One of the greatest difficulties in severe storm forecasting is deciding where and when storms will initially form. Current numerical models struggle with this problem and often have large errors in their 1-6-hour forecasts for convective initiation (CI). The Advanced Baseline Imager (ABI) aboard GOES-R will provide an unprecedented array of spectral bands at improved spatial and temporal resolution relative to the current geostationary satellites, and offers great promise in improving skill in short-term CI forecasts. The overall goal of this project is to develop a single objective system that predicts where and when storms will form 1-6 hours prior to initiation. Up to this point, the majority of satellite-based CI research has focused on nowcasting which growing cumulus or towering cumulus clouds will develop into precipitating thunderstorms during the next 30-60 minutes. Despite the importance of this forecast problem, very little research has focused on determining what short-term predictive (e.g., during the next 1-6 hours) information is available from satellite data prior to cumulus cloud formation, or whether there is information within an initial field of shallow cumulus clouds about whether (and which) of these clouds will eventually become storms, or what regions are more likely to generate new deep convection.

Through this effort, five components that increase understanding of the factors leading to CI will be used, all of which exploit forthcoming ABI datasets. First, channel differences can be used to forecast where clouds and storms will form. Although there are 3 “water vapor” bands on the ABI, even the 7.3 µm band's weighting function peaks relatively far from the Earth's surface, so little (if any) information about boundary layer moisture exists in those bands. However, the 12.3 µm band is very sensitive to boundary layer moisture, and the split window difference (10.35 – 12.3 µm) can be used to identify regions of enhanced or deepening low-level water vapor. Second, mesoscale atmospheric motion vectors (AMVs) will be used to provide information on the mesoscale wind field. The added value of the simulated MAMVs in locating low-level moisture convergence, upper-level divergence, and identifying jet streaks will be explored in the context of pre-convective environments, including nearcasting tools. The GOES-R MAMV and split-window moisture data will be compared and blended with wind observations from the WSR-88D network using the VDRAS to provide a more complete picture of the pre-storm wind and moisture fields. The information on boundary layer convergence (from the Doppler and/or GOES MAMV wind data), changes in low-level water vapor deduced from the split window, and other relevant physical parameters (such as CAPE and CIN), will be used as predictors in developing an algorithm for thunderstorm formation. The intensity of the CI signal developed above will then be correlated to the intensity and timing of convective storms. Lastly, a method that identifies locations of land-surface sensible and latent heating gradients using GOES insolation and MODIS NDVI will be implemented to develop a “surface-heating index.” The heating (which applies to lands with minimal heterogeneity) and land surface variability (which alternatively considers land-surface heterogeneity) indices will be used to help identify the development of mesoscale circulations, resembling inland sea breezes and so-called “non-classical mesoscale circulations”. The quantified probabilistic results will integrate the results from CIRA, UW-CIMSS, UAHuntsville, CREST and NSSL for the formation of a 1-6 hour CI nowcasting system. Maps of 1-6 hour CI probability will be an end product.