**1. Introduction**

Vegetation within protected areas such as game reserves provides wildlife and society with indispensable ecosystem goods and services [1] including food, medicinal resources, aesthetic value, and recreational opportunities [2]. However, inappropriate management and other disturbances affect the potential productivity and spatial extent of this resource [3]. Thus, any factor that poses a threat to vegetation and its associated benefits which could affect their productivity in the protected areas needs to be identified and monitored. One such threat is an increase in temperature above normal as well as a prolonged decline in precipitation and soil moisture, leading to extreme climatic events such as droughts, which severely affect vegetation productivity [4]. Drought-related impacts are becoming more multifaceted, as explained by their rapidly growing consequences in sectors such as recreation and tourism, agriculture, and energy [5].

The influence of drought on vegetation varies in the spatial and temporal scales, and these are projected to increase with climate change [6,7]. This behavior affects wildlife, particularly in semi-arid and arid environments where herbivory is strongly restricted by vegetation extent and water availability [8]. In the north-east part of KwaZulu-Natal, South Africa, for example, droughts are becoming a recurrent and prominent feature [9,10], affecting vegetation, water and wildlife resources notably in the Hluhluwe-iMfolozi Park (HiP), the oldest proclaimed game reserve in Africa, as reported in this paper. Furthermore, these impacts have potential consequences that could incapacitate this game reserve's support of its specialist grazers such as rhinos [11].

Understanding the association between vegetation productivity and climatic variables such as precipitation and temperature has, therefore, become a high priority. To address this, spatiotemporal tools that can integrate climate data with other information of interest are required. Remotely sensed data provide the opportunity to monitor vegetation dynamics in a systematic manner [12]. They play a growing role in drought detection and management as they afford up-to-date information over various time and geographic scales and complement alternative techniques such as field surveys [4] and interviews [13]. Remote sensing's systematic observation allows us to track vegetation conditions from the 1970s to the present [14] and provides the means to integrate the record with causal factors. This study investigates vegetative drought which is the vegetation stress as a function of moisture deficit [15].

Several drought studies based on satellite-derived measurements have exploited key indicators such as the (a) normalized difference vegetation index (NDVI), a ratio of the difference between the near-infrared and red bands of the spectrum over the sum of the near-infrared and red bands [16,17], which is a robust indicator of vegetation productivity [18]; (b) the normalized difference infrared index (NDII) which contain additional information on water availability in the soil for use by vegetation [19] as measured by the ratios of the near-infrared and short-wave infrared [20]; and (c) the evapotranspiration (ET) which includes both the loss of root zone soil water through transpiration (influenced by stomatal conductance), as well as evaporation from bare soil [21]. These studies have enhanced our understanding of how vegetation reacts to drought events over time [22–25].

Hitherto, numerous studies have explored vegetation changes using NDVI in response to climatic variability. Most have shown that vegetation is largely swayed by the El Niño/Southern Oscillation (ENSO) phenomenon and have been established to respond well to climatic variables [10,24,26,27]. These studies in different climatic regions have revealed climate-induced effects in key economic sectors such as agriculture [28] and forestry [24,29]. Most recently, Huang et al. [30] used MODIS-derived NDVI to demonstrate how vegetation responds to climate variation in the Ziya-Daqing basins of China. Their results showed that the trends of growing season NDVI were significant in the forest, grassland, and highlands of Taihang but insignificant in most plain drylands [27]. They also showed how grassland, as the primary vegetation on the Qinghai-Tibet Plateau, has been increasingly influenced by water availability due to droughts over the last decade.

Several factors make the HiP an ideal site for assessing the effects of drought on wildlife. First, Bond et al. [31] established that droughts largely influence the extent of grazing vegetation in the reserve. More recently, Xulu et al. [10] showed how the recent intense drought moderated the vegetation health of commercial plantations located ~70 km from the park. Second, the HiP is an important conservation area and ecotourism destination in South Africa [32], so the resultant socio-economic impacts of ecosystem changes are of great concern. In this study, therefore, we aim to evaluate the influence of climatic variability on vegetation in the game reserve over the period of 2002 to 2017. This is the first attempt to demonstrate the spatial dimension of the drought effects in the HiP using satellite data. We show how to construct a MODIS-derived NDVI time series in the GEE platform, and perform statistical tests to determine the causal influence of climatic variables in the reserve.
