In this study, Surface solar radiation downward (SSRD) estimation and occupational radiation monitoring, both important for environmental planning and radiation protection, are jointly evaluated despite often being assessed separately. Monthly mean SSRD and meteorological variables for 37 stations (2010-2020) were obtained from the NASA Prediction of Worldwide Energy Resources (POWER) database, while occupational annual effective dose records (2012-2016) were obtained from Nigerian Nuclear Regulatory Authority (NNRA) monitoring reports. Random Forest (RF) regression was applied and evaluated against linear regression (LR) using stratified train-test splitting and five-fold cross-validation to analyze the spatial and temporal variability of SSRD across Nigeria's major climatic zones and assess the annual effective dose for regulated worker categories using a machine-learning modeling approach. SSRD showed spatial variability of higher mean in the Sahel (6.5 ± 0.74 kWh m$^{-2}$ day$^{-1}$), moderate mean in the Savannah (5.8 ± 0.68 kWh m$^{-2}$ day$^{-1}$), and lower mean values along the Guinea Coast (5.2 ± 0.62 kWh m$^{-2}$ day$^{-1}$). RF achieved an improved performance for SSRD (R$^2$ = 0.761, RMSE = 0.420 kWh m$^{-2}$ day$^{-1}$, MAE = 0.334 kWh m$^{-2}$ day$^{-1}$) relative to LR. Although annual effective doses were low overall, higher mean doses were observed in Nuclear Medicine and Radiological Research. RF predictive modelling outperformed LR, with a stronger dose prediction of R$^2$ = 0.872. The findings suggest that predictive modeling may support environmental radiation assessment and occupational dose monitoring by improving the understanding of spatial and temporal variability in radiation-related conditions, especially in settings with limited observational data.