Enterprises are making machine learning for production as an integral part of their future roadmaps and Earth science domain is no exception. However, there is common problem in transitioning machine learning from science to production due to a major difference in constructing a model versus deploying it for people to use to make decisions. Phases of machine learning lifecycle that includes model transition to production using a successful application is discussed.

Tropical cyclone intensity estimation is a challenging task as it required domain knowledge while extracting features, significant pre-processing, various sets of parameters obtained from satellites, and human intervention for analysis. The inconsistency of results, significant pre-processing of data, complexity of the problem domain, and problems on generalizability are some of the issues related to intensity estimation. In this study, we design a deep convolutional neural network architecture for categorizing hurricanes based on intensity using graphics processing unit. Our model has achieved better accuracy and lower root-mean-square error by just using satellite images than 'state-of-the-art' techniques. Visualizations of learned features at various layers and their deconvolutions are also presented for understanding the learning process.

The laboratory calibration of airborne Hurricane Imaging Radiometer's C-Band multi-frequency receivers is described here. The method used to obtain the values of receiver front-end loss and injected noise diode temperature is presented along with the expected RMS uncertainty in the final calibration. Internal Warm load was excluded from the calibration due to an apparent anomaly.

The performance characteristic of a digital correlation radiometer signal processing back end (DBE) is analyzed using a computer simulator. The particular design studied here corresponds to the airborne Hurricane Imaging radiometer which was jointly developed by the NASA Marshall Space Flight Center, University of Michigan, University of Central Florida and NOAA. Laboratory and flight test data is found to be in accord with the simulation results. Overall design seems to be optimum for the typical input signal dynamic range at the DBE input. It is found that the performance of the digital kurtosis could be improved by lowering the DBE input power level. An unusual scaling between digital correlation channels, observed in the instrument data, is confirmed to be a DBE characteristic.

The Hurricane Imaging Radiometer (HIRAD) is an airborne microwave radiometer developed to provide wide-swath hurricane surface wind speed and rain rate imagery for scientific research. This letter presents a geometric evaluation of the brightness temperature (Tb) images produced by HIRAD for high-contrast land/water targets. Methodologies used to validate geolocation accuracy and spatial resolution are discussed, and results are presented to provide quantitative pixel geolocation accuracy and the effective image spatial resolution of the Tb image.

On board of the NASA's Global Hawk (AV1) aircraft there are two microwave, namely: the passive microwave Hurricane Imaging Radiometer (HIRAD), and the active microwave High-altitude Imaging Wind and Rain Airborne Profiler (HIWRAP). This paper presents results from an unplanned rain rate measurement validation opportunity that occurred in 2013, when the Global Hawk aircraft flew over an intense tropical squall-line that was simultaneously observed, by the Tampa NEXRAD meteorological radar. During this experiment, Global Hawk flying at an altitude of 18 km made 3 passes over the rapidly propagating thunderstorm, while the TAMPA NEXRAD perform volume scans on a 5-minute interval. NEXRAD 2D images of rain rate (mm/hr) were obtained at two altitudes (3 km & 6 km), which serve as surface truth for the HIRAD rain rate retrievals. In this paper, results are presented of the three-way inter-comparison of HIRAD Tb, HIWRAP dbZ and NEXRAD rain rate imagery.

The Hurricane Imaging Radiometer (HIRAD) is an airborne passive microwave radiometer designed to provide high resolution, wide swath imagery of surface wind speed in tropical cyclones from a low profile planar antenna with no mechanical scanning. Wind speed and rain rate images from HIRAD's first field campaign (GRIP, 2010) are presented here followed, by a discussion on the performance of the newly installed thermal control system during the 2012 HS3 campaign. The paper ends with a discussion on the next generation dual polarization HIRAD antenna (already designed) for a future system capable of measuring wind direction as well as wind speed.