phtrempe / sknni

Spherical k-nearest neighbors interpolation (geospatial interpolator)

Hi there! 👋 I'm Philippe Trempe (PhTrempe on GitHub).

I'm a Machine Intelligence Research Engineer. 🧠🖥️

• 🔭 I’m currently working on a variety of projects pertaining to AI and deep learning.
• 🌱 I’m continuously learning to better bridge the gap between academia and industry.
• 👯 I’m looking to collaborate on various public and private academic and industrial projects.
• 💬 Ask me about anything related to AI/ML/DL, data science/engineering, Python, etc.
• ⚡ Fun fact: I really enjoy creating (worldbuilding, story writing, music composition, vectoring, games and other software).

Connect with me

Languages and Tools

sknni's Issues

How to use for use for real weather data

Can you please show how it can be used for weather time series data for converting time series to weather maps,

Like here on your monograph

The range of latitude angle

Hi Philippe,
I just tried to use your code for spherical data interpolation and it worked very well! While I was trying to figure out which coordinate system is being used, it occured to me that the range of your altitude angle is between [-90,90]. However, in your __polar_to_cartesianfunction, you use
z = r * np.cos(lat)
that means, the z-coordinate is always positive? I'm not sure if this will affect the result.

Thank you very much!

mean value of longitude angles

Hi,
I noticed, that you are calculating the mean value of the longitudinal angles in __nddnisd_interp_fn(). The problem is, if you are using np.mean you are adding e.g. -175 and 180 to 2.5 and not the -177.5. This can be fixed by using scipy.stats.circmean().

However, when I changed it, I noticed that the results look wrong, but I don't understand why at the moment. Do you have any ideas?