Accurate estimation of Downward Longwave Radiation (DLR) is essential for climate modeling, hydrology, agriculture, and renewable energy applications, yet the reliability of widely used empirical DLR models across different climatic zones remains uncertain. This study evaluated eleven existing clear-sky DLR empirical models and developed five new locally calibrated regression models for Ibadan and Makurdi, within the Guinea Savannah climatic zone of Nigeria. Forty years (1984-2023) of monthly mean air temperature, relative humidity, and DLR data were used, with thirty-two years for model development and eight years for validation. Model performance was assessed using Mean Bias Error, Root Mean Square Error, Mean Percentage Error, t-statistic, Index of Agreement, and correlation coefficient, followed by a composite ranking procedure. Results showed that, among existing models, Kruk et al., performed best at Ibadan while Idso was most suitable for Makurdi. Among the newly developed models, the modified Guest model gave the highest accuracy at Ibadan, while the modified Dilley and O'Brien model performed best at Makurdi, with all modified models outperforming their existing counterparts. Seasonal analysis revealed that DLR was consistently higher during the rainy season, driven predominantly by atmospheric water vapour and cloud cover rather than temperature. These findings demonstrate that site-specific calibration substantially improves DLR estimation accuracy and provide validated, region-appropriate models recommended for practical application across the Guinea Savannah climatic zone of Nigeria in the absence of direct pyrgeometer measurements.