What is the proper way to treat geographic coordinate data (latitude and longitude) for a spatio-temporal time series model for ML algorithms?

Question

I am processing a CHIPRS dataset for rainfall prediction, where I have as inputs the geographical coordinates (latitude and longitude), the day of the rainfall reading and the rainfall value, with respect to the date of reading I am extracting the month and the year and put them in a sine and cosine function separately as follows:

sin-day, cos-day, sin-month, cos-month: in this way I can tell the model which day is closer to the other, as well as the months, but I do not know how to treat the latitude and longitude variables, all this thinking in predictive models of monthly precipitation as for example LTMS or GRU.

I appreciate your help in advance.