Spatial and temporal variation of phenological growing season and climate change impacts in temperate eastern China

Using phenological and normalized difference vegetation index (NDVI) data from 1982 to 1993 at seven sample stations in temperate eastern China, we calculated the cumulative frequency of leaf unfolding and leaf coloration dates for deciduous species every 5 days throughout the study period. Then, we determined the growing season beginning and end dates by computing times when 50% of the species had undergone leaf unfolding and leaf coloration for each station year. Next, we used these beginning and end dates of the growing season as time markers to determine corresponding threshold NDVI values on NDVI curves for the pixels overlaying phenological stations. Based on a cluster analysis, we determined extrapolation areas for each phenological station in every year, and then implemented the spatial extrapolation of growing season parameters from the seven sample stations to all possible meteorological stations in the study area. Results show that spatial patterns of growing season beginning and end dates correlate significantly with spatial patterns of mean air temperatures in spring and autumn, respectively. Contrasting with results from similar studies in Europe and North America, our study suggests that there is a significant delay in leaf coloration dates, along with a less pronounced advance of leaf unfolding dates in different latitudinal zones and the whole area from 1982 to 1993. The growing season has been extended by 1.4–3.6 days per year in the northern zones and by 1.4 days per year across the entire study area on average. The apparent delay in growing season end dates is associated with regional cooling from late spring to summer, while the insignificant advancement in beginning dates corresponds to inconsistent temperature trend changes from late winter to spring. On an interannual basis, growing season beginning and end dates correlate negatively with mean air temperatures from February to April and from May to June, respectively.     

中国东部温带植被生长季节的空间外推估计

利用地面植物物候和遥感归一化差值植被指数（NDVI）数据,以及一种物候-遥感外推方法,实现植被生长季节从少数站点到较多站点的空间外推。结果表明：（1）在1982～1993年期间,中国东部温带地区植被生长季节多年平均起讫日期的空间格局与春季和秋季平均气温的空间格局相关显著;（2）在不同纬度带和整个研究区域,植被生长季节结束日期呈显著推迟的趋势,而开始日期则呈不显著提前的趋势,这与欧洲和北美地区植被生长季节开始日期显著提前而结束日期不显著推迟的变化趋势完全不同;（3）北部纬度带的植被生长季节平均每年延长1．4～3．6d,全区的植被生长季节平均每年延长1．4d,与同期北半球和欧亚大陆植被生长季节延长的趋势数值相近;（4）植被生长季节结束日期的显著推迟与晚春至夏季的区域性降温有关,而植被生长季节开始日期的不显著提前则与晚冬至春季气温趋势的不稳定变化有关;（5）在年际变化方面,植被生长季节开始和结束日期分别与2～4月份平均气温和5～6月份平均气温呈负相关关系。     

Effects of precipitation and soil water potential on drought deciduous phenology in the Kalahari

We utilized an ecosystem process model to investigate the influence of precipitation and soil water potential on vegetation phenology in the semi‐arid, drought‐deciduous ecosystems in the Kalahari region of South Africa. The timing of leaf flush was assumed to be the first day during which a rainfall event exceeded that day's estimate of potential evapotranspiration after a defined dry season. Leaf senescence was assumed to be a dynamic feedback between soil water potential and net plant carbon gain and was determined by dynamically modeling the effects of concomitant trends in soil water potential and net primary production on leaf area index (LAI). Model predictions of LAI were compared with satellite‐derived normalized difference vegetation indices (NDVI) for 3 years at two sites along the Kalahari transect. The mean absolute error for the prediction of modeled leaf flush date compared with leaf flush dates estimated from NDVI were 10.0 days for the Maun site and 39.3 days for the Tshane site. Correlations between model predicted 10‐day average LAI and 10‐day composite NDVI for both Maun and Tshane were high (ρ=0.67 and 0.74, respectively, P<0.001), suggesting that this method adequately predicts intra‐annual leaf area dynamics in these dry tropical ecosystems.     

Remote sensing of larch phenological cycle and analysis of relationships with climate in the Alpine region

This research aims at developing a remote sensing technique for monitoring the interannual variability of the European larch phenological cycle in the Alpine region of Aosta Valley (Northern Italy) and to evaluate its relationships with climatic factors. Phenological field observations were conducted in eight test sites from 2005 to 2007 to determine the dates of completion of different phenological phases. MODerate Resolution Imaging Spectrometer (MODIS) 250 m 16‐days normalized difference vegetation index (NDVI) time series were fitted with double logistic curves and the dates corresponding to different features of the curves were determined. Comparison with field data showed that the features of the fitted NDVI curve that allowed the best estimate of the start and end of the growing season were the zeroes of its third derivative (MAE of 6 and 4 days, respectively). The start and end of season were also estimated with the spring warming (SW) and growing season index (GSI) phenological models. MODIS start and end of season dates generally agreed with those obtained by the SW and GSI climate‐driven phenological models. However, phenological models provided erroneous results when applied in years with anomalous meteorological conditions. The relationships between interannual variability of the larch phenological cycle and climate were investigated by comparing the mean start and end of season yearly anomalies with air temperature anomalies. A strong linear relationship (R²=0.91) was found between mean spring temperatures and mean start of season dates, with an increase of 1 °C in mean spring temperature leading to a 7‐day anticipation of mean larch bud‐burst date. Leaf coloring dates were found to be best related with mean September temperature (R²=0.77), but with higher spring temperatures appearing to lead to earlier leaf coloring.     

Simulating phenological shifts in French temperate forests under two climatic change scenarios and four driving global circulation models

After modeling the large-scale climate response patterns of leaf unfolding, leaf coloring and growing season length of evergreen and deciduous French temperate trees, we predicted the effects of eight future climate scenarios on phenological events. We used the ground observations from 103 temperate forests (10 species and 3,708 trees) from the French Renecofor Network and for the period 1997–2006. We applied RandomForest algorithms to predict phenological events from climatic and ecological variables. With the resulting models, we drew maps of phenological events throughout France under present climate and under two climatic change scenarios (A2, B2) and four global circulation models (HadCM3, CGCM2, CSIRO2 and PCM). We compared current observations and predicted values for the periods 2041–2070 and 2071–2100. On average, spring development of oaks precedes that of beech, which precedes that of conifers. Annual cycles in budburst and leaf coloring are highly correlated with January, March–April and October–November weather conditions through temperature, global solar radiation or potential evapotranspiration depending on species. At the end of the twenty-first century, each model predicts earlier budburst (mean: 7 days) and later leaf coloring (mean: 13 days) leading to an average increase in the growing season of about 20 days (for oaks and beech stands). The A2-HadCM3 hypothesis leads to an increase of up to 30 days in many areas. As a consequence of higher predicted warming during autumn than during winter or spring, shifts in leaf coloring dates appear greater than trends in leaf unfolding. At a regional scale, highly differing climatic response patterns were observed.     

Determining the growing season of land vegetation on the basis of plant phenology and satellite data in Northern China

The objectives of this study are to explore the relationships between plant phenology and satellite-sensor-derived measures of greenness, and to advance a new procedure for determining the growing season of land vegetation at the regional scale. Three phenological stations were selected as sample sites to represent different climatic zones and vegetation types in northern China. The mixed data set consists of occurrence dates of all observed phenophases for 50–70 kinds of trees and shrubs from 1983 to 1988. Using these data, we calculated the cumulative frequency of phenophases in every 5-day period (pentad) throughout each year, and also drew the cumulative frequency distribution curve for all station-years, in order to reveal the typical seasonal characteristics of these plant communities. The growing season was set as the time interval between 5% and 95% of the phenological cumulative frequency. Average lengths of the growing season varied between 188 days in the northern, to 259 days in the southern part of the research region. The beginning and end dates of the surface growing season were then applied each year as time thresholds, to determine the corresponding 10-day peak greenness values from normalized difference vegetation index curves for 8-km² pixels overlying the phenological stations. Our results show that, at the beginning of the growing season, the largest average greenness value occurs in the southern part, then in the northern, and finally the middle part of the research region. In contrast, at the end of the growing season, the largest average greenness value is measured in the northern part, next in the middle and lastly the southern part of the research region. In future studies, these derived NDVI thresholds can be applied to determine the growing season of similar plant communities at other sites, which lack surface phenological data. Received: 29 November 1999 / Revised: 14 March 2000 / Accepted: 15 March 2000     

Exploring two methods for statistical downscaling of Central European phenological time series

In this study we set out to investigate the possibility of linking phenological phases throughout the vegetation cycle, as a local-scale biological phenomenon, directly with large-scale atmospheric variables via two different empirical downscaling techniques. In recent years a number of methods have been developed to transfer atmospheric information at coarse General Circulation Model's grid resolutions to local scales and individual points. Here multiple linear regression (MLR) and canonical correlation analysis (CCA) have been selected as downscaling methods. Different validation experiments (e.g. temporal cross-validation, split-sample tests) are used to test the performance of both approaches and compare them for time series of 17 phenological phases and air temperatures from Central Europe as microscale variables. A number of atmospheric variables over the North Atlantic and Europe are utilized as macroscale predictors. The period considered is 1951–1998. Temporal cross-validation reveals that the CCA model generally performs better than MLR, which explains 20%–50% of the phenological variances, whereas the CCA model shows a range from 40% to over 60% throughout most of the vegetation cycle. To show the validity of employing phenological observations for downscaling purposes both methods (MLR and CCA) are also applied to gridded local air temperature time series over Central Europe. In this case there is no obvious superiority of the CCA model over the MLR model. Both models show explained variances from 40% to over 70% in the temporal cross-validation experiment. The results of this study indicate that time series of phenological occurrence dates are very compatible with the needs of empirical downscaling originally developed of local-scale atmospheric variables.     

Divergent phenological response to hydroclimate variability in forested mountain watersheds

Mountain watersheds are primary sources of freshwater, carbon sequestration, and other ecosystem services. There is significant interest in the effects of climate change and variability on these processes over short to long time scales. Much of the impact of hydroclimate variability in forest ecosystems is manifested in vegetation dynamics in space and time. In steep terrain, leaf phenology responds to topoclimate in complex ways, and can produce specific and measurable shifts in landscape forest patterns. The onset of spring is usually delayed at a specific rate with increasing elevation (often called Hopkins' Law; Hopkins, 1918), reflecting the dominant controls of temperature on greenup timing. Contrary with greenup, leaf senescence shows inconsistent trends along elevation gradients. Here, we present mechanisms and an explanation for this variability and its significance for ecosystem patterns and services in response to climate. We use moderate‐resolution imaging spectro‐radiometer (MODIS) Normalized Difference Vegetation Index (NDVI) data to derive landscape‐induced phenological patterns over topoclimate gradients in a humid temperate broadleaf forest in southern Appalachians. These phenological patterns are validated with different sets of field observations. Our data demonstrate that divergent behavior of leaf senescence with elevation is closely related to late growing season hydroclimate variability in temperature and water balance patterns. Specifically, a drier late growing season is associated with earlier leaf senescence at low elevation than at middle elevation. The effect of drought stress on vegetation senescence timing also leads to tighter coupling between growing season length and ecosystem water use estimated from observed precipitation and runoff generation. This study indicates increased late growing season drought may be leading to divergent ecosystem response between high and low elevation forests. Landscape‐induced phenological patterns are easily observed over wide areas and may be used as a unique diagnostic for sources of ecosystem vulnerability and sensitivity to hydroclimate change.     

High‐resolution prediction of leaf onset date in Japan in the 21st century under the IPCC A1B scenario

Reports indicate that leaf onset (leaf flush) of deciduous trees in cool‐temperate ecosystems is occurring earlier in the spring in response to global warming. In this study, we created two types of phenology models, one driven only by warmth (spring warming [SW] model) and another driven by both warmth and winter chilling (parallel chill [PC] model), to predict such phenomena in the Japanese Islands at high spatial resolution (500 m). We calibrated these models using leaf onset dates derived from satellite data (Terra/MODIS) and in situ temperature data derived from a dense network of ground stations Automated Meteorological Data Acquisition System. We ran the model using future climate predictions created by the Japanese Meteorological Agency's MRI‐AGCM3.1S model. In comparison to the first decade of the 2000s, our results predict that the date of leaf onset in the 2030s will advance by an average of 12 days under the SW model and 7 days under the PC model throughout the study area. The date of onset in the 2090s will advance by 26 days under the SW model and by 15 days under the PC model. The greatest impact will occur on Hokkaido (the northernmost island) and in the central mountains.     

Regional unified model‐based leaf unfolding prediction from 1960 to 2009 across northern China

Using first leaf unfolding data of Salix matsudana, Populus simonii, Ulmus pumila, and Prunus armeniaca, and daily mean temperature data during the 1981–2005 period at 136 stations in northern China, we fitted unified forcing and chilling phenology models and selected optimum models for each species at each station. Then, we examined performances of each optimum local species‐specific model in predicting leaf unfolding dates at all external stations within the corresponding climate region and selected 16 local species‐specific models with maximum effective predictions as the regional unified models in different climate regions. Furthermore, we validated the regional unified models using leaf unfolding and daily mean temperature data beyond the time period of model fitting. Finally, we substituted gridded daily mean temperature data into the regional unified models, and reconstructed spatial patterns of leaf unfolding dates of the four tree species across northern China during 1960–2009. At local scales, the unified forcing model shows higher simulation efficiency at 83% of data sets, whereas the unified chilling model indicates higher simulation efficiency at 17% of data sets. Thus, winter temperature increase so far has not yet significantly influenced dormancy and consequent leaf development of deciduous trees in most parts of northern China. Spatial and temporal validation confirmed capability and reliability of regional unified species‐specific models in predicting leaf unfolding dates in northern China. Reconstructed leaf unfolding dates of the four tree species show significant advancements by 1.4–1.6 days per decade during 1960–2009 across northern China, which are stronger for the earlier than the later leaf unfolding species. Our findings suggest that the principal characteristics of plant phenology and phenological responses to climate change at regional scales can be captured by phenological and climatic data sets at a few representative locations.