**1. Introduction**

The Tibetan Plateau (TP) is situated in the subtropical area of eastern Eurasia and is referred to as "The Third Pole" and the "Roof of the World", exceeding 4000 m above sea level on average. Its dynamic and thermal forcing can affect the occurrence and development of climate and weather in China and East Asia [1–4]. Land–air hydrothermal exchange processes in the TP (such as surface heat sources, atmospheric heat sources, vegetation cover, and snow cover) regulate the thermodynamic forcings of the TP, which have an important effect on the monsoon, Asian atmospheric circulation, as well as global climate [4–9].

**Citation:** Mao, X.; Ren, H.-L.; Liu, G.; Su, B.; Sang, Y. Influence of the Indian Summer Monsoon on Inter-Annual Variability of the Tibetan-Plateau NDVI in Its Main Growing Season. *Remote Sens.* **2023**, *15*, 3612. https:// doi.org/10.3390/rs15143612

Academic Editor: Fernando Camacho

Received: 13 June 2023 Revised: 14 July 2023 Accepted: 17 July 2023 Published: 20 July 2023

**Copyright:** © 2023 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/).

As one of the major elements on the TP land surface, vegetation is critical to land– atmosphere interactions. Local weather and climate change affect vegetation growth [10,11]. Owing to the geography and altitude of the TP, the vegetation on the TP responds to climate change more rapidly than that in other regions at the same latitude. In turn, changes in vegetation can alter the surface properties of the TP, including surface albedo and soil moisture, thus affecting the land–air hydrothermal exchange and carbon cycle over the TP [12,13]. Model experiments showed that the enhanced greening of the TP vegetation induces changes in heat consumption by plant transpiration and surface evaporation and surface heat sources, thus influencing the climate and weather of the TP and its adjacent surroundings [14]. Also, the enhanced vegetation dynamics on the TP may attenuate the local surface warming [15], meaning that thermal conditions over the TP can be affected by changes in the TP vegetation. Therefore, it is essential to understand the characteristics of the TP vegetation in terms of its variability and trends.

Several indices can reflect vegetation activity, such as the normalized difference vegetation index (NDVI), soil-adjusted vegetation index (SAVI), enhanced vegetation index (EVI), leaf area index (LAI), and net primary productivity (NPP). Most of these vegetation indices are derived from remote-sensing images of vegetation; many factors (e.g., atmospheric conditions, soil types, topography, shading effects, and solar angle) may introduce noise into these indices [16]. Of these vegetation indices, the NDVI is calculated as the ratio of the difference between near-infrared (NIR) and red (Red) light to the sum [17]. Due to the ratio of band intensities, the NDVI can eliminate a large proportion of noise caused by instrument calibration, solar angle, topography, cloud shadows, and atmospheric attenuations existing in visible red and infrared bands [18,19], which enhances the response to vegetation and reduces the susceptibleness of illumination conditions [20]. Matsushita et al. [21] also noted that the NDVI may be indirectly affected by topography, which can be somewhat neglected. Moreover, the NDVI seems to have good performance in the TP with complex and fragile ecosystems and vegetation species [22]. Considering the altitude and ecosystem of the study area (i.e., the TP exceeding 3000 m above sea level; Figure 1) in this paper, the NDVI is a suitable vegetation index for indicating the growth activity and cover of vegetation on the TP [23]. Therefore, we chose the NDVI as an indicator to explore the characteristics and trends of vegetation on the TP in the current study.

**Figure 1.** Map showing the location of the study area within the boundary of the Tibetan Plateau (TP) (3000 m above sea level; gray shading).

The NDVI on the entire TP generally exhibits a rising trend under global warming and some human activities [24–27], while certain regions of the TP are suffering from vegetation degradation [28,29]. The apparent inconsistency in the regional and overall NDVI trends on the TP may be due to its relatively distinct climatic characteristics and geographical location. From one perspective, complicated climate change, such as different change trends and intensities of climatic factors, leads to inconsistent vegetation growth trends in different areas of the TP [24,28]. Moreover, when the vegetation on the TP is becoming denser and reaches a certain threshold, the NDVI may no longer increase with the anomalous increase in climatic factors due to the saturation effect of NDVI. From the other perspective, at different altitudes and geographical locations (windward and leeward slopes), vegetation on the TP responds to climatic anomalies in very distinct ways [16,29–31]. For example, the

intensity of solar radiation varies greatly between windward and leeward slopes, which can lead to differences in the amount of water lost from vegetation to the atmosphere due to its evapotranspiration, thus affecting the growth of vegetation [16].

The above studies mainly concentrated on the characteristics and trends of longstanding variations in the TP vegetation rather than its inter-annual variations. Considering the inter-annual variability of climatic factors affecting vegetation growth [18,32–34] and the fact that the inter-annual variability of vegetation can also adjust the TP thermal conditions, the inter-annual characteristics of vegetation on the TP in its growing season deserve exploration. Therefore, this study focuses on the growth of TP vegetation on inter-annual scales. Our previous study revealed that several local climatic factors jointly regulate the inter-annual variability of two NDVI patterns dominating the TP in June–September (JJAS, i.e., the main growing season) [35]. However, it is not enough to merely understand the local factors modulating the inter-annual variability of NDVI on the TP, as the variations in local climatic factors over the TP are inseparable from external influences.

Earlier studies have demonstrated that the complex surface environment and anomalously variable ocean, as well as associated atmospheric teleconnections, can alter hydrothermal conditions over the TP. For example, the Indian summer monsoon (ISM) [36–38], North Atlantic Oscillation [39,40], El Niño-Southern Oscillation (ENSO) [41], and Indian Ocean Basin Mode (IOBM) [42] can all modify precipitation over the TP. Sea surface temperature anomalies in several key oceans [42–45], and Indian soil moisture [46], can modulate the TP thermal conditions. Among these external climatic factors, the ISM seems to be a factor more closely related to the TP. The ISM is essential to the variation in summer precipitation over the TP [37,42,47–50]. Precipitation over the TP can be governed by the deep convection in the Indian subcontinent, which is connected to the ISM [37]. The ISM can also affect precipitation over the southern TP by modulating the transportation of water vapor entering the TP [47,48]. In turn, the ISM onset is directly associated with the TP's atmospheric heat source [51–53]. The TP diabatic heating is pivotal in modulating the location and intensity of the ISM [47,54].

The current study intends to explain what role the ISM plays in the inter-annual variability of vegetation on the TP and whether the vegetation can in turn affect the ISM. This study may be of great practical significance to TP ecological environmental protection and the fields of short-term climate prediction. The following section describes the study period and datasets, index definitions, and analysis methods involved in the study. The findings are introduced in Section 3, where Section 3.1 examines relationships between the ISM, precipitation over the TP, and the NDVI on the TP, and Section 3.2 explores the influence of the ISM on vegetation on the TP in its main growing season. Section 4 (i.e., Discussions) and Section 5 (i.e., Conclusions) explore and sum up the relationship between the ISM and vegetation on the TP, respectively.

#### **2. Data and Methods**

#### *2.1. Data*

#### 2.1.1. Remote-Sensing-Based NDVI Datasets

Remote-sensing images have been widely used in monitoring vegetation dynamics at a regional scale due to their wide coverage and frequent capture of surface information. Based on a comprehensive detailed review related to the use of remote-sensing products (such as LANDSAT, SPOT-Vegetation, Sentinel-2, Himawari-8/9) in the estimation of vegetation growth, we employed two remote-sensing-based NDVI datasets to reflect the variability of the TP vegetation exceeding 3000 m above sea level [35]; the study area is shown in Figure 1. One is GIMMS NDVI3g derived from the National Oceanic and Atmospheric Administration (NOAA), and the other is MCD19A3CMG NDVI of the MODIS products derived from the National Aeronautics and Space Administration (NASA). The former has an 8-km spatial resolution and spans from January 1982 to December 2014, and the latter has a horizontal precision of 0.05◦ × 0.05◦ grid and spans from February 2000 to December 2020.

The GIMMS and MODIS NDVIs have been considered as the more commonly used remote-sensing-based NDVI datasets. This is because as the third generation of AVHRR sensor data, GIMMS NDVI3g has been proven to be a better dataset for describing vegetation dynamics in applications [55]. The GIMMS NDVI3g has a longer time scale and has been extensively applied in regional and global-scale studies of vegetation dynamics and degradation [56,57]. In addition, the MCD19A3CMG is a MAIAC BRDF corrected product in the MODIS sensor datasets, which improves spatial resolution (0.05◦ × 0.05◦ grid), the accuracy of atmospheric correction, aerosol retrievals, and cloud detection [58,59]. Therefore, they are combined to study vegetation activities due to the linear correlation and compatibility between the two datasets [60–62].

#### 2.1.2. Reanalysis and Meteorological Observation Datasets

The monthly geopotential height, zonal/meridional wind, water vapor, and vertical pressure velocity were derived from the European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ERA5) [63], utilizing a reduced horizontal precision of 2.5◦ × 2.5◦ grid. The ERA5 provides a total of 37 vertical pressure levels ranging from 1000 to 1 hPa. In this paper, we used the ERA5 from 1000 to 200 hPa. The monthly outgoing longwave radiation was obtained from NOAA satellite observations [64]. Additionally, the monthly latent and sensible heat flux were derived from the long-term Japanese 55-year Reanalysis (JRA-55) [65], utilizing a horizontal precision of 1.25◦ × 1.25◦ grid. These datasets were used to examine how the ISM influences the vegetation on the TP for the period 1982–2020. To explore the relationship between the TP vegetation and ENSO/IBOM, this study also used the Niño 3.4 index obtained from the NOAA CPC and the IOBM index obtained from the National Climate Centre of China Meteorological Administration (NCC/CMA).

The monthly meteorological variables were derived from the station-observed dataset of the National Meteorological Information Center of China, including precipitation, surface air temperature, ground surface temperature, and sunshine duration. Daily meteorological elements at 88 observational stations in the TP (with the average altitude exceeding 3000 m above sea level) were processed into the monthly data on a 0.5◦ × 0.5◦ grid by daily accumulation and Cressman spatial interpolation [66]. Besides the Cressman interpolation, we performed the elevation correction of the meteorological variables following the elevation correction equation of He et al. [67], which involves the calculation of the elevation lapse rate. These processed elements were used to illustrate the local effects on the TP vegetation.

#### *2.2. The Study Period and Methods*

The ISM is a crucial element of the Asian summer monsoon system [36] and a major source of water vapor for India that is responsible for over 2/3 of the annual precipitation over India [68]. Based on precipitation in Kerala, the southernmost state of the Indian subcontinent, the Indian Meteorological Department defines early June (early October) as the time for the ISM onset (demise) [69]. The main growing season for the TP vegetation is generally from June to September [35]. Considering the overlapped period of the ISM and vegetation growing season, June to September (JJAS) was selected as the main study period.

Following previous studies [25,60,61], we spliced the GIMMS and MCD19A3CMG NDVIs datasets. Before splicing the datasets, the MCD19A3CMG NDVI was downscaled and interpolated to the same resolution as the GIMMS NDVI. Further comparisons revealed that the two NDVIs on the TP have consistent characteristics in the growing seasons and significantly correlate with each other. As such, we can establish linear regression equations between the GIMMS and MCD19A3CMG NDVIs and eventually obtain a longer JJAS TP NDVI dataset by fitting and splicing the two datasets [35]. To capture the varying characteristics of the TP vegetation on inter-annual scales, the rotated empirical orthogonal function (REOF) decomposition and the North test were applied in this paper, ensuring that the leading REOF modes are homogeneous and independent [70].

All data involved in this study were subtracted from the monthly mean climatology, seasonally averaged, and detrended. Several frequently used statistical analysis methods

were utilized to examine the influence of the ISM on vegetation on the TP in its growing season, and the specific methods were as follows: (1) REOF analysis and regional average were used to derive the uniform NDVI pattern (UNP) and ISM indices, respectively. (2) Linear (partial) correlation analyses were applied to determine the correlations of the TP vegetation with the ISM, IOBM, ENSO, precipitation on the TP, etc., where partial correlation analysis was used to determine the actual correlations between two of the three related variables. (3) Univariate linear regression analyses were employed to explore the process of ISM influencing the JJAS TP NDVI and the influence of ISM on climatic factors on the TP. (4) The contributions of ISM and four climatic factors to the inter-annual variability of vegetation on the TP in the growing season were also explored through multiple linear regression analysis. (5) The potential feedback between the TP thermal conditions and vegetation growth was discussed through linear correlation and composite analysis. The importance of the findings was determined via Student's *t*-test. Note that all data do not have the same horizontal precision, which we unified in our processing.
