Smoothing metrics such as RMSE and absolute error work well for smooth variables like temperature but are unsuitable for highly skewed variables like precipitation, making it difficult for models to predict sporadic events such as heat waves, cyclones, and extreme precipitation.