0
$\begingroup$

I have daily weather data for almost 50 years for a weather station, and I want to predict the next 100 days' weather. I use python and all of the tools I've used so far (pmdarima.auto_arima, statsmodels.ARIMA, etc) run either infinitely or just crash my Colab or Jupyter notebook. What would be the best way to forecast this data with ARIMA?

The ACF and PACF plots also generally give such results that I cannot interpret meaningful p,d,q orders. For instance, adf-test gives p-value 0.000 for the seasonal part and this is its PACF and PACF plots:

enter image description here

enter image description here

adf-test suggests the seasonal part is stationary but PACF plot still gives significant lag values even up to 35 (which would make running an ARIMA model longer). Any advice would be appreciated.

$\endgroup$

1 Answer 1

0
$\begingroup$

Predicting 100 days is probably impossible with SARIMA and 1 station.

Even current supercomputers can forecast 7 days with good accuracy and up to 15 days with average accuracy (depending on regions): They cannot forecast 100 days.

https://www.mprnews.org/story/2020/01/02/forecast-models-keep-hinting-at-subzero-air-ahead

Having one station, I recommend starting to predict 1 or 2 days.

If you want to predict longer, you will probably need more data to cross information, complex models, and a lot of computing power.

$\endgroup$
1
  • $\begingroup$ Does it answer your question? If not, please let me know to provide additional information. $\endgroup$ Dec 13, 2022 at 13:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.