**1. Introduction**

Vegetation phenology is the seasonal timing of lifecycle events, such as leaf emergence, flowering, leaf coloration and fall, and it has become an important topic in the field of climate and ecology as a sensitive and precise indicator that is responsive to climate warming [1,2]. Shifts in spring and autumn vegetation phenology caused by climate warming can differentially alter the length of the growing season, which affects carbon, water, and energy exchange between terrestrial ecosystems and the atmosphere [3–5]. Recent studies have reported that in addition to climatic factors, soil and biological factors also influence shifts in vegetation phenology by affecting plant growth processes in the context of ongoing global climate change [6,7], due to the poor interpretation of phenology shifts among different vegetation types [8,9]. Hence, it is essential to study the dynamics

**Citation:** Guo, J.; Liu, X.; Ge, W.; Ni, X.; Ma, W.; Lu, Q.; Xing, X. Specific Drivers and Responses to Land Surface Phenology of Different Vegetation Types in the Qinling Mountains, Central China. *Remote Sens.* **2021**, *13*, 4538. https:// doi.org/10.3390/rs13224538

Academic Editors: Xuanlong Ma, Jiaxin Jin, Xiaolin Zhu, Yuke Zhou and Qiaoyun Xie

Received: 9 October 2021 Accepted: 8 November 2021 Published: 11 November 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

and drivers of phenology among different vegetation types to improve phenology models and enrich our understanding of the carbon cycle of terrestrial ecosystem.

With the application of remote sensing in monitoring vegetation phenology, we traditionally use the term land surface phenology (LSP) to denote the dynamic variations in vegetation land surface as observed from satellite imagery [10]. Satellite-derived LSP metrics are usually focused on the start (SOS) and end (EOS) of growing seasons [11]. Satellite-based studies have shown that SOS was advanced by 10.6 days (i.e., 5.4 days per decade) throughout Europe and by 14 days (i.e., 7.9 days per decade) in temperate China before 2000 [12,13]. However, this trend of SOS advancement may have slowed or even reversed since the 2000s. For instance, in the entire northern hemisphere, it was advanced by only 0.2 days during 2000 to 2008, but a delayed SOS was revealed in the Tibetan Plateau [14,15]. Regarding satellite-derived EOS, the published results have not always been consistent. Across the entire Northern Hemisphere, EOS was delayed at a rate of 2.2 days per decade during 2000–2008 [14]. In the Yellow River Basin, EOS was delayed by 5.6 and 3.4 days in 1982–1999 and 2000–2015, respectively [16]. In the Qinghai-Tibet Plateau, however, Wang et al. [17] reported the opposite phenology change trends, in both the east and west zones. Overall, these varied results might be due to different study areas, periods, and methods of extracting phenology metrics. However, few studies have focused on the diversity of phenology across different vegetation types. In particular, the dynamic phenology characteristics of herbaceous or shrubs and evergreen forests in the Qinling Mountains have been little studied.

To date, the processes and drivers governing LSP remain poorly understood. Several studies reported that temperature is a major driver of early spring leaf development and delayed autumn leaf fall in plants and has less control over autumn phenology than spring phenology [18,19]. The impact of precipitation on LSP processes is mainly directed at plants in arid and semiarid regions, where water deficits limit the use of light and heat conditions by plants in arid and semiarid areas [20]. Some studies further considered solar radiation (i.e., shortwave radiation) and found that increasing photosynthetic active radiation can promote earlier leaf germination and delaying leaf senescence [21]. Besides meteorological factors, soil factors and biological factors have also been shown to be drivers of LSP change processes [22–24]. Soil temperature and moisture information, due to the prevalent freezing—Thawing process of soil in alpine and arctic regions, could more directly control vegetation growth; for example, soil wetting will, to some extent, reduce the effect of soil warming on LSP changes [22,25]. Slight fluctuations in the time interval between the middle date (MD) of the growing season and the autumn phenology have a strong effect on regulating EOS [7]. Peak growth in summer (i.e., maximum NDVI during the growing season, MN) can have an impact on vegetation greening and senescence, and its unique vegetation growth patterns may result in different allocations of green or carbon across the growing season [24]. To date, how these meteorological, soil, and biological factors affect regional-scale LSP variations has not been clearly and consistently studied, which seriously affects our ability to predict LSP periods.

The responses of vegetation growth processes to climate change are inherently nonlinear [26]. Random forest (RF), as a nonparametric multivariate method, can explain nonlinear processes to a large extent [27]. The advantage of the RF model is that it can consider many predictor variables and nonlinearly determine the relative importance of each predictor variable [28]. Due to its high efficiency in handling the potentially complex relationships between LSP periods and meteorological, soil, and biological factors, RF has been widely and successful applied in recent ecology studies [6,29].

The Qinling Mountains (QMs), rich in vegetation types, represent a demarcation line of climate in China and are also an area characterized by sensitivity or response to climate change, with a significant upward temperature trend in the past half century [30]. Here, we used the satellite derived normalized difference vegetation index (NDVI) records (2001–2019) from MOD13A2 to extract the LSP dates of QMs. The objectives were (a) to explore the temporal and spatial trends of LSP in the QMs, (b) to quantify the relative

contribution of SOS and EOS of different vegetation types to the length of growing season (LOS) and to determine the dominant growth pattern during the growing season, and (c) to simulate the LSP dates and assess the relative importance of meteorological, soil, and biological factors on the interannual variations in LSP. This study focuses on the specificity of different vegetation growths in the QMs, and the results are helpful for future accurate prediction of vegetation growth and to develop scientific managemen<sup>t</sup> strategies.

#### **2. Materials and Methods**
