There are a great deal of obvious issues about effects of scale and scaling conversion that need to be solved for agricultural condition monitoring using remote sensing. In this paper, a serial of researches about them were performed so as to service practical requirements of agricultural condition remote sensing monitoring, especially in the dynamic monitoring of crops growth using remote sensing technology, such as suitability of satellite images on different spatial resolution (i.e. spatial scales); associated relationships of some same kind of remote sensing information on different spatial scales in terms of time changes; a set of corresponding spatial-and-temporal associating models developed; and their effectiveness relatively assessed. As for validating the above, a case study was provided in Hengshui region of North China Plain, using multi-temporal satellite imageries of MODIS (Moderate-resolution Imaging Spectroradiometer, spatial resolutions of 500m & 250m) aboard EOS and AVHRR (Advanced Very High Resolution Radiometer, a spatial resolution of 1km) aboard NOAA, wherein the spectral characteristic values such as NDVI (Normalized difference vegetation index) were retrieved and applied in the corresponding spatial resolutions (spatial scales) of the satellite images. Practically, the last results showed that the model and approach of remote sensing scale and scaling conversion were effective and available in the actual crop growing remote sensing monitoring, while preserving informational integrity of satellite imageries, and hence they are able to be more widely applied and further integrated into agricultural condition remote sensing monitoring systems. 1.

The experiments were conducted for monitoring crop area by spatial sampling methods in a winter wheat main production region with a size of 42Km×42Km in Hengshui City, Hebei Province in China to optimize designs of the frameworks and elements (sample size, sample-square dimension), based on Remote Sensing (RS) and Geographical Information System (GIS) techniques. 5 sample-square dimension levels (3000m×3000m, 2000m×2000m, 1000m×1000m, 500m×500m and 300m×300m) were selected and 3 sampling techniques (simple random sampling, systematic sampling and stratified sampling) were applied. The experimental results demonstrate that the sampling efficiency by stratified sampling was the maximal (the average relative error was 0.15%, the average sample size varied from 9 to 10); and that of the systematic sampling was inferior (the average relative error varies from 0.74%~2.06%, the average sample size is 229); the sampling efficiency by simple random sampling was the minimum (the average relative error is 2.04%, the average sample size is 229) among 3 sampling techniques. Sampling relative error decreased with sample-square dimension simultaneously. When the sample-square dimension was reduced to a certain extent (the size of sample is 500m×500m), the error was not declining any more. The sampling relative error was the minimum using the sample-square with a size of 500m×500m among 5 sample-square dimension levels. 1.

www.mdpi.com/journal/remotesensing