**3. Results**

### *3.1. Wheat, Processing Tomato, and Cotton Height, LAI, and Kc Models Based on the σ<sup>0</sup> Normalization Method*

The effect of the proposed normalization (Equation (2)) on the SAR backscatter from two incidence angles is illustrated in Figure 6. Following the normalization process, the difference is greatly reduced, and a considerable improvement in the R<sup>2</sup> and RMSE of all the σ0-based models is observed. In wheat, processing tomatoes, and cotton, the height, LAI, and Kc models' R<sup>2</sup> improved in the range of 0.0172–0.367, and the RMSE improved in the range of 5–52%. Table 3, Tables S4–S6, and Figure 7 show the performance of σ0-based height, LAI, and Kc models in wheat, processing tomatoes, and cotton.

**Figure 6.** A time series of VV polarization values recorded by Sentinel-1 on its ascending overpasses with local incidence angles of 36.5◦ and 47.7◦ during the wheat experiment in Saad: (**A**) prior to applying the σ0 normalization; (**B**) post-normalization.

**Table 3.** R<sup>2</sup> and RMSE improvements following the local incidence angle normalization procedure. Significance is marked by \*. The percentage values in the brackets show improvement in RMSE after normalization (i.e., the reduction in prediction error).


**Figure 7.** Models based on field measurements and the σ0 normalization method: (**A**) wheat height; (**B**) wheat LAI; (**C**) wheat Kc; (**D**) processing tomato height; (**E**) processing tomato LAI; (**F**) processing tomato Kc; (**G**) cotton height; (**H**) cotton Kc.

*3.2. Processing Tomato and Cotton Height, LAI, and Kc Models Based on the β0 Normalization Method*

The effect of the β0-based normalization (Equation (8)) that reduces the difference in β0 images acquired at different angles is shown in Figure 8.

**Figure 8.** A time series of VV polarization values recorded by Sentinel-1 on its descending overpasses with local incidence angles in the ranges of 30.8◦–31.2◦ and 42.2◦–43.6◦ during the processing tomato experiment in Gadash, 2018: (**A**) prior to applying the β0 normalization; (**B**) post-normalization.

The β0-based normalization method permitted achieving the improvement in the R<sup>2</sup> and RMSE of all the β0-based models. For the processing tomato and cotton height, LAI, and Kc models, the R<sup>2</sup> improved in the range of 0.1143–0.668, and the RMSE improved in the range of 18–44%. Table 3, Tables S7 and S8, and Figure 9 show the performance of processing tomato and cotton β0-based height, LAI, and Kc models. Table 3 shows the performance of all the σ0-based and β0-based normalized models developed in this study and their R<sup>2</sup> and RMSE improvements over the non-normalized models.

**Figure 9.** Models based on field measurements and the β0 normalization method: (**A**) processing tomato height; (**B**) processing tomato LAI; (**C**) processing tomato Kc; (**D**) cotton height; (**E**) cotton Kc.

### *3.3. Performance of the Dual-Polarized RVI*

The dual-polarized RVI-based wheat, processing tomato, and cotton models are shown in Table 4. The wheat and cotton RVI models were compared against the σ0-based models, while the processing tomato RVI models were compared to the β0-based processing tomato models.

**Table 4.** RVI models for wheat, processing tomato, and cotton height, LAI, and Kc. The differences in R<sup>2</sup> and RMSE indicate the difference in performance compared to models based on either the σ0 (wheat and cotton models) or β0 (processing tomatoes) local incidence angle normalization methods in Table 3. Negative values represent lower R<sup>2</sup> and higher RMSE of the RVI models.

