Investigating the random relocation of gauges below the canopy by means of numerical experiments

Roving, or the random relocation of N gauges among M positions, was studied via numerical experiments. Continuous throughfall measurements taken during 8 months with N = 22 fixed gauges placed below an evergreen heath tree-laurel forest in the Garajonay National Park (La Gomera, Spain) were randomized numerically, simulating the effect of spatial repositioning and time frequency relocation. In general, we found that the roving strategy was superior to the fixed gauge arrangement in terms of minimization of both the coefficient of variation (CV) and the dispersion around the mean cumulative throughfall. These conclusions were not constrained by the number of locations M = 22 where throughfall was measured in our plot, since similar patterns were reproduced using synthetic data. Furthermore as M was synthetically increased from 22 up to 220, the reduction in the dispersion of the CV and mean throughfall in the roving versus the fix arrangement was shown to be more evident. Reducing the number of roved gauges lowered the CV but increased the dispersion around the estimated mean throughfall, such that a compromise arose between the two statistics. Furthermore, we found that an increase in the relocation frequency implied a quadratic exponential reduction in the 97.5% CV confidence limit. In our case, an optimum roving arrangement consisted of 14 gauges, repositioned daily among 22 positions, yielding a CV range of 21.5–38.9% and a mean throughfall (computed as the 14 cumulative measurements average) lying within the interval 373–422 mm.