Change in spring phenology is a sensitive indicator of ecosystem response to climate change, and exerts first‐order control on the ecosystem carbon and hydrological cycles. The start of season (SOS) in spring can be estimated from satellite data using different spatiotemporal scales, data sets and algorithms. To address the impacts of these differences on trends of SOS, a Bayesian analysis is applied to investigate the rate of SOS advance and whether that advance has slowed down or changed abruptly over the last three decades.

30°–75° N in the Northern Hemisphere.

We applied four algorithms to three different satellite data sets (AVHRR, Terra‐MODIS and SPOT) to obtain an ensemble of SOS estimates. A Bayesian analysis was applied to test different hypotheses of SOS trends.

Over the period 1982–2011, SOS is likely (74%) to have experienced a significant advance best described by a linear trend (−1.4 ± 0.6 days decade⁻¹). At hemispheric and continental scales, deceleration or abrupt changes in the SOS trend are unlikely (< 30%) to have occurred. Trend analysis restricted to the last decade suggests no significant SOS advance since 2000. This lack of trend can be explained by large interannual variations of SOS and uncertainties in SOS extraction, in the context of a short‐term decadal‐period analysis. Spatial analyses show that SOS advance could have slowed down over parts of western North America, and the SOS trend could have abruptly changed over parts of Canada and Siberia.

SOS advance is unlikely to have slowed down or changed abruptly at a hemispheric scale over the last three decades. At a regional scale, SOS advance could have slowed down or abruptly changed due to changes in winter chilling or fire regimes. Trends of SOS derived from different satellites were within the uncertainties of SOS extraction.