Autumn Phenological Grid Dataset of Typical
Deciduous Broad-leaved Woody Plants in the
Warm Temperate Zone of China

Zhu, M. Y.^1,2  Dai, J. H. ^1,2,*  Tao, Z. X.^1,2  Wang, H. J.¹  Liu, H. L.¹  
Dong, X. Y.³  Hu, Z.^1,2

1. Key Laboratory of Land Surface Pattern and Simulation, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China;

2. University of Chinese Academy of Sciences, Beijing 100049, China;

3. School of Earth Science and Resources, Chang??an University, Xi??an 710054, China

Abstract: Plant autumn phenology is affected by many environmental factors, such as temperature, water, and daylight. Because of the sensitivity to environmental factors, changes in the timing and distribution of autumn phenology is more complicated and may be important indicators of global changes. In this paper, the leaf coloring date (LCD) of five woody plants (Fraxinus chinensis Roxb. (Oleaceae), Salix babylonica L. (Salicaceae), Robinia pseudoacacia L. (Leguminosae), Ulmus pumila L. (Ulmaceae Mirb.), Armeniaca vulgaris Lam. (Rosaceae)) from 1963?C2015 was obtained from the Chinese Phenological Observation Network and used to establish and compare three LCD models (multiple regression, temperature?Cphotoperiod, and spring-influenced autumn). The 0.5?? ?? 0.5?? LCD grid data of the five species from 1963 to 2015 were developed by simulation and scale expansion. The cross-validation results showed that the average simulation error of the LCD was approximately 10?C18 days. The autumn phenological grid data set of China is the basic data representing the spatial?Ctemporal pattern and changes in autumn phenology in China over the last 50 years. The data set includes three parts: header file, phenophase, and spatial distribution, which consist of three data folders and 541 data files. The data is stored in TXT, GEOTIFF, and ARCGIS ASCII formats. The data volume is 26.8 MB before compression (4.9 MB after compression).

Keywords: autumn phenology; woody plants; leaf coloring date; China

1 Introduction

Plant autumn phenology is affected by both internal factors (e.g., growth hormones, age of reproductive development) and external environmental factors (e.g., temperature, water, daylight, nutrient deficiency) ^[1]. Compared with spring phenology, the changes are more complicated, and the mechanism of autumn phenology has not been fully explored. In the absence of in situ observation data, phenological data can be interpolated by models of autumn phenology to provide data support for comprehensive studies on plant geography and global change^[2]. In our previous paper, three models were established to examine spatial?Ctemporal patterns of leaf coloring date (LCD) for three woody plants in China, and the results provided a better understanding of the autumn phenological process and its response to climate change^[3]. Currently, great uncertainties remain in how to observe and model autumn phenology in China. Large systematic errors are associated with observations because of the difficulty in identifying the degree of autumn leaf coloring, and because of the complex relationship between autumn phenology and environmental factors, extensive verification of autumn phenological models is lacking^[4]. Therefore, in this study, the applicability of three different models for the simulation and scale expansion of autumn phenology of five woody plants was evaluated. This paper was prepared to publish the systematic development methodology and the basic results of a spatial?Ctemporal distribution grid data set of the autumn phenology of woody plants in China.

2 Metadata of Dataset

The metadata of the autumn phenological grid data set of five typical deciduous broad-leaved woody plants in the warm temperate zone of China^[5] are summarized in Table 1. It includes the data set full name, short name, authors, year of the data set, temporal resolution, spatial resolution, data format, data size, data files, data publisher, and data sharing policy.

Table 1  Metadata summary of the autumn phenology data set of five typical deciduous broad-leaved woody plants in the warm temperate zone of China from 1963?C2015

(To be continued on the next page)

(continued)

3 Methods

3.1 Basic Data Collection

The LCD of five well-observed typical deciduous broad-leaved woody plants was derived from the Chinese Phenological Observation Network (CPON, www.cpon.ac.cn). The five species were Fraxinus chinensis Roxb. (Oleaceae), Salix babylonica L. (Salicaceae), Ro­binia pseudoacacia L. (Leguminosae), Ulmus pumila L. (Ulmaceae Mirb.), Armeniaca vul­garis Lam. (Rosaceae). Species spatial distribution data were obtained from the Atlas of Woody Plants in China: Distribution and Climate^[7]. The daily air temperature data of both station and grid were obtained from the China Meteorological Data Service Center (data.cma.cn), which were used for model parameterization and scale expansion, respectively.

3.2 Algorithm Principle

The autumn phenological models are mostly based on different assumptions on the response of plants to environmental factors such as temperature and photoperiod^[8]. Compared with spring phenological models, autumn phenological models include more drivers, such as photoperiod and spring phenophase, and thus, the process of parameter estimation is more complicated. In this study, three autumn LCD models were tested: multiple regression (MR) model^[8], which considers the influence of temperature in different months; temperature?Cphotoperiod (TP) model^[9,10]; and spring-influenced autumn (SIA) model^[11], which considers the combined influence of temperature and photoperiod. The functional forms and structures of the relationship between phenophase and meteorological factors are different in these models.^[12]

(1) In the MR model, the influence of average temperature on autumn LCD is different in different months. Increasing temperatures in May and June may lead to the advance of LCD, whereas increasing temperatures in August and September may lead to the delay of LCD.^[8] A multiple regression model (eq. (1)) was established on the basis of the correlations (R₅?CR₉) between LCD (P_l) and average temperature from May to September (T₅?CT₉), where a, b, c, d, and e are model coefficients and ?? is a constant term.

                                (1)

(2) The TP model assumes that the autumn LCD is affected by both temperature and photoperiod.^[7] When the photoperiod is lower than the threshold P_start, the cold state CDD(d) starts to accumulate (eq. (2)). When the accumulated iCDD(d) exceeds the threshold Y_crit, the day d is the exact date of leaf coloring (Y_mod, eq. (3)). The daily cold state CDD(d) is codetermined by daily temperature T(d) and daily photoperiod P(d) (eq. (4, 5)).

                                              (2)

                                                                                        (3)

                                       (4)

                                                                (5)

The TP model includes five parameters: P_start, Y_crit, T_b, x, and y, where x and y are 0, 1, or 2. x = 0 or y = 0 indicates that the LCD is independent of temperature or photoperiod; x = 1 or y = 1 indicates that the LCD is linearly correlated with temperature or photoperiod; x = 2 or y = 2 indicates a nonlinear correlation with temperature or photoperiod.

(3) In the SIA model, which is based on the TP model, the influence of spring phenology on autumn phenology is considered^[11]. The model assumes that the threshold Y_crit at the beginning of the LCD in autumn is linearly correlated with spring phenological departure S_a (eq (6)). The SIA model has six parameters: P_start, T_b, x, y, a, and b.

                                                                                                      (6)

3.3 Technical Route

The steps in the development of the grid data include the building of phenological models, the comparison of models, and simulation and scale expansion. First, the autumn LCD models were built using the LCD data from CPON and the station temperature data from the China meteorological data website. Next, the three LCD models were validated via internal validation using the LCD in odd years and via cross validation using the LCD in even years. The optimal LCD model was determined by the minimum root mean square error (RMSE). Finally, the LCD grid data of five woody plants in China were developed by simulation and scale expansion based on the optimal LCD model and the grid temperature data from the China meteorological data website.

4 Results and Validation

4.1 Data Composition

The autumn phenology grid data set consists of header file, phenophase, and spatial distribution (Table 2). Each of the five species folders contains a total of 106 LCD files from 1963 to 2015. Data files are stored in visual GEOTIFF format and ARCGIS ASCII standard format. The LCD data outside the range of a species spatial distribution need to be masked.

Table 2  Composition of the autumn phenological grid data set of five typical deciduous broad-leaved woody plants in the warm temperate zone of China from 1963?C2015

4.2 Results

The results of autumn phenophase simulation showed that the LCD of all species was spatially different depending on latitude and altitude, with the latest LCD in the eastern and southern regions in China (Figure 2). The time distribution of each species indicated that the average LCD of Armeniaca vulgaris (Figure 2e) was relatively early, whereas that of Robinia pseudoacacia (Figure 2c) was relatively late. The LCD time range was also different. The range in LCD was larger in Fraxinus chinensis, Salix babylonica, and Ulmus pumila than that in the other species. A linear regress between LCD and year was used to determine the trend of autumn phenology. The results of the trend analysis of the LCD time series (Table 3) showed that the LCD was delayed in 83.7%?C99.5% of the pixels for most species, except for Ulmus pumila. Among the other species, the delaying trends were most distinct in Robinia pseudoacacia and Armeniaca vulgaris, with 83.7% and 81.6%, respectively, of the pixels showing significant delay. The LCD of Ulmus pumila was advancing in 70.8% pixels, but only 3.2% of the pixels the advancing trends were significant (p < 0.05). In general, except for Ulmus pumila (−0.4 d/10a), the LCD of all species showed a distinct delaying trend over the last 50 years, with an average trend of 0.5?C0.8 days per decade.

Figure 2  Average leaf coloring date (LCD) of five typical deciduous broad-leaved woody plants in the warm temperate zone of China from 1963 to 2015. Panels (a)?C(e) represents mean LCD of each species. DOY represents day of year refer to a Julian calendar.

Table 3  Statistics of the trend in variation of autumn phenological time of five typical deciduous broad-leaved woody plants in the warm temperate zone of China from 1963 to 2015

4.3 Data validation

The validation results of the three autumn LCD models are shown in Figure 3. Internal validation results showed that the MR model was the best fit to the LCD of Salix babylonica, Robinia pseudoacacia, and Armeniaca vulgaris (r² = 0.34?C0.51; RMSE = 10.89?C15.62 days). The SIA model was most suitable for Fraxinus chinensis (r² = 0.52; RMSE = 10.45 days) and Ulmus pumila (r² = 0.64; RMSE = 11.71 days). The cross-validation results showed that the MR model had the highest simulation accuracy for Fraxinus chinensis, Salix babylonica, and Armeniaca vulgaris (RMSE = 9.76?C18.23 days), whereas the SIA model best simulated the LCD of Robinia pseudoacacia and Ulmus pumila (RMSE = 11.49 and 11.05 days, respectively). Overall, the simulation of the autumn phenology data was less accurate than that of the spring phenology data,^[13] and the average simulation error was approximately 10?C18 days.

Figure 3  Internal and cross validation of three phenological models for the simulation of the leaf coloring date (LCD) of five typical deciduous broad-leaved woody plants in the warm temperate zone of China from 1963 to 2015. Panels (a)?C(j) represents internal and cross-validation on LCD of each species. DOY represents day of year refer to a Julian calendar.

Note: The circle, cross, and triangle represent the simulation results of the MR, TP, and SIA models, respectively. Solid lines are linear regression curves fit to observed and predicted values. The number after the symbol and the number in parentheses represent the r² and root mean square error, respectively, of the model. The root mean square minimum value represents the simulation result of the optimal phenological model.

5 Discussion and Conclusion

In this paper, the methodology and research results of autumn phenological data from the observations of CPON were systematically reviewed, and the LCD time series data set of five typical deciduous broad-leaved woody plants in the warm temperate zone in China from 1963?C2015 is presented for publication. The time resolution is one year, and the spatial resolution is 0.5?? ?? 0.5??. The average simulation error of the autumn LCD for each species was approximately 10?C18 days, which is a greater error than that obtained for the simulation of spring leaf unfolding date or first flowering date^[13]. Research in North America^[10] Europe^[9], and Asia^[14] based on a single site also shows great differences in the simulation accuracy of autumn models. In these previous studies of different species, the absolute error of simulation of the LCD varies from 1 to 13 days. Therefore, the accuracy of the simulation of the autumn phenology data set in this paper was moderate and was within the range of previous measurements. Furthermore, the data set shows the spatial?Ctemporal distribution as well as the changes in the autumn phenology of woody plants in China. This data set can provide the basis for a better understanding of the response of autumn phenology to global climate change.

Author Contributions

Dai Junhu was responsible for the overall design and development of the data set. Dong Xiaoyu and Hu Zhi collected and processed autumn phenological data. Tao Zexing and Wang Huanjiong designed the model and the algorithm. Zhu Mengyao and Dong Xiaoyu performed data verification. Zhu Mengyao and Dai Junhu wrote the data set papers.

References

[1]       Lim, P. O., Kim, H. J., Nam, H. G. Leaf senescence [J]. Annual Review of Plant Biology, 2007, 58: 115?C136.

[2]       Cleland, E. E., Chuine, I., Menzel, A., et al. Shifting plant phenology in response to global change [J]. Trends in Ecology & Evolution, 2007, 22(7): 357?C365.

[3]       Tao, Z. X., Wang, H. J., Dai, J. H., et al. Modeling spatiotemporal variations in leaf coloring date of three tree species across China [J]. Agricultural and Forest Meteorology, 2018, 249: 310?C318.

[4]       Dai, J. H., Tao, Z. X., Wang, H. J., et al. Applications of phenological models in quantitative climate reconstruction [J]. Quaternary Sciences, 2016, 36(3): 703?C710.

[5]       Zhu, M. Y., Dai, J. H., Liu, H. L., et al. Autumn phenological grid data set of typical deciduous broad-leaved woody plants in the warm temperate zone of China [DB/OL]. Change Research Data Publishing & Repository, 2019. DOI: 10.3974/geodb.2019.06.06.V1.

[6]       GCdataPR Editorial Office. GCdataPR data sharing policy [OL]. DOI: 10.3974/dp.policy.2014.05 (Updated 2017).

[7]       Fang, J. Y., Wang, Z. H., Tang, Z. Y. Altas of woody plants in China: distribution and climate [M]. Beijing: Higher Education Press, 2009.

[8]       Estrella, N., Menzel, A. Responses of leaf colouring in four deciduous tree species to climate and weather in Germany [J]. Climate Research, 2006, 32(3): 253?C267.

[9]       Delpierre, N., Dufrene, E., Soudani, K., et al. Modelling interannual and spatial variability of leaf senescence for three deciduous tree species in France [J]. Agricultural and Forest Meteorology, 2009, 149(6-7): 938?C948.

[10]    Jeong, S. J., Medvigy, D. Macroscale prediction of autumn leaf coloration throughout the continental United States [J]. Global Ecology and Biogeography, 2014, 23(11): 1245?C1254.

[11]    Keenan, T. F., Richardson, A. D. The timing of autumn senescence is affected by the timing of spring phenology: implications for predictive models [J]. Global Change Biology, 2015, 21(7): 2634?C2641.

[12]    Tao, Z. X. Simulation of the distributions of five widespread tree species in China under the influence of phenological changes [D]. Beijing: University of Chinese Academy of Sciences, 2018.

[13]    Dai, J. H., Zhu, M. Y., Wang, H. J., et al. Spring phenological grid data set of typical deciduous broad-leaved woody plants in the warm temperate zone of China [DB/OL]. Global Change Research Data Publishing & Repository, 2019. DOI: 10.3974/geodb.2019.06.04.V1.

[14]    Park, C. K., Ho, C. H., Jeong, S. J., et al. Spatial and temporal changes in leaf coloring date of Acer palmatum and Ginkgo biloba in response to temperature increases in South Korea [J]. PLoS ONE, 2017, 12(3): e0174390.