T e r r a  D e w Research Project
 
 

Home

Contact

Overview

Study site

Experimental data

Results

Publications

Links

Effect of Plant Surface Wetness on Radar Backscatter

An essential task towards image analysis is an improved knowledge on the effect of free vegetation water on the radar signal. The examinations in the framework of this study indicated significant diurnal variation in radar backscatter for cereals and high grassland in the morning after the rain event only. For the other crops cultivated in the study area the diurnal variations do not exceed 1 dB. Indeed, most of them were still very small and sparse in the middle of June. Maybe later in the season the radar signal will be affected by plant surface wetness, especially for corn. The impact on cereals and grassland differ with respect to growth stage, vegetation structure, frequency and polarization (see Figures). Greatest differences in the effect of plant surface wetness on the radar backscatter with respect to various crop types were found at C-band. Green, full-developed cereals show an decrease in radar backscatter in the morning after the rain event, whereas ripe cereals tend to show an opposite trend. For all other frequencies the radar signal rose due to the presence of plant surface wetness, especially at X-VV and L-HV. For single fields changes in radar backscatter up to 4 dB were found. Regarding the second day of the campaign no changes were obtained for all crops. However, the amount of dew deposited on the plant surfaces in the morning of June 16, 2000 was very small. Maybe, under more favourable conditions for dew accumulation there will be an impact on the radar images.

Short time changes in the radar signal as observed in this study could be generally attributed to variations of following factors: structural characteristics, soil moisture, plant water content, and plant surface wetness. During the field campaign no significant changes in vegetation geometry were recognized except for grassland. Due to the heavy rainfall at night the grass was blown down in the morning of June 14. At the time of the third SAR data acquisition it was back in a vertical position. Thus, only for grassland an impact of structural changes have to be considered. Furthermore, the analysis of the groundtruth data indicate no significant changes in soil moisture conditions during the whole campaign. Regarding the impact of varying plant water contents on the radar backscatter former studies documented only little changes of less than 1 dB. Thus, this factor is expected to have only little influence on the radars signal. Consequently, the observed changes in radar backscatter obtained in this study could be most probably related to changes in plant surface wetness.

Effect of Plant Surface Wetness on Crop Recognition

Regarding the thematic analysis of SAR images affected by intercepted rainfall the investigations indicated, that the overall classification accuracies (see Table) as well as the user’s and producer’s accuracies of single crops change. With respect to all observed frequencies, the strongest impact of intercepted rainfall on crop recognition was detected for C-band. Especially the separability of corn and rape decreases significant due to moisture. For X- and C-band former studies ascertained, that the separability of wheat and barley increases under wet conditions. These results were not confirmed by this study. At C-band the separability of cereals was very low in general. At X-band high classification accuracies were found particularly for summer barley. Due to plant surface wetness an increased mix-up with wheat was found. The investigations indicated higher classification accuracies for summer barley and wheat at midday, whereas for summer rye an opposite trend was found. L-band data showed lowest variations in the classification accuracies of different crops during the day. One typical problem arising under wet conditions is an increased mix-up between rape and high grassland fields. In consequence, classification accuracy of rape decreased significantly due to moisture. This phenomenon was detected at all observed bands and band combinations to a more or less extent. With respect to future TerraSAR applications it is important to notice, that for the combination of X- and L-band data higher classification accuracies were achieved when plant surface water was present. All cereals and rape fields were mapped more accurate under wet conditions.

By combining SAR data acquired under wet and dry conditions, the classification result could be improved significantly. For all band combinations both the number of separable classes and the overall classification accuracy (except C-band) increased (see Table). The figure shows the classification result for the E-SAR data as well as for the TerraSAR-simulations when incorporating X- and L-band information in the classification process. Table below illustrates the derived confusion matrix for the E-SAR data. Moreover, for all crop types the changes in classification accuracy due to the combination of radar scenes acquired under different moisture conditions is shown. In particular, significant changes in classification accuracy were found for crops, which showed high variations in separability due to interception (ripe cereals, wheat and rape).



[Home] [Contact] [Overview] [Study site] [Experimental data] [Results] [Publications] [Links]