Friday, March 28, 2014

Presenting in Machine Learning Conference in NYAS today

Creating High-Resolution Climate Meteorological Forecasts by Application of Machine Learning Techniques

Nabin Malakar, PhD, Emmanuel Ekwedike, BS, Barry Gross, PhD, Jorge Gonzalez, PhD, and Charles Vorosmatry, PhD
The City College of New York, New York, New York, United States;

In order to study the effects of global climate change on a regional scale, the low resolution GCM forecast data needs to be intelligently adapted (downscaled) so that it can be injected into high resolution models such as terrestrial ecosystems. Our study region is the North East domain [{35N, 45N} x {-85W,-65W}]. In particular, we focus on High and Low temperature extremes within the Daymet data set, while the low resolution climatology (at 0.5 deg) MET data are obtained from the The Inter-Sectoral Impact Model Intercomparison Project (ISI-MIP) climatology forecast database.  Although the injection of regional Meteorological Models such as Weather Research and Forecasting (WRF) can be attempted where the GCM conditions and the forecasted land surface properties are encoded into a future time slices, this approach is extremely computer intensive. We present a two-step mechanism by using low resolution meteorological data including both surface and column integrated parameters, and then by combining high resolution land surface classification parameters to improve on purely interpolative approaches by using machine learning techniques. 

Application of Machine-Learning for Estimation of PM2.5 by Data Fusion of Satellite Remote Sensing, Meteorological Factors, and Ground Station Data

Lina Cordero, MS, Nabin Malakar, PhD, Yonghua Wu, PhD,  Barry Gross, PhD,  Fred Moshary, PhD
Optical Remote Sensing Laboratory, CCNY, New York, New York, United States;

Particulate matter with dimension less than 2.5 micrometers (PM2.5) can have adverse health effects. These particles can enter into the blood streams via lungs, reach vital organs and cause serious damages by oxidative inflammations. We present our latest progress in obtaining correct estimates of PM2.5 on regional scale by using machine learning techniques. Specifically, we apply a neural network method for better describing the non-linear conditions surrounding the PM2.5-MODIS AOD while at the same time investigating dependencies on additional factors or seasonal changes.  In our local test, we find very good agreement of the neural network estimator when AOD, PBL, and seasonality are ingested (R~0.94 in summer). Next, we test our regional network and compare it with the GEOS-CHEM product. In particular, we find significant improvement of the NN approach with better correlation and less bias in comparison with GEOS-CHEM. We also show that further improvements are obtained if additional satellite information and land surface reflection, is included. Finally, comparisons with Community Multi-scale Air Quality Model (CMAQ) PM2.5 are also presented.

Using NN techniques to ingest Meteorological Weather Satellite data in support of Defense Satellite Observations

Crae Sosa, BS, Gary Bouton, MS, Sam Lightstone, MS, Nabin Malakar, PhD,  Barry Gross, PhD and Fred Moshary, PhD
The City College of New York, New York, New York, United States;

The need to observe thermal targets from space is crucial to monitoring both natural events and hostile threats. Satellites must choose between high spatial resolution with high sensitivity and multiple spectral channels. Defense satellites ultimately choose high sensitivity with a small number of spectral channels. This limitation makes atmospheric contamination due to water vapor a significant problem which can not be determined from the satellite itself. For this reason, we show how it is possible to ingest meteorological satellite data using NN to allow for the compensation of water absorption and re-emssion in near-real time