Temporal autocorrelation refers to the correlation of a signal with a delayed version of itself over time, indicating that observations close in time are more similar than those further apart. It is crucial in time series analysis as it affects the assumptions of statistical models, potentially leading to biased estimates if not properly accounted for.