15.3 Enhanced Precipitation Estimation using GOES-R

Thursday, 26 January 2017: 2:00 PM
Conference Center: Yakima 2 (Washington State Convention Center )
Christopher J. Mattioli, MIT Lincoln Laboratory, Lexington, MA; and M. S. Veillette, H. Iskendarian, P. M. Lamey, E. P. Hassey, J. Mecikalski, and G. T. Stano

In areas that lack ground-based weather sensing radar, estimations of radar-like precipitation intensity rely heavily on visible and infrared data from geostationary satellite and lightning data from ground based detection networks. The geostationary satellite imagers on GOES-13 and GOES-15 provide one visible and four infrared channels over both land and ocean, and lightning networks typically suffer a loss in detection efficiency over ocean. The advent of GOES-R brings two new instruments that can be used for improving estimation of precipitation:  the Advanced Baseline Imager (ABI) and the Geostationary Lightning Mapper (GLM). The ABI will provide more timely data at a higher resolution than currently available. Additionally, it will increase the number of channels to sixteen including multiple new visible channels, water vapor channels, and window channels. The GLM, a technology currently unavailable on the sensing devices today, will provide 70-90% detection efficiency of total lightning over land and ocean. In this work, we seek to assess the potential benefits these new and augmented instruments can create for enhanced precipitation estimation.

First, we quantify benefit of the extra channels in the estimation of precipitation intensity. Using the Advanced Himawari Imager on the Himawari-8 geostationary satellite as a proxy for the ABI, we have constructed a supervised machine learning methodology for the estimation of radar-like precipitation intensity.  We derive the truth precipitation intensity data from the Global Precipitation Measurement Mission (GPM) Dual-Frequency Precipitation Radar (DPR) and time-align these data with Himawari-8 satellite images. A model is first trained using only those wavelengths available in GOES-13/15 as features. Once this baseline model is established, a series of models are trained by adding in a new GOES-R channel one at a time to the feature set. The performance of these models is then compared to the performance of the baseline GOES-13/15 model.

To assess the benefit of GLM, GOES-13/15 imager data is fused with total lightning data to create a virtual radar-like analysis of precipitation intensity. To do so, we post-process total lightning data obtained from NASA’s Pseudo Geostationary Lightning Mapper (PGLM) dataset. This data is used as a proxy for the lightning data generated from the GLM. This work will describe the use and benefits of combining lightning data with other data sources to construct an improved picture of radar-like precipitation intensity.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner