Hi Mike, that's very interesting! I would say that the different sampling strategies can generate different (but equally interesting) sets of data. Based on the few information you gave, I would stretch the following ideas..

If sampling always the same site, you will be able to determine population variations through time, which is surely extremely useful to understand how pest species develop during the year. On the other hand, sampling different sites (possibly randomized) may result helpful in detecting the presence of recently arrived pest species. This would give you more chances of detecting pest species as soon as they arrive into a field, since you will be covering many different areas each week and not always the same ones. In short, if your forecasting model aims to understand how species spread and grow into a field, I would suggest to keep checking the same area through time. While, if you want to understand how many new species can arrive into the crops, then randomization of the sampling sites could be more useful..

Hope this will help!

Thanks Francesco. Further information:

We use this data to predict future pest numbers, and to recommend possible control actions for farmers to take. Given your answer we would be better to monitor populations at the same sites each sampling session.

Hi Mike,

In case you cannot perform both the samplings, then probably yes, a continuous sampling of the same area can enable a better understanding of the pest population through time (how it grows, how the numbers change seasonally etc etc).

However, focusing uniquely on a few sites, you may risk to "miss" arrivals of new insects/pests, and just realize that once the 'newly arrived pest' has a stable population you cannot eradicate anymore. I suppose your decision should be informed also by the dimension of the farmer's field.. The bigger is the field, the riskier it would be to focus always on the same sites, I think. Unfortunately, you will probably have to deal with time and money spent in this activity and you won't be able to keep performing both the samplings. But if this is not the case, and going back to your original question, I think both the methods are equally 'statistically valid' and both the dataset obtained could be extremely interesting!

Yes I agree with previous colleagues. By sampling same sites for more times, allow to you to better describe the life cycle performances of a more homogeneous genetic population. Specially for the production of forecasting models, it is better focus attention on populations whose metabolic responses to environmental variables, depends on the most similar genetic pool.

Hi Mark,

I agree with colleagues. I think, your predictions relate to some well known species, and if the stage of population dynamic (increasing of number, outbreak, depression) can be determined due to your sampling, you can recommend to farmers the most adequate tools for number control. So, I think, you'd better to sample in the same sites every week.

Hi Mike,

I think using the two strategies are of valuable indications. After a period of time you could compare the two results and make your statistical differentiation.

Regards,

Wafaa