Quantifying uncertainty: The benefits of removing snow cover from remote sensing time series on the extraction of climate-influenced grassland phenology on the Qinghai-Tibet Plateau

Vegetation phenology is a sensitive indicator of climate change. Remote sensing technology is often utilized to estimate vegetation phenology over expansive regions, but it contains uncertainties stemming from methods of data preprocessing and phenology extraction. Previous studies have primarily focused on individual causes of uncertainty, neglecting potential interactions between various causes and the consequent effects on analyzing vegetation's response to climate change. In this study, the uncertainty introduced by multiple time-series processing procedures in extracting the start of the growing season (SOS) of grassland on the Qinghai-Tibetan Plateau (QTP) was assessed using remote sensing data. This includes data preprocessing (time-series smoothing and snow cover removal) and phenology extraction. How these uncertainties might influence the analysis of SOS response to climate change was further quantified. Our findings demonstrated that the average uncertainty in grassland SOS extraction on the QTP for the period 2001-2015 reached 20.01 d, and snow-induced noise emerged as the principal factor affecting SOS estimates. Temporal trends of SOS revealed a regional average uncertainty of 1.06 d/y, both snow cover removal and the smoothing method were influential, each introducing uncertainty of 1.04 d/y and 1.05 d/y, respectively. After the removal of snow cover, uncertainties were reduced by 35 % for the SOS temporal trend. SOS response to climate change was extracted by various methods, leading to divergent and at times conflicting outcomes. The uncertainty for SOS sensitivity to temperature was averagely 10.50 d/degrees C on the QTP, while that for SOS sensitivity to precipitation was 1.75 d/mm. As the duration of snow cover lengthened, uncertainty increased in SOS response to temperature, while uncertainty increased in arid regions for SOS response to precipitation. The insights provided here contribute to enhancing phenology extraction accuracy and understanding the interactions between global climate change and vegetation phenology.