TL;DR
Current Earth system models struggle to accurately predict the timing of spring leaf onset in the Northern Hemisphere, largely because they rely on daily mean temperature. This study demonstrates that daytime temperature (Tmax) is a significantly stronger driver of leaf onset than nighttime temperature (Tmin) or mean temperature (Tmean). This finding suggests a crucial refinement for phenology models, potentially improving predictions of carbon cycle feedbacks and ecosystem responses to climate change.
What Problem Does This Paper Address?
Phenology, the study of the timing of biological events, is a critical indicator of climate change impacts. In the Northern Hemisphere, spring leaf onset has been advancing, a trend observed in both satellite data and ground-based measurements. However, current dynamic global vegetation models (DGVMs) that simulate this phenomenon often exhibit significant errors. These models typically use daily mean temperature (Tmean) to calculate the accumulation of heat (like Growing Degree Days or GDDs) required for leaf unfolding. The problem is that global warming is not uniform; nighttime temperatures are increasing faster than daytime temperatures. This disparity means that using Tmean might be an increasingly inaccurate proxy for the actual thermal cues plants use to initiate spring growth. The paper aims to identify which temperature metric (Tmax, Tmin, or Tmean) is the most accurate predictor of interannual variations in leaf onset and to propose a more effective modeling approach.
Methods
The researchers employed a two-pronged approach using both in situ observations and satellite data.
- In Situ Observations: They analyzed leaf unfolding dates (LUDs) for numerous plant species across Europe and for lilac shrubs in the United States, spanning from 1982 to 2011. To determine the most influential temperature period, they calculated "preseason" lengths (0-3 months before LUD) that showed the strongest negative partial correlation with LUD. Crucially, they used partial correlation analysis to isolate the effect of Tmax and Tmin while controlling for other climate variables like precipitation and cloudiness. This allowed them to compare the independent influence of daytime versus nighttime temperatures. They also calculated the "temperature sensitivity" of LUD to Tmax and Tmean using multiple linear regressions.
- Satellite Observations: To assess the generality of their findings beyond specific ground sites, they analyzed satellite-derived vegetation green-up dates (VGDs) for the Northern Hemisphere (>30°N) over the same 1982-2011 period. They used multiple VGD algorithms to ensure robustness and applied similar partial correlation and sensitivity analyses as with the in situ data.
Datasets
- In Situ Phenology Data:
- European Pan European Phenological Database (PEP725): Data from 2,400 phenological sites and 24 plant species across Europe.
- USA National Phenology Network (USA-NPN): Data from 35 phenological sites for lilac shrubs.
- Satellite Vegetation Data:
- NDVI3g data set (1982-2011) from the Global Inventory Modeling and Mapping Studies (GIMMS) group. They also used MODIS NDVI data for a shorter period (2000-2010) for comparison.
- Climate Data:
- CRU TS 3.20 (Climate Research Unit): Monthly averaged daily maximum (Tmax) and minimum (Tmin) temperatures, precipitation, and cloudiness at a 0.5° spatial resolution.
- WATCH Forcing Data Methodology to ERA-Interim reanalysis data (WFDEI): Used for future climate scenario modeling.
Key Insights and Ideas
- Daytime Dominance: The most striking insight is that daytime temperature (Tmax) is a significantly stronger driver of interannual variations in spring leaf onset than nighttime temperature (Tmin) or even the commonly used daily mean temperature (Tmean). This suggests that the heat accumulated during daylight hours is more critical for triggering budburst.
- Model Deficiencies: The study highlights a fundamental limitation in current Earth system models. Their reliance on Tmean, which is increasingly influenced by faster-warming nights, likely leads to inaccurate predictions of phenological shifts.
- Regional Differences: While Tmax is generally dominant, the paper notes regional variations. In temperate dry regions, the relationship between Tmax and green-up is weaker or even positive, potentially due to Tmax-regulated water stress or the influence of Tmin on frost risk.
- GDD Re-evaluation: The concept of Growing Degree Days (GDDs), a cornerstone of phenology modeling, needs re-evaluation. If Tmin often remains below the critical threshold for GDD accumulation, it contributes less to fulfilling the heat requirement for spring green-up, making Tmax a more relevant variable.
- Future Projections: When projecting future changes, models using Tmax predict a smaller advance in leaf onset compared to Tmean-based models, primarily because Tmin is projected to increase faster than Tmax. This implies that Tmean-based models might overestimate the speed of phenological shifts.
Key Results
- In Situ Data: In Europe, 33% of species-site combinations showed a significant negative partial correlation between LUD and Tmax preseason, compared to less than 8% for Tmin. In the US (lilac sites), this was 54% for Tmax versus 14% for Tmin.
- Satellite Data: 42% of the Northern Hemisphere land area showed a significant negative partial correlation between VGD and Tmax preseason, compared to only 11% for Tmin.
- Temperature Sensitivity: An increase of 1°C in Tmax advances LUD by an average of 4.7 days in Europe and 4.3 days in the US. This sensitivity is greater than that estimated using Tmean.
- Ratio of Sensitivities: In 60% of the study area, satellite-derived Tmax sensitivity (SVTmax) was greater than Tmean sensitivity (SVTmean). In Eastern Europe and Siberia, SVTmax was 1-2 times larger than SVTmean.
- Future Projections: For 85% of the Northern Hemisphere, Tmean-based models predicted a larger advance in VGD under future climate scenarios compared to Tmax-based models.
Limitations and Future Work
- Data Resolution and Algorithms: While multiple VGD algorithms and climate datasets were used for robustness, discrepancies between satellite and in situ sensitivities were noted, potentially due to differences in temporal and spatial footprints.
- Other Drivers: The study acknowledges that while temperature is a primary driver, other factors like photoperiod, water availability, and soil conditions can also influence phenology. The analyses controlled for precipitation and cloudiness but did not explicitly include soil moisture or temperature.
- Non-linearity and Synoptic Events: The authors mention that the relationship between temperature and phenology can be non-linear, and unusually warm synoptic events can also be critical. These complexities are not fully captured by simple statistical models.
- Future Work: The paper strongly suggests incorporating daytime temperature (Tmax) into phenology modules of Earth system models to improve their performance. Further research is needed to understand the precise mechanisms of interaction between spring leaf unfolding and daytime temperature, especially through well-designed manipulation experiments. This improved understanding is crucial for more accurate simulations of spring phenology and its impact on the global carbon balance and ecosystem feedbacks.
Summary updated