Synthetic Aperture Radar is a low-cost alternative widely employed to estimate soil properties, especially over small map scales (i.e. large area). In contrast to deeper soil layers, topsoil is more consistent with the capability of a C-band SAR signal data (e.g., Sentinel-1) to reach the soil surface. However, few studies address the use of microwave-based sensors to estimate particle size fractions and soil texture classes (e.g., loam, sandy clay). As soil texture consists of the relative proportions of sand, silt, and clay, this soil property is compositional in nature (i.e., the sum of the components is equal to 100%), and such a constraint is not always considered in either explicitly spatial or non-spatial models. Moreover, retrieving information on soils from radar signals is a challenging task, particularly under vegetated soil conditions. This research seeks to address this challenge by employing H-alpha dual-pol decomposition over a study domain located in the Republic of Ireland, where soils are typically covered by grass/pasture. No study to date has employed this method for estimating sand, silt, and clay, or soil texture. Employing both a spatial and non-spatial framework and explicitly considering the compositional nature of soil texture, two different statistical modelling approaches – linear and tree-based – are used to derive soil estimates from Sentinel-1 data, in tandem with topographical and geophysical covariates. Five primary conclusions can be drawn from this work: (i) it is beneficial to treat soil texture as compositional data in the models employed; (ii) radar-based derivatives are not able to predict sand, silt or clay without the aid of covariates, since the models do not identify direct relationships between the backscattering coefficients (� ��0 , � ��0) and the soil particle size fractions; (iii) the non-spatial modelling approaches yield better estimates when fitted without the geophysical covariates, however, the geophysical covariates are useful in obtaining soil texture in the regression models with interactions terms; (iv) the H-alpha Dual-Pol Decomposition method provides, to a certain extent, soil information over low vegetation, and improved estimates of sand, silt and clay; and (v), the spatial models do not outperform the non-spatial models in estimating soil particle size fractions (PSF) in terms of numerical estimates, but are useful in capturing patterns and trends in the response variables. Hence, this research provides a methodological framework to inform the design of in situ soil surveys and a means to estimate soil properties over large spatial areas to generate new data products for use in hydrological, land surface, climate, and other model-based approaches that currently employ coarse global scale soil texture products. This research is also timely in light of the European and Irish policy initiatives around soils such as “A Soil Deal for Europe 2021-2030 (Mission Soil)” and “A Signpost for Soil Policy in Ireland 2021-2030”.