One of the factors that influences the most on the sampling design is the control of the spatial and temporal patterns during the study period. Let's give an example. If collaborators were allowed to choose among the 386 squares without restrictions, as it was the case for the Breeding Bird Atlas, many squares in areas far from big towns would probably remain unsurveyed until the last year of study, such as those in the Ebro comarques and Pyrenees. Interannual changes in distribution during the breeding season appear to be of minor importance, and their bias is acceptable. It is not the case for the winter season when the distribution of some species may significantly vary due to weather and irruptive processes. Therefore, if some areas were left unsurveyed until the last year, it would not be possible to conclude whether the obtained distribution was due to a specific area effect or to a year effect. In order to adequately tackle with this issue, the sampling must be carefully distributed over the study period. The random design of squares to be surveyed every year is a first attempt, which, theoretically, we believe appropriate. However, we consider necessary to apply a restriction that minimizes the number of adjacent squares surveyed, the same protocol used for the design of the SOCC priority squares (Figure 7). This design enables an accurate control on the interannual changes whether at a Catalan or at a biogeographical region scale, but does not prevent that, at a UTM 10x10 square scale, the spatial and temporal patterns modify our observations.

The adoption of this design has important consequences for the project's coordination. 129 squares will have to be surveyed in the first two winters, and 128 in the last. Therefore, the field work in each square will have to be done in its pre-established winter and only then. Nevertheless, collaborators will be encouraged to submit complementary data belonging to other winters in order to complete the distribution maps, although these data won't be used for to evaluate interannual differences or for comparisons among squares. Indeed, this design involves the need of professionals surveying these areas so that these could be adequately covered from the beginning.



Figure 8. Distribution of the extensive sampling of the Winter Bird Atlas over the three study years. 129 squares will be surveyed every year (only 128 the last year) in a random arrangement although minimizing the number of adjacent squares within a year so that it can be obtained an evenly distributed representation of the territory for this time period. It is a scheme that combines systematic sampling with some degree of randomness.