Over the past 50 years, global land surface air temperature has been rising at a much higher rate at night than during the day. Understanding plant responses to the asymmetric daytime and nighttime warming in the context of climate change has been a hot topic in global change biology and global ecology. It has been debatable whether the asymmetric warming has opposite effects on vegetation activity (e.g., phenology, productivity). Here we settle the debate by scrutinizing the underpinnings of different statistical methods and revealing how the misuse or improper use of these methods could mischaracterize the effects of asymmetric warming with in situ and satellite observations. The use of the ordinary least square (OLS) methods including both daytime (Tmax) and nighttime (Tmin) temperature in the multiple regression models could overlook the multicollinearity problem and yield the misinterpretations that Tmax and Tmin had opposite effects on spring phenology, autumn phenology, gross primary production (GPP), and normalized difference vegetation index (NDVI). However, when the OLS methods were applied with Tmax and Tmin included in separate models or alternatively the ridge regression (RR) method with properly selected ridge parameter was used, the effects of Tmax and Tmin on vegetation activity were generally in the same direction. The use of the RR method with improperly selected ridge parameter could also mischaracterize the effects of asymmetric warming. Our findings show that daytime and nighttime warming has no opposite effects on vegetation phenology and productivity in the northern hemisphere, and properly dealing with the multicollinearity problem is critical for understanding the effects of asymmetric warming on vegetation activity.
