Crop height is a crucial indicator in understanding crop growth, providing insights into developmental stages that are essential for precision agriculture. Sentinel-1’s capability to acquire images regardless of atmospheric conditions makes it ideal for monitoring these dynamics, though modelling wheat height across its growth cycle remains challenging due to structural changes in the plant. This research aimed to effectively capture and predict the height variability in wheat phenology, using clustering as a method to segment data by growth stage for stage-specific analysis. This study was conducted across three wheat fields in Umbria, Italy, from January 30^th to June 10^th 2024, where in-field measurements of plant height and phenology were carried out while acquiring twelve Sentinel-1 images. For image processing, two distinct speckle filters (Lee 7x7 and Refined Lee) were applied and several radar-derived variables were extracted, including VH, VV, CR, Entropy, Anisotropy, Alpha, and the Radar Vegetation Index (RVI). Visual analysis of the variables in relation to plant height suggested clusters within the dataset, which were confirmed through fuzzy C-means clustering. This approach successfully separated data into two phenological groups, allowing us to implement multiple linear regression models tailored to each cluster. The results highlight strong model performance in the early growth stages (from tillering to stem elongation) for both filters (R² 0.76 and RMSE 6.88 for the Lee 7x7 filter; R² 0.79 and RMSE 6.35 for the Refined Lee), while in later stages (from booting to maturity) model performance declined, with Lee 7x7 achieving a higher result (R² 0.51, RMSE 9.33) than the Refined Lee (R² 0.33, RMSE 10.89). These findings provide promising insights into height prediction in the early growth stages of wheat that are crucial for optimizing management strategies and ensuring high yields.