Cited by 42 publications

References 32 publications

Understanding the drivers of marine liquid-water cloud occurrence and properties with global observations using neural networks

Understanding the drivers of marine liquid-water cloud occurrence and properties with global observations using neural networks

Long Term Ground Based Precipitation Data Analysis: Spatial and Temporal Variability

Urban Health Related Air Quality Indicators over the Middle East and North Africa Countries Using Multiple Satellites and AERONET Data

Contact Info

Product

Resources

About