Ranking Strategy for Large Temperature Dataset

I have a large dataset of daily global temperatures for 43 years. Imagine a 43x365x720x1440 matrix (43 years, 365 days, 0.25 degree latitude from 90N to 90S, 0.25 degree longitude from 0 to 360).

I'd like to take the daily temperature for the current year (Imagine a 1x365x720x1440 matrix) and rank each lat/long temperature based on the full 43 years of data. For example, for each lat/long grid point on a given day, determine from 1 to 43 (1 being warmest, 43 being coolest) where it ranks over the entire 43 years of data for that day. Ideally, I'd also do this as a 7-day moving average (in other words, each daily data point is centered on 7 days to smooth out the daily variation).

I'm trying to determine the most efficient MATLAB ranking solution/strategy to this problem. It feels like creating a gigantic 43x365x720x1440 matrix would clog up a ton of memory.

Any creative ideas on where I might start to tackle this problem?

Answer 1

[Taylor]

timetables are quite useful when working with data that is regularly spaced in time. You can index to specific time points and ranges and perform your calculations on that subset of data.