| 英文摘要: | The timing of phenological events exerts a strong control over ecosystem function and leads to multiple feedbacks to the climate system1. Phenology is inherently sensitive to temperature (although the exact sensitivity is disputed2) and recent warming is reported to have led to earlier spring, later autumn3, 4 and increased vegetation activity5, 6. Such greening could be expected to enhance ecosystem carbon uptake7, 8, although reports also suggest decreased uptake for boreal forests4, 9. Here we assess changes in phenology of temperate forests over the eastern US during the past two decades, and quantify the resulting changes in forest carbon storage. We combine long-term ground observations of phenology, satellite indices, and ecosystem-scale carbon dioxide flux measurements, along with 18 terrestrial biosphere models. We observe a strong trend of earlier spring and later autumn. In contrast to previous suggestions4, 9 we show that carbon uptake through photosynthesis increased considerably more than carbon release through respiration for both an earlier spring and later autumn. The terrestrial biosphere models tested misrepresent the temperature sensitivity of phenology, and thus the effect on carbon uptake. Our analysis of the temperature–phenology–carbon coupling suggests a current and possible future enhancement of forest carbon uptake due to changes in phenology. This constitutes a negative feedback to climate change, and is serving to slow the rate of warming. Changes in phenology greatly affect the carbon balance of terrestrial ecosystems. Warmer springs, for example, stimulate an early emergence from winter dormancy, leading to an extension of an ecosystem’s carbon uptake period10. Warmer autumns, on the other hand, are thought to lead to carbon losses from ecosystems due to a greater increase in respiration than photosynthesis4, 9. Global mean temperatures have risen over the past decades11. It is therefore imperative to develop a robust understanding of both the temperature sensitivity of phenology2, and the associated changes in carbon cycling1. Given the importance of phenology to the earth system, and the recent changes in global temperatures, much attention has been focused on the detection of climate-induced trends in phenology. Long-term ground observations of phenology have shown an increase in growing season length12, 13, 14, 15. Independent studies based on satellite reflectance corroborate this evidence, showing an earlier spring and later autumn in temperate and boreal forests5, 16, 17, 18, 19. However, the long-term impacts of changes in phenology on temperate forest carbon uptake and storage have yet to be quantified at the regional scale. Here we report multi-decadal phenological trends in temperate forests in the eastern US, and quantify the subsequent impact on regional carbon cycling. We combine three different remote sensing greenness indices (daily MODIS enhanced vegetation index (EVI), normalized difference vegetation index (NDVI), and green chromatic coordinate (GCC)), two date extraction techniques, and the MODIS land cover dynamics product20, with two decades of ground observations of individual tree phenology, and measurements of CO2 exchange between forests and the atmosphere at seven long-term research sites. Across all scales (organism, ecosystem, landscape) we detect a consistent trend of earlier spring and later autumn over the past two decades. We derive the temperature sensitivity of spring and autumn phenology, and show how it can be used to improve the representation of seasonality by land surface models. Using the observed ecosystem–atmosphere carbon exchange, we quantify the impact that both interannual variability and long-term changes in phenology are having on forest photosynthesis and respiration, and consequently CO2 uptake. Spatially coherent trends of earlier spring phenology were evident across the different remotely sensed measures of greenness we analysed (Fig. 1). Spring phenology advanced on average by 0.48 ± 0.2 d yr−1 (P < 0.01; panel analysis) for the period 2001–2012. The detection of false but statistically significant phenological trends is possible using individual remote sensing metrics21, 22. The magnitude of the spring trends we detect is largely independent of the metric used (Fig. 1 and Supplementary Fig. 1), and matches changes recorded in ground observations (see below), thus enhancing our confidence in these results.
|
