Table 1.
Description of the models we consider for interpolating Argo temperature data. Model 1 is a reimplementation of the procedure developed by Roemmich & Gilson (RG) in [7] and models 2–6 are variants of locally stationary interpolation. The models differ in terms of how time is taken into account, in the distribution of the fine-scale variation represented by the nugget effect and in the length of the temporal window used in the fit (see text for more details).
| model | time window | mean | covariance | nugget |
|---|---|---|---|---|
| 1 | February | RG (spatial) | RG-like | Gaussian |
| 2 | February | RG (spatial) | local (spatial) | Gaussian |
| 3 | February | RG (spatial) | local (spatial) | Student |
| 4 | January–March | RG (spatio-temporal) | local (spatial) | Gaussian |
| 5 | January–March | RG (spatio-temporal) | local (spatio-temporal) | Gaussian |
| 6 | January–March | RG (spatio-temporal) | local (spatio-temporal) | Student |