Globally, remotely sensed agricultural monitoring is impeded by cloudy conditions which render the acquired images useless. In semi-arid landscapes, rainfed croplands dominate agricultural production; thus, the majority of planting occurs in the rainy season, characterized by erratic cloud cover. The clear-sky pixels in partly-cloudy images can be used to increase the number of useful observations for quantitative analysis. This is achieved by cloud screening and atmospheric correction (AC) processes. However, the effectiveness of various AC approaches under partly-cloudy conditions is still unknown. Many studies validate surface reflectance (SR) under clear-sky conditions, with only a few focusing on the validation of SR under partly-cloudy conditions. This study sought to validate Sentinel-2 SR products derived from various AC approaches, i.e., MAJA, Sen2Cor, iCor, and FORCE, using in-situ spectral measurements. A partly-cloudy image with >60% cloud cover, acquired over a semi-arid agricultural landscape and ±3 days of the field measurements, was used as a test case. The results showed R2 of <0.2 with in-situ SR in the VIS and SWIR, and R2 of ~0.6 in the Red-edge and NIR regions, attributable to low aerosol scattering effect in the NIR. However, RMSE, MAE, and BIAS error metrics showed consistently higher errors across all AC approaches and crop types, with varying magnitude per spectral band. This finding can be attributed to the high presence of clouds in the image, which enhanced the apparent path radiance, causing an overestimation of the aerosol optical thickness, hence an underestimation of SR. The need for further validation of SR under various cloud conditions emanates from this study, to ascertain the utility of AC approaches under such conditions. Overall, the results have implications for the utility of remotely sensed images for agricultural crop monitoring in partly-cloudy conditions.