Vegetation plays a critical role in evaluating ecosystems, particularly in drylands. Medium resolution satellite imagery, high-resolution drone data and physical structure data provide effective tools for both long-term monitoring and rapid assessment. This study developed an URG framework by integrating U-net, Random Forest, and Genetic Algorithm to upscale and evaluate the spatial distribution of floodplain shrubs and trees in drylands from drone imagery (~ 1 cm), Sentinel-2 imagery (10 m) and a global Canopy Height Model (CHM) (30 m). Gwydir, Macquarie Marshes, Lachlan, Middle Murray and Murrumbidgee, across the Murray-Darling Basin (the MDB) in Australia, were selected as five study areas. Target vegetation includes the common and ecologically important floodplain shrub Duma florulenta (lignum), and tree species, e.g., Eucalyptus camaldulensis (river red gum). Drone RGB imagery and published CNN-derived classifications were used as references for the accuracy validation of the URG framework. The results show the URG framework achieved overall accuracy (OA) from 0.5 to 0.80 across the MDB, categorized into low (OA < 60%), moderate (60% OA < 80%) and high-accuracy (OA > 80%) classification groups, which illustrates an OA decline from the north-east region (about 80%) to south-west region (< 60%) across the MDB, reflecting strong positive relationships between lignum cover and classification accuracy (R² = 0.8023). This is directly related to the lower percentage cover and density of vegetation, decreased water availability and rainfall, highlighting the importance of hydrological conditions in shaping vegetation cover and structure in drylands. The sparse cover and mixed structural layer of vegetation cover in drylands increased uncertainty, resulting in lower OA (< 60%). Sensitivity analysis and spatial autocorrelation analysis demonstrate the robustness of the URG framework with higher composite optimization score (0.47). Overall, the URG framework provides an operationally relevant and transferable tool for supporting ecosystem evaluations at a basin scale.