Modelling the transport of pollen by wind

Dr Eva Novotny, SGR, and Dr Jean Perdang, University of Liege, present in-depth analysis of the transport of pollen by wind, which has significant implications for setting separation distances between fields of GM and non-GM crops.

Two in-depth papers, 18 May 2014

In May 2002, SGR submitted evidence to a UK government hearing on whether to approve a variety of genetically-modified maize called Chardon LL. As part of this evidence, we carried out mathematical modelling of the transport of pollen by wind, to examine how much pollen is deposited at various distances from a source. This is of relevance to the setting of separation distances between GM and non-GM-crops to minimise cross-pollination. Our work showed that ‘islands’ with high levels of pollen could occur far from the crops of origin, leading to a significant unanticipated risk of cross-pollination.

In this latest work, we have taken the analysis further and written two in-depth papers outlining our mathematical modelling, results and conclusions as follows.

Paper I [pdf, 0.35MB] describes the computer program and assumptions, including a key parameter, the ‘sticking probability’, which quantifies the likelihood of pollen deposition at any particular place at any time during the simulation. The computational technique employed is a cellular automaton (CA) model, which divides the area under investigation into rectangular cells and then operates on each cell individually according to prescribed rules, which may incorporate stochastic (random) processes. All calculations are performed in dimensionless units, which can be translated into physical units at the end of the simulation by assigning the required values to certain quantities. Calculations begin at one end of the bounding rectangle (the field) and proceed row-by-row, and cell-by-cell within the row, across the field.  At each step and at each cell, wind direction and speed are pre-assigned, but conditions such as amount of pollen deposited are calculated. The CA results are compared in particular with observed data on diminishing pollen deposit with distance, as well as with the traditional theoretical representation by a decaying exponential. Our CA results provide a significantly better simulation of the empirical data than does the exponential decay pattern. Dips and peaks similar to those in observational data can also be produced in our models. A degree of validation of these methods is obtained by calculating the profile of airborne pollen created instantaneously in a single row of plants and then blown across a field (without deposition) by a uniform wind: this shows the expected gradual broadening of the profile. The specification of winds, pollen creation, number of layers of air in a three-dimensional model (which was simplified in the present work), terrain and other conditions can be made in a highly flexible way, allowing the program to be used in as simple or as complex a way as desired.

Paper II [pdf, 1.6MB] uses the CA modelling program to simulate pollen deposition for various values of the ‘sticking parameter’ and for various distances from the source of pollen for various formulations of the wind, represented either analytically or by an analytical approximation to actual wind data.  We evaluate the ‘sticking parameter’ by comparing CA calculations of relative amounts of deposited pollen as a function of distance with observed values. As expected, the stronger the wind, the farther does the deposited pollen reach. Irregularities in the wind produce non-uniform deposition.

Our results are particularly relevant for considerations of establishing so-called ‘co-existence’ guidelines by government, to govern planting regimes of GM and non-GM crops. We find long fingers and isolated patches of high pollen deposit well beyond what might be regarded as the distance by which most of the pollen has been deposited. While this is not important for crops like grains that are mixed (assuming they are well mixed) over a whole field after harvesting, crops like ‘corn-on-the-cob’ that are sold individually could be over the legal limit for unlabelled GM content even if they grow much farther than the required separation distance from a GM source. We also bring attention to the observed phenomenon of a long ‘tail’ in the distribution of deposited pollen, far beyond any value being considered as a ‘separation distance’ between GM and non-GM crops.