Vegetation phenology, specifically the onset of greening (OG) in spring and the onset of dormancy (OD) in autumn, is a sensitive indicator of climate change. Global warming has generally led to earlier OG in the Northern Hemisphere, supported by ground measurements and satellite data. However, the impact of warming on OD is less clear, with some studies reporting delayed OD and others reporting earlier OD. Delayed OD might be due to increased enzyme activity, decreased chlorophyll degradation, reduced frost risk, or increased growth capacity; while earlier OD can result from limited leaf life or reduced productivity due to water deficits. Previous studies showed that earlier OG and increased evapotranspiration (ET) from 1982-2011 led to summer soil moisture deficits and earlier OD in the Northern Hemisphere. However, the relative roles of summer soil moisture deficits, reduced precipitation, and increased atmospheric water demand (AWD) in determining OD remain unclear. Increased AWD, reflected in vapor pressure deficit (VPD), can disrupt the soil-plant-atmosphere continuum, causing stomatal closure and earlier OD, even with sufficient soil moisture. This study aimed to investigate long-term trends in OG, OD, and growing season length (GSL) in Europe from 1982 to 2020; to determine if meteorological variables and soil moisture can explain these trends; and finally, to elucidate the roles of water and energy supply in GSL changes.

Numerous studies have explored the impacts of climate change on vegetation phenology. While earlier onset of greening is well-documented across the Northern Hemisphere, the effects on the onset of dormancy are less consistent. Some studies report delayed autumn dormancy due to factors such as increased temperature leading to higher photosynthetic activity and reduced chlorophyll degradation. Conversely, others have observed earlier dormancy, linked to limitations on leaf lifespan or water stress later in the growing season. The relationship between evapotranspiration, soil moisture, and the onset of dormancy has also been investigated, with some studies suggesting that earlier spring greening can lead to increased evapotranspiration and subsequent soil moisture deficits, resulting in earlier dormancy. However, the interplay between increased atmospheric water demand (reflected in vapor pressure deficit) and soil moisture availability in determining the timing of dormancy requires further investigation. Prior research has shown lengthening growing seasons in Europe in earlier decades but the influence of more recent decades needs further scrutiny.

This study utilized a novel logistic function derivative of NDVI (LFD-NDVI) method along with long-term satellite-based NDVI data (GIMMS, AVHRR, and MODIS) to determine OG and OD, and subsequently GSL, for Europe from 1982 to 2020. The data were preprocessed to achieve consistency, accounting for differences in spatial and temporal resolutions. A scaling approach, using GIMMS as a benchmark due to its established quality, adjusted OG and OD from MODIS and AVHRR to match the GIMMS data, considering different land cover types. The Mann-Kendall test, with AR(1) correction for temporal autocorrelation, assessed the significance of trends in GSL, OG, and OD. The Group Method of Data Handling (GMDH) was used for a pixel-wise analysis to identify the main control factors for OG, OD, and GSL anomalies, using meteorological variables (temperature, precipitation, VPD) and soil moisture (surface and root zone) data from various reanalysis datasets (GLDAS, ERA5-Land, GLEAM). Data from these datasets were validated against in-situ FLUXNET measurements. Principal component analysis (PCA) and biplotting further investigated the relationships between the control factors and the phenological events.

The analysis revealed a significant trend towards earlier greening across 35% of the European pixels from 1982 to 2020, aligning with previous findings of early spring greening. However, the rate of advancement in greening slowed in recent years. In contrast to earlier studies reporting delayed OD, this study found that in 17% of pixels OD occurred earlier in the recent decade (2011–2020). About two thirds of pixels showed no significant trend in GSL (remaining constant at approximately 185 days). A significant lengthening trend in GSL was observed in only 28% of pixels, whereas a shortening of GSL occurred in only 7% of pixels. This contrasts with studies that reported a greater lengthening trend in GSL during earlier periods. Analysis using a 15-year moving window revealed a trend reversal after 2003-2004, potentially linked to a severe European drought in 2003. While GSL lengthening continued until 2013-2014 (though at a slower rate), shortening commenced afterwards. The GMDH analysis identified temperature and vapor pressure deficit in late summer, as well as temperature in early spring and spring, as the most important control factors for GSL. Temperature and VPD in early spring were the most influential on OG, while temperature and VPD in late summer were the most significant for OD. Soil moisture showed minimal influence on GSL, OG, and OD, suggesting that water availability was not the primary limiting factor for plant growth, in contrast to some previous hypotheses. PCA further confirmed that temperature and VPD in summer and late summer primarily drove earlier OD, while precipitation had a positive effect, delaying OD. Spatial analysis of control factors revealed that temperature, VPD, and precipitation influenced early greening, with VPD dominant in coastal areas, precipitation in Central Europe, and temperature elsewhere. For early dormancy, temperature and VPD were the main drivers. Different land use types exhibited varied responses to these drivers. For example, earlier OD was mainly influenced by croplands and grasslands, likely due to their shallow root systems.

This study challenges previous findings suggesting continually lengthening growing seasons in Europe. The observed recent shortening of GSL can be mainly attributed to the increased atmospheric water demand during summer months, which limited transpiration and resulted in earlier onset of dormancy even when soil moisture levels remained similar to previous years. This highlights the increasing importance of energy limitations (as captured by VPD) alongside water availability in regulating vegetation phenology under a changing climate. The minimal role of soil moisture in this study's results for Europe contrasts with some earlier hypotheses that emphasized soil moisture deficits as the primary driver of shifts in dormancy. The findings emphasize that temperature and atmospheric water demand are crucial factors influencing vegetation dynamics in Europe, warranting a reevaluation of management strategies that rely on historical relationships between soil moisture and phenological patterns.

This research provides comprehensive evidence for a recent trend reversal in European growing season length, shifting from lengthening to shortening, mainly driven by increased atmospheric water demand, reflected in VPD, during summer. Soil moisture was found to be a less dominant factor than previously believed. The findings highlight the urgent need to incorporate energy limitations alongside water availability into future climate change impact assessments and ecosystem management strategies. Further research should investigate the implications of this trend reversal for various aspects of ecosystem functioning, such as carbon cycling, biodiversity, and agricultural production, and explore the specific effects across diverse vegetation types and geographical regions.

This study uses satellite-based NDVI data at a relatively coarse spatial resolution (0.25 degrees), potentially masking finer-scale variations in phenological patterns. This limitation could affect the accuracy of the results, especially in heterogeneous landscapes. The influence of harvest timing on the onset of dormancy in croplands could not be comprehensively assessed due to the coarse spatial resolution. While the reanalysis data used in the study was validated with FLUXNET data, inherent uncertainties associated with remote sensing and reanalysis products might affect the results. Furthermore, the study focuses on the period from 1982-2020, and future research needs to encompass further time to monitor longer-term trends and account for future climatic shifts.