Authors:L. Barkaszi, A. Arutyunjan and K. Takács-György
It has been proved that because of the different past of the parcels regarding their soil, agronomical and technological parameters, weed sampling results may not be generalised. Therefore it is necessary to study those solutions how to determine on an acceptable confidence level a parcel’s weed infestation with optimised sampling techniques.For studying the question we have delimited on wheat stubble a total sample area of 36×54 metres (using it as reference) and divided it into 2×2 cells giving a total of 486 sample cells. Then we surveyed the weed infestation and GPS recorded the location of each cell.We have analysed the weed infestation data with mathematical and statistical methods comparing the results of cells with each other and with the total sample area. We found that in several cases of different sample cells weed infestation displayed a diverse picture. This way sampling of weeds is extremely difficult.We found close relation between relative frequency of weeds and sampling accuracy. Therefore sampling is reliable only for surveying the frequent weeds in a parcel, while more rarely found weeds (e.g. spots of perennials) are to be scouted only by means of going over the parcel and GPS recording them. Otherwise, in the case of a traditional sampling process, the number of sampling cells required for acceptable reliability is unnecessarily high.Consequently, it is necessary to further study the economic and cost efficiency aspects of the needed weed sample density from the point of view of reasonable sample density, accuracy and optimal yield.