Spatial variability of methane emissions from rice fields and implications for experimental design

It is observed that side by side plots in rice fields, managed by the same practices, produce methane emissions that are often different by factors of 2 to 4. Similarly on a given day when many plots are sampled, the emissions from one plot may differ from another by factors of 3 - 4 on average. These large variations must be taken into account if these data are extrapolated to larger scales such as countries or the world. In this paper we analyze and delineate the nature of this variability. We distinguish temporal and spatial variability and examine the effects of both on the observed seasonal average emission rates. While temporal variability is managed well in most experiments because frequent measurements are taken, the spatial variability remains a potential major uncertainty in many experiments. Using the characteristics of the observed variability we use simulations to show that with a proper and realistic sampling strategy, it is possible to get to within 15% - 20% of the true flux for most cases of interest. We calculate how many plots are needed and how often they should be sampled during the growing season to obtain a seasonal average flux within prescribed limits relative to the true flux.