Representativity of a mesoscale network for weather-related factors governing pollen dispersal

Abstract

The cultivation of transgenic crops, such as maize, requires successful gene isolation in field environments. Five spatial statistical techniques are used to evaluate the use of a regional mesoscale observation network (Iowa Environmental Mesonet) as a means to drive field-scale pollen dispersion modeling. The Nearest Neighbor Index, Fractal Dimension, Morisita Index, Thiessen Polygons, and Coefficient of Representativity are computed showing the positive and negative impacts of sequential addition of observation networks into a mesonet framework (a collection of pre-existing networks). While it is shown that the arbitrary combination of disparate observing networks increases spatial resolution, this improvement is often at the expense of increased clustering due to co-location of observation sites near urban areas. Network composition in terms of density and degree of clustering was evaluated with a grid analysis using the Barnes scheme as a means to mitigate clustering and improve prediction accuracies when mesonet data are applied to modeling. This paper shows the importance of understanding and accounting for the spatial characteristics of an observational network before applying it to a modeling effort such as field scale pollen dispersion.