xarray interp with method="cubic" fails for time/lon/lat interpolation #11342

Activity

1. mnergizci

   added

   needs triageIssue that has not been reviewed by xarray team memberIssue that has not been reviewed by xarray team member

   on May 15, 2026
2. headtr1ck commented on May 19, 2026

   

   headtr1ck

   on May 19, 2026

   Collaborator

   More actions

   Can you please post a minimal complete code example that we can test? Thanks!
3. headtr1ck

   added

   needs mcvehttps://matthewrocklin.com/blog/work/2018/02/28/minimal-bug-reportshttps://matthewrocklin.com/blog/work/2018/02/28/minimal-bug-reports

   on May 19, 2026
4. ghostiee-11 commented on Aug 10, 2026

   

   ghostiee-11

   on Aug 10, 2026

   More actions

   mcve

   import numpy as np, pandas as pd, xarray as xr
   
   nt = 288
   da = xr.DataArray(
       np.random.rand(nt, 71, 73), dims=("time", "lat", "lon"),
       coords={"time": pd.date_range("2026-01-01", periods=nt, freq="15min"),
               "lat": np.linspace(-87.5, 87.5, 71),
               "lon": np.linspace(-180, 180, 73)},
   )
   pt = dict(time=da.time[nt // 2], lon=12.3, lat=45.6)
   
   da.interp(**pt, method="linear").item()   # 0.24s, +42 MB
   da.interp(**pt, method="cubic").item()    # 3.93s, +1.8 GB

   dont think its xarray though. same call straight to scipy costs the same, 3.69s and 2.2 GB

   from scipy.interpolate import interpn
   interpn((np.arange(nt, dtype=float), da.lat.values, da.lon.values),
           da.values, np.array([[nt / 2, 45.6, 12.3]]), method="cubic")

   nd cubic builds a spline over the whole grid even for one point, linear only touches the cells around it. so it scales with the array not with how many points you ask for.

   @mnergizci if your tec is a month of 15 min maps thats ~2880 steps, 10x the above. that one got oom killed here (exit 137), so it looks like a hang but its memory.

   workaround, cubic on time only then linear on lat lon, 0.04s

   da.interp(time=pt["time"], method="cubic").interp(lon=12.3, lat=45.6, method="linear")

   scipy 1.18.0, xarray 2026.7.1.dev24
5. headtr1ck

   added

   topic-performance

   and removed

   needs mcvehttps://matthewrocklin.com/blog/work/2018/02/28/minimal-bug-reportshttps://matthewrocklin.com/blog/work/2018/02/28/minimal-bug-reports

   needs triageIssue that has not been reviewed by xarray team memberIssue that has not been reviewed by xarray team member

   on Oct 11, 2026