Yes, spatial econometric models can be applied to non-exhaustive survey data, provided that the spatial structure is appropriately defined and that potential sampling biases are accounted for. Many studies have successfully used spatial models with sampled micro-level data, particularly in agriculture and environmental economics.

The key consideration is how spatial relationships such as contiguity or distance matrices are constructed when not all units in the population are observed. As long as the spatial weights matrix reflects the structure of observed interactions (e.g., between surveyed winegrowers), spatial dependence can be effectively modeled. However, care must be taken in interpreting the results, as unobserved neighbors may influence behavior but are not captured in the data.

The study by Lardeux and Merly-Alpa (2018) is indeed relevant; it demonstrates that spatial models can provide meaningful insights even with non-exhaustive data. I also recommend looking into foundational texts such as LeSage and Pace (2009) on spatial econometrics, and recent applied work that deals with spatial spillovers in agriculture using sampled data.

You may also consider robustness checks using non-spatial models or simulation-based approaches to assess how non-exhaustiveness might affect your estimates. If your sample is large and representative, spatial econometric models remain a valid and valuable tool for your research.