Statistical analysis on geospatial data? Converting projection to x,y?

[Jessica]

Hello Geopandas!

Graduate student here working on a thesis project. I have been using your module for said project, which involves taking tornado tracks, snapping them all to the same point (right now 0,0 latitude and longitude), and seeing how the spread of tracks differs based on different environments. This work is meant to try and quantify uncertainties in tornado movement to help make tornado warnings more efficient.

There are two problems I am facing right now, and I was hoping I could get help or advice on them. I have spoken already to two GIS professors who did have some advice but do not know if Geopandas has the tools for it (ArcGIS might?).

The first problem is that all of the tornado tracks are located in the United States and my code moves them to the equator, which is over 40 degrees of latitude and 80 degrees of longitude difference. The projection skew is likely going to cause errors in representing the tracks to how they would look like to a National Weather Service forecaster who needs to make a tornado warning polygon.

The tornado tracks are obtained from the NWS in shapefile format. Each tornado track is a linestring of latitude and longitude pairs, and graphing them results in the x and y axis being in degrees, not miles of distance. I was thinking of first setting a CRS that's suited for the United States (easily doable in Geopandas), but then converting that lat and long into y and x, possibly in miles east/north away from some center point (much harder). One of the professors suggested I instead snap all of the tornado tracks to a point in the US, such as (40, -80), but this does not fix the axis issue. I know Geopandas has functionality to convert lat/lon into x and y, as after all that is necessary to plot shapefiles in the first place, but I couldn't find any function to do this easily either through StackOverflow or through documentation.

The second problem ties into the first. I am looking to run statistical analysis on all of those tornado tracks once they are snapped at the same point to get standard deviations and try and quantify how much those tracks differ in different tornado environments. I should be able to get help from this from statistics professors, but I would first need to convert those linestrings from geographic into geometric, which ties into my first problem. Otherwise, is there any Geopandas tools that allows for straight up statistical analysis on a bunch of linestring elements?

Any sort of advice is appreciated. According to one of the GIS professors, this might take an entire literature review to solve, but I am hoping that perhaps someone here might be able to point me in the right direction.

Here's a preliminary plot of about 80% of all tornadoes from the past 5 years that I have generated and snapped to a single point (ignore the color bar, it is an atmospheric environmental variable I am looking at):

-Jessica