Influence of sampling size in the estimation of mean throughfall in two Mediterranean holm oak forests

Throughfall has a large spatial variability due to the heterogeneous structure of the canopy and to variable rainfall patterns. This study has been conducted in two Mediterranean holm oak (Quercus ilex, L.) forests, here named LC and RP, to determine the optimum number of collectors needed to obtain a mean throughfall value within certain limits of error. To this end, we have applied a resampling technique that created mean values for a variable number of collectors (n) ranging from 2 to 31. This resampling has allowed us to relate some statistical parameters of the obtained distributions (the mean, ±90 and 95% confidence limits, coefficient of variation) with the number of collectors used in the generation of these distributions. Results are presented for the whole data set and for a partition considering: (1) weeks of rainfall <5 mm; (2) weeks of rainfall >40 mm; and (3) alternate observations. For the whole data set, with 9 collectors at LC and 11 at RP, the error in the mean weekly throughfall is around 10%. This error is reduced to 5% when using 22 and 23 collectors at LC and RP, respectively. The number of observations considered has not modified the spatial variability (therefore not influencing the number of collectors needed to obtain a mean with certain error), but the variability is clearly higher for small rainfalls.