The authors present algorithms for selecting optimal study sites for landscape ecology studies which (a) minimise spatial autocorrelation between sites, and (b) maximise the values of particular metrics of interest. In general, the methods presented are sound, and very well explained. I feel that it would be fairly straightforward to replicate this method. This is made even easier by the code that has been shared - thanks for this helpful addition to the paper. I only really have one major and one minor comment:

Major: A method was previously published in MEE which aims to do something similar to the UDA approach (although without reducing spatial autocorrelation). I would expect to see some discussion around how this new method adds to what has already been done in this case, and how it compares.

Gillespie, M.A.K., Baude, M., Biesmeijer, J., Boatman, N., Budge, G.E., Crowe, A., Memmott, J., Morton, R.D., Pietravalle, S., Potts, S.G., Senapathi, D., Smart, S.M. and Kunin, W.E. (2017), A method for the objective selection of landscape‐scale study regions and sites at the national level. Methods Ecol Evol, 8: 1468-1476. https://doi.org/10.1111/2041-210X.12779

Minor: Ensure consistency in the algorithm names (the manuscript switches between SDA and stratefied, which confused me because I'd forgotten what SDA stood for). On a related note, part of my confusion was that the naming of the algorithms didn't really make much sense. UDA seems more like stratefied sampling (across different landscape metrics). Could these names be changed to reflect better what the algorithms do?