SAMPLING DESIGNS OVER TIME BASED ON SPATIAL VARIABILITY OF IMAGES FOR MAPPING AND MONITORING SOIL EROSION COVER FACTOR

摘要

In the Revised Universal Soil Loss Equation, cover factor reflects the effect of ground and vegetation covers on the reduction of soil loss and it controls change of soil erosion for a specific area. Developing optimal sampling designs over time for data collection of this factor is thus critical to monitor the dynamics of soil erosion. In this study we developed an image-inferred semivariogram based method to determine optimal sample size. We further explored spatial and temporal variability, and the change of sample sizes needed over time for this cover factor. In addition, we studied application of historical ground data and uncertainties to infer semivariograms by combining the Landsat thematical mapper (TM) images to determine sample sizes. Compared to the results using ground data, the semivariogram and its dynamics of the cover factor could be successfully inferred using the multi-temporal TM images. The accuracy of sample sizes obtained using the image-inferred semivariograms could meet the requirement for regional estimation, but for local estimation for mapping it was very much dependent on the quality and correlation of the images with the factor. Moreover, historical ground data should be used with great caution for sampling design.