Analysis of sampling precision in low-density weed populations
2022
Hamouz, Pavel | Hamouzová, Kateřina
Site-specific weed management (SSWM) can provide significant herbicide savings but requires the reliable detection of weed populations. In this work, the effect of sample size on sampling precision was studied in four sub-field areas of winter cereals infested with Galium aparine. Plots of 1170 m², 4212 m², 2592 m² and 2880 m² were marked and the positions of all G. aparine individuals were recorded using a precise navigation system. Mapping cells of 3 m × 3 m were arranged over experimental data and then various sampling intensities of G. aparine populations were simulated. The relative root mean square error, mean absolute percentage error (MAPE), and false negative rate were calculated and plotted against individual sampling intensities. Regressions were fitted to non-linear models. Experimental plots had relatively low mean densities of G. aparine (1.00–2.98 plants m⁻²) and the spatial distributions of plants were patchy in all cases. The values of all error measures decreased with increasing sampling intensity. On the basis of the fitted model, keeping MAPE below 30% required the sampling of 32.2% of the total area for the low-density population and 13.4% for the highest-density population. The results presented demonstrate that reliable sampling of low-density populations requires a very high sampling intensity and could be helpful in the design of reliable sampling strategies for SSWM.
显示更多 [+] 显示较少 [-]