uarenforecasting / advi Goto Github PK

About 1Mb of RGBA values (4 bytes per pixel) are sent for each map update. This translates to slow updating when off campus on slower internet. Consider only setting the alpha value initially and only sending RGB values for 3 bytes per pixel. Or try sending only the color bin for 1 byte per pixel.

similar to https://ensemble.ucar.edu/sounding.php?d=2017100500 where one can click on the map to get the forecasted skew-t for that location. Probably easiest to use https://github.com/rsobash/d3-skewt, send the data via bokeh, and have a bokeh javascript callback to update the d3 data.

Following the HRRRE methodology:

https://rapidrefresh.noaa.gov/hrrr/HRRRE/

“HRRRE probabilities are the fraction of members that exceed a given threshold, or predict a given precipitation type, at a point. The final probability field (100*(n/total)) is smoothed using a Gaussian filter of width 25 km.”

let's start with composite reflectivity > 50 dbz, then move on to hourly precip, then 24 hour precip. Might want to do precip in a separate pull request due to the time complication.

@lboeman I will sketch out the algorithm for this and then leave it to you to move to production.

Need to add wind direction to variables. Right now that would require a WSPD and WDIR file and for the bokeh app to open both files. Seems more reasonable to put spd and dir in the same file. It is nice that each variable has it's own file so that not all models/initializations need all variables. A reasonable approach might be to make a WIND file with /wspd/time and /wdir/time arrays and make other variable files in consistent way.

try a light gray background