The junction temperature of red (AlGaInP), green (GaInN), blue (GaInN), and ultraviolet (GaInN) light-emitting diodes (LEDs) is measured using the temperature coefficients of the diode forward voltage and of the emission-peak energy. The junction temperature increases linearly with DC current as the current is increased from 10 mA to 100 mA. For comparison, the emission-peak-shift method is also used to measure the junction temperature. The emission-peak-shift method is in good agreement with the forward-voltage method. The carrier temperature is measured by the high-energy-slope method, which is found to be much higher than the lattice temperature at the junction. Analysis of the experimental methods reveals that the forward-voltage method is the most sensitive and its accuracy is estimated to be ± 3°C. The peak position of the spectra is influenced by alloy broadening, polarization, and quantum confined Stark effect thereby limiting the accuracy of the emission-peak-shift method to ±15°C. A detailed analysis of the temperature dependence of a tri-chromatic white LED source (consisting of three types of LEDs) is performed. The analysis reveals that the chromaticity point shifts towards the blue, the color-rendering index (CRI) decreases, the color temperature increases, and the luminous efficacy decreases as the junction temperature increases. A high CRI > 80 can be maintained, by adjusting the LED power so that the chromaticity point is conserved.