Over the years, countries around the world have been spending a lot to establish their own national weather forecasting systems in order to provide accurate weather forecasts information to their people. Following the revolutions in computational technology and satellites, modern weather forecasts have become significantly predictive about real weathers. This paper then studies the value of accurate weather forecasts today accessible to the general public in many places all around the world. Using the daily maximum temperature forecasts in China, I exploit the spatial and temporal variations of the forecasts accuracy represented as the rolling half-year root-mean-squared-error (RMSE). I use an interactive regression design on the Chinese labor time-use survey data to estimate the differential labor responses to temperature forecasts under varying forecasts RMSE. My analysis shows that more accurate weather forecasts with 1C lower RMSE can lead to significantly more reduction of labor working hours per day by 2.5 hours under extreme hot maximum temperature forecast above 40C and 1 hour under medium-cold forecast around 20C, relative to the reference of 25C. In the end, these leads to the large labor decrease relative to 25C up to 7 hours under heat (40C) or 2.5 hours under medium-cold (20C) that disappears when RMSE increases. Using a single-period utility maximization model, I estimate that the value of 1C decrease in forecasts RMSE is about 1500 2015 Yuan (239 USD) per person per year. The 0.041C yearly average RMSE decrease from 2011 to 2015 can generate a large annual social benefit of at least 54.8 billion 2015 Yuan (8.72 billion USD) from the labor sector alone, enough to cover twice the annual expenditure of the country’s national meteorological administration.