Advances in Optical and Thermal Remote Sensing of Vegetative Drought and Phenology

Abstract

In recent decades, remote sensing of vegetative drought and phenology has gained considerable attention from researchers, leading to a significant increase in research activity in this area. While new drought indices are being proposed, there is also growing attention on how variations in phenology affect drought detection. This review begins by exploring the crucial role of satellite optical and thermal remote sensing technologies in monitoring vegetative drought. It presents common methods after revisiting the foundational concepts. Then, the review examines remote sensing of land surface phenology (LSP) due to its strong connection with vegetative drought. Subsequently, we investigate vegetative drought detection techniques that consider phenological variability and recommend approaches to improve the detection of vegetative drought, emphasizing the necessity to incorporate phenological metrics. Finally, we suggest potential future work and directions. Unlike other review papers on remote sensing of vegetative drought, this review uniquely surveys the comprehensive advancements in both detecting vegetative drought and estimating LSP through optical and thermal remote sensing. It also highlights the necessity and potential applications for these practices.

1. Introduction

Droughts are among the most devastating natural disasters, impacting ecosystems, agriculture, and water resources. This phenomenon occurs when insufficient rainfall results in a water deficit. With accelerating global warming, droughts have been striking Earth more frequently and severely, posing severe threats to the environment, human life, and sustainable economic and social development. The ability to accurately detect and monitor drought is essential for developing effective mitigation and adaptation strategies. According to current generally accepted definitions, drought can be classified into vegetative drought, meteorological drought, hydrological drought, and socio-economic drought [1]. In this article, we focus on vegetative drought. Vegetative drought impacts not only agriculture but also natural vegetation, including forests, grasslands, and other forms of plant life. It refers to a situation where vegetation experiences stress or reduced growth due to inadequate water availability. Vegetative drought has far-reaching impacts on both the environment and socio-economic conditions. It leads to changes in ecosystems and a reduction in biodiversity, as well as soil degradation and water resource shortages, which in turn affect agricultural production and food security, causing economic losses for farmers. Drought exacerbates economic losses, raises the government’s disaster relief burden, and can lead to population migration, social instability, and resource competition. The definitions of the other drought types can be found in [2,3] the literature. Figure 1 illustrates the evolutionary relationships among these drought types.

Vegetative droughts present formidable challenges to agricultural productivity, ecosystem health, and water resource management, necessitating timely detection and response to mitigate their impacts on the environment and society. Traditional drought monitoring techniques primarily rely on methods such as meteorological stations, soil moisture sensors, and field surveys to assess vegetation health by collecting and analyzing meteorological data and on-site samples. However, these methods have several limitations. First, traditional monitoring has limited spatial coverage, typically reflecting only localized conditions, which makes it challenging to effectively monitor large regions. Second, data collection and processing take time, resulting in an inability to reflect drought conditions in real time. Additionally, traditional methods are relatively expensive, are significantly affected by weather conditions, and have low update frequencies, limiting their ability to timely and widely capture changes in drought conditions. In contrast, remote sensing technology can provide broader, real-time, and more efficient data, overcoming the limitations of traditional monitoring. Remote sensing acquires radiation reflected and emitted from the Earth’s surface across multiple electromagnetic bands [4] using sensors on satellites and aircraft. Remote sensing techniques are categorized by wavelength ranges, including visible light remote sensing (0.4–0.7 µm), near-infrared (NIR) remote sensing (0.75–1.5 µm), shortwave infrared (SWIR) remote sensing (1.5–3 µm), midwave infrared (MIR) remote sensing (3–6 µm), thermal infrared (TIR) remote sensing (6–15 µm), and microwave remote sensing (1 mm–1 m). Optical and thermal remote sensing is a well-established approach in remote sensing, utilizing solar and infrared wavelengths between 0.4 and 15 μm [5]. This range includes visible light, NIR, SWIR, MIR, and TIR. In recent decades, optical and thermal remote sensing have been widely used for detecting vegetative drought and estimating phenological metrics. By capturing data across these spectral bands, it is possible to retrieve drought-related variables, such as vegetation health, chlorophyll content, water content, temperature, soil moisture, and evapotranspiration, from space. These variables can be further combined into drought severity indicators to achieve effective drought detection and monitoring [6,7]. Satellite remote sensing has been used to detect and monitor drought from space since the first Landsat satellite was launched into orbit in 1972 [8]. Given its extensive and comprehensive spatial coverage, along with regular and frequent revisits, satellite remote sensing has established itself as a primary methodology for monitoring vegetative drought across vast geographic areas [9,10]. In this article, we focus on satellite optical and thermal remote sensing.

Vegetation phenology, the study of recurring life cycle events in plants, provides critical insights into ecosystem dynamics, biodiversity patterns, and responses to environmental change [11]. Land surface phenology (LSP), a subset of phenology focusing on vegetation dynamics on the Earth’s surface, has emerged as a valuable tool for monitoring and understanding terrestrial ecosystems [12]. Satellite remote sensing techniques offer valuable tools for monitoring LSP over large spatial scales. In LSP research, the main tasks involve extracting various phenological metrics from the time series data of remote sensing products, primarily those related to vegetation and temperature. These phenological metrics include Start of Growth Season (SOS), End of Growth Season (EOS), Peak of Growth Season (POS), and Length of Growth Season (LOS), among others [13].

It has been established that drought and phenology are closely connected, with mutual influences and feedback. As an important environmental stress factor, drought can significantly alter the growth and development rhythms of plants [14,15,16]. The water deficit caused by drought can delay key phenological stages of plants, such as green-up and senescence, while also shortening their growth periods. These phenological changes can, in turn, provide important information for identifying and forecasting drought events [17,18,19,20]. Interestingly, the long-term accumulation and analysis of phenological data can help us better understand the patterns of drought occurrence [21,22,23]. By tracking the abnormal changes in phenological indicators, we can detect the precursor signals of drought events, facilitating an early warning and response to drought [18,24,25,26,27].

The phenological response to frequent and prolonged drought events has been widely investigated in recent decades. Recent research has highlighted how changes in phenology can impact the identification of drought events. Figure 2 presents publications related to the remote sensing of drought and phenology, sourced from the core library of the Web of Knowledge. As shown, there has been a significant increase in publications on both remote sensing of drought and remote sensing of phenology over the past two decades. However, despite early attention to the relationships between remote sensing of phenology and drought, there has been less focus on studying both aspects simultaneously. Nonetheless, it is important to note that there has been an increasing number of publications in recent years that address both drought and phenology in the context of remote sensing.

Overall, this review aims to investigate how phenology-informed approaches can enhance the detection of vegetative drought using remote sensing. It emphasizes the importance of incorporating phenological variability into these methods. The review provides a comprehensive examination of the connections between vegetative drought and LSP, with the goal of advancing our understanding and monitoring of these critical environmental phenomena. This review focuses on the application of satellite optical and thermal remote sensing for detecting vegetative drought, with three main objectives:

To delve into the crucial role of remote sensing in monitoring vegetative drought, revisiting its underlying principles, and exploring prevalent methods;

To examine the interrelation between LSP with vegetative drought, focusing on the challenges in remote sensing of vegetative drought caused by phenology variability;

To review the recent advancements in detecting vegetative drought while considering phenology variability.

This work is unique in that it simultaneously examines recent progress in the detection of vegetative drought and LSP using satellite optical and thermal remote sensing. It also explores the relationship between vegetative drought and phenology, an area that has previously received limited attention. Furthermore, unlike some existing review papers, this study emphasizes the importance of exercising caution when identifying vegetative drought and not neglecting variations in phenology.

2. Satellite Data for Vegetative Drought Detection and Phenology Estimation

The satellite optical and thermal remote sensing data used for drought monitoring and phenology extraction involve multiple spectral bands and sensor types. Monitoring the temporal dynamics of vegetation indices (VIs) is crucial for detecting vegetative drought and extracting LSP, as vegetative drought and phenology are processes that evolve over time, though usually occurring gradually and often taking several weeks to months to become significantly apparent. High-frequency (e.g., daily or weekly) remote sensing observations are required to capture the temporal patterns of vegetation changes. To investigate significant advances in drought and phenology research, we focus on sensors with low to moderate spatial resolution and relatively high temporal frequency. These sensors should cover the optical-to-thermal spectrum and be capable of monitoring vegetation dynamics on a large scale. These data can then be used to derive various remote sensing indicators related to vegetative drought and phenological parameters, including VI, leaf/canopy temperature, land surface temperature (LST), soil moisture, evapotranspiration, and more. After more than four decades of research across multiple space missions, we have now accumulated a rich dataset of observations from space, covering a wide range of sensors operating on different physical principles.

Table 1. Commonly used data sources for remote sensing of drought and phenology.

Data Source	Platform	Resolution	Band	Period	Advantages	Limitations
Moderate Resolution Imaging Spectroradiometer (MODIS)	Terra and Aqua	250 m (visible light and NIR), 500 m (NIR), 1 km (TIR)	36 bands (0.4–14.4 µm)	1999–Present	Frequent temporal coverage, wide spectral range, global coverage	Moderate spatial resolution, data quality affected by cloud cover
Landsat Series	Landsat 8, ETM+ on Landsat 7, TM on Landsat 5	30 m (visible light, NIR, NIR, SWIR), 15 m (panchromatic), 60 m (TIR, Landsat 5), 100 m (TIR, Landsat 8)	11 bands (0.43–12.51 µm)	1972–Present	High spatial resolution, long-term data record	16-day revisit period, data gaps (Landsat 7 SLC issue)
Sentinel-2	Sentinel-2A and 2B	10 m (visible light and NIR), 20 m (Red Edge, SWIR), 60 m (Coastal Aerosol, Water Vapor, SWIR Cirrus)	13 bands (0.443–2.19 µm)	2015–Present	High spatial resolution, frequent revisit time, free data access	Data quality affected by cloud cover, complex data processing
AVHRR (Advanced Very High-Resolution Radiometer)	NOAA and MetOp	1.1 km	6 bands (0.58–12.5 µm)	1978–Present	Long-term data availability, wide swath width	Coarse spatial resolution, data quality variability
VIIRS (Visible Infrared Imaging Radiometer Suite)	Suomi NPP and NOAA-20	375 m (I-bands), 750 m (M-bands)	22 bands (0.412–12.01 µm)	2011–Present	Moderate spatial resolution, comprehensive spectral bands, near-daily global coverage	Large data volumes, relatively new and shorter time series
ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer)	Terra	15 m (visible light and NIR), 30 m (SWIR), 90 m (TIR)	14 bands (0.52–11.65 µm)	1999– Present	High spatial resolution, multiple spectral bands, global coverage	SWIR sensor inoperable, 16-day revisit period, data processing requirements

summarizes some of the most commonly used satellite optical and thermal remote sensing data for monitoring drought and extracting phenology, along with their key characteristics. The comprehensive utilization of these data can provide multifaceted and comprehensive information support for monitoring drought and extracting phenology.

3. Detection of Vegetative Drought Using Optical and Thermal Remote Sensing

3.1. Fundamentals

Over the past decades, researchers worldwide have developed various methods for monitoring drought that correspond to different remote sensing technologies. These methods enable the detection of vegetative drought-related variables, such as vegetation growth status, temperature differences between vegetation and air, soil moisture, and evapotranspiration, using specific spectral bands that are sensitive to these factors. Researchers utilize these variables either individually or systematically to create mathematical indices for monitoring and providing early warnings of drought. Methods based on moisture, VI, temperature, and evapotranspiration are currently the most widely used techniques for detecting vegetative drought through optical and thermal remote sensing. Additionally, constructing drought indices based on the spectral (or variable) feature space is an effective approach for monitoring vegetative drought using remote sensing.

Remote sensing of vegetative drought primarily involves the calculation of drought indices. These indices can be categorized into two types based on their calculation process: simple drought indices and composite drought indices. Simple drought indices are directly computed through band arithmetic, while composite drought indices are further calculated using simple drought indices as inputs. Based on several review papers [8,28,29,30,31] published in renowned journals, such as Remote Sensing of Environment, Remote Sensing, ISPRS of Photogrammetry and Remote Sensing, and IEEE Transactions on Geoscience and Remote Sensing, we have summarized the most common drought indices from satellite optical and thermal remote sensing. Figure 3 illustrates their relationships and connections.

3.2. Methods

Many drought indices have been proposed in recent decades that utilize optical and thermal remote sensing. Qin et al. provided a comprehensive summary of monitoring agricultural drought using optical and thermal remote sensing [6]. Alahacoon and Edirisinghe reviewed 111 drought indices [28]. Vegetative drought indices can be classified into several categories based on the variables involved. These variables include vegetation conditions, temperature, moisture, and evapotranspiration. In the following subsections, we introduce these methods. However, before we do this, we list these representative methods and their application examples, as shown in

Table 2. Prevalent vegetative drought monitoring methods and their application examples.

No.	Variable	Name	Formula	Reference	Examples	Advantages	Disadvantages
1	Moisture	Moisture Stress Index (MSI)	$MSI={\displaystyle \frac{{\rho }_{MIR}}{{\rho }_{NIR}}},$ $\mathrm{where} \rho$ is the reflectance.	[32]	[33,34]	Effectively reflects vegetation moisture conditions; sensitive to changes in soil and vegetation moisture.	Highly dependent on soil type and vegetation species, which may lead to inaccuracies; requires ground truth data for calibration.
2		Reciprocal of the Moisture Stress Index (RMSI)	$RMSI={\displaystyle \frac{1}{MSI}}$	[35]	[36,37]	Enhances sensitivity to low moisture conditions through its reciprocal form; can be combined with other indices to provide more comprehensive information.	May perform poorly under high moisture conditions, leading to misinterpretation; data processing is complex and requires careful interpretation.
3		Simple Ratio Water Index (SRWI)	$SRWI={\displaystyle \frac{{\rho }_{NIR}}{{\rho }_{SWIR}}}$	[38]	[39,40]	Simple to calculate and easy to understand and apply; shows good responsiveness to vegetation moisture changes.	Sensitive to variations in light conditions and soil background, which may affect results; provides relative information, lacking absolute assessments of moisture content.
4		Normalized difference water index (NDWI)	$NDWI={\displaystyle \frac{{\rho }_{NIR}-{\rho }_{SWIR}}{{\rho }_{NIR}+{\rho }_{SWIR}}}$	[41]	[42,43]	Effectively distinguishes water bodies from vegetation, suitable for monitoring moisture changes; can utilize existing remote sensing data for calculations.	May be affected in areas with high vegetation cover, leading to errors in results; highly influenced by weather conditions, which may cause fluctuations in data quality.
5		Normalized Difference Infrared Index (NDII)	$NDII={\displaystyle \frac{{\rho }_{NIR}-{\rho }_{MIR}}{{\rho }_{NIR}+{\rho }_{MIR}}}$	[44]	[42,45]	Sensitive to vegetation moisture conditions; effectively distinguishes different types of vegetation.	Sensitive to atmospheric conditions and surface characteristics; needs to be used in conjunction with other indices to improve accuracy.
6	Temperature	Temperature Condition Index (TCI)	$TCI={\displaystyle \frac{T_{max}-T}{T_{max}-T_{min}}}$	[46]	[47,48]	Effectively reflects the temperature conditions of vegetation; can be easily calculated using remote sensing data, making it suitable for large-scale monitoring; can provide a more comprehensive drought assessment when combined with other vegetation indices.	May be affected by climate change and seasonal temperature fluctuations, leading to unstable results; does not directly measure soil or vegetation moisture; needs long-term temperature data for baseline comparisons.
7	Vegetation conditions	Anomaly Vegetation Index (AVI)	$AVI=NDVI-\overline{NDVI}$	[43]	[49,50,51]	Reflects the relative condition of vegetation compared to historical data; can be easily calculated from satellite imagery.	Sensitive to seasonal variations, which may complicate interpretation; requires a long-term dataset for accurate historical comparisons.
8		Vegetation Condition Index (VCI)	$VCI={\displaystyle \frac{NDVI-{NDVI}_{min}}{{NDVI}_{max}-{NDVI}_{min}}}$	[46]	[52,53]	Reflects the relative condition of vegetation compared to historical data; can be easily calculated from satellite imagery.	Sensitive to seasonal variations, which may complicate interpretation; requires a long-term dataset for accurate historical comparisons.
9		Vegetation Health Index (VHI)	$VHI=\alpha VCI+\beta TCI$	[46]	[54,55]	Combines information from both the VCI and the TCI, providing a comprehensive view of vegetation health.	Complexity in calculation due to the integration of multiple indices; may require ground truth data for calibration to improve accuracy.
10		Solar Induced Fluorescence (SIF)	May be measured using narrow absorption lines such as the Fraunhofer lines or the O2 absorption lines.	[56]	[57,58]	Directly measures photosynthetic activity, providing insights into vegetation stress and health; sensitive to changes in water availability and can indicate drought conditions effectively.	Requires advanced sensors and technology, which may not be widely available; interpretation can be complex, as fluorescence signals can be influenced by various environmental factors.
11	Feature space	Perpendicular Drought Index (PDI)	$PDI={\displaystyle \frac{1}{\sqrt{M^2+1}}}\left({\rho }_{Red}+M{\rho }_{NIR}\right),$ where M represents the slope of the soil line in the NIR/Red feature space.	[59]	[60,61]	Simple to compute and interpret, making it accessible for various applications; effectively distinguishes between drought and non-drought conditions based on vegetation health.	May not account for all environmental factors influencing vegetation health, potentially leading to inaccuracies; limited sensitivity to subtle changes in moisture levels.
12		Modified Perpendicular Drought Index (MPDI)	$MPDI=$ ${\displaystyle \frac{{\rho }_{Red}+M{\rho }_{NIR}-f_v\left({\rho }_{v,Red}+M{\rho }_{,v,NIR}\right)}{\left(1-f_v\right)\sqrt{M^2+1}}},$ $\mathrm{where} f_v \mathrm{is} \mathrm{the} \mathrm{fraction} \mathrm{of} \mathrm{vegetation}. {\rho }_{v,Red}$$\mathrm{and} {\rho }_{v,NIR}$ are pure vegetation reflectances.	[59]	[62,63]	Enhances the original PDI by incorporating additional spectral information, improving sensitivity to drought conditions; provides a more nuanced assessment of vegetation health and moisture stress.	Increased complexity in calculation compared to PDI, which may require more data and processing; still dependent on accurate input data, which can vary with sensor quality and environmental conditions.
13		Temperature Vegetation Dryness Index (TVDI)	$TVDI={\displaystyle \frac{LST-{LST}_{min}}{a+bNDVI-{LST}_{min}}},$ where LST is land surface temperature, and a and b are parameters defining the dry edge in the LST/NDVI feature space.	[64]	[65,66]	Combines temperature and vegetation indices to provide a clear indication of moisture stress; effective in distinguishing between wet and dry conditions.	Sensitive to atmospheric conditions, which can affect temperature readings; requires accurate temperature and vegetation data.
14		Vegetation Temperature Condition Index (VTCI)	$VTCI={\displaystyle \frac{{LST}_{{NDVI}_{i^{max}}}-{LST}_{{NDVI}_i}}{{LST}_{{NDVI}_{i^{max}}}-{LST}_{{NDVI}_{i^{min}}}}},$ where ${LST}_{{NDVI}_{i^{max}}}=a+b{NDVI}_i,$${LST}_{{NDVI}_{i^{min}}}=a^{\prime }+b^{\prime }{NDVI}_i,$${NDVI}_i$$\mathrm{is} \mathrm{the} \mathrm{NDVI} \mathrm{value} \mathrm{at} \mathrm{a} \mathrm{certain} \mathrm{pixel} \mathrm{location} i, a, b \mathrm{and} a^{\prime }$$, b^{\prime }$ are edge line coefficients in LST and fractional vegetation coverage (LST/FVC) feature space.	[67]	[68,69]	Integrates temperature and vegetation data to assess plant stress and drought conditions effectively; can provide timely information for decision-making in drought management.	Complexity in calculations may require advanced data processing techniques; interpretation can be affected by local climatic conditions, which may lead to variability in results.
15	Evapotranspiration	Water Deficit Index (WDI) or Crop Water Stress Index (CWSI) for full-cover canopies	$WDI=1-f_{PET},$ where $f_{PET}={\displaystyle \frac{ET}{PET}},$ ET: evapotranspiration, PET: potential evapotranspiration.	[70]	[71,72]	Specifically designed to assess water stress in crops, providing relevant information for agricultural management; effective for full-cover canopies, allowing for accurate monitoring of plant water status.	Limited applicability to non-crop vegetation types, which may reduce its overall utility; requires precise temperature and humidity data, which may not always be readily available.
16		Evaporative Stress Index (ESI)	$ESI={\displaystyle \frac{〈f_{PET}〉-\overline{〈f_{PET}〉}}{{\sigma }_{〈f_{PET}〉}}}$	[73]	[74,75]	Reflects the evaporative demand on vegetation, making it sensitive to changes in moisture availability; useful for monitoring drought conditions across various vegetation types and ecosystems.	May be influenced by local climatic conditions, which can affect the accuracy of the index; requires comprehensive meteorological data, which may not be accessible in all regions.
17	Combination	Drought Severity Index (DSI)	$DSI={\displaystyle \frac{z-\overline{z}}{{\sigma }_z}},$ where $Z=Z_{f_{PET}}+Z_{NDVI},$ $Z_{f_{PET}}={\displaystyle \frac{f_{PET}-\overline{f_{PET}}}{{\sigma }_{f_{PET}}}},$ $Z_{NDVI}={\displaystyle \frac{NDVI-\overline{NDVI}}{{\sigma }_{NDVI}}}.$	[76]	[77,78]	Provides a quantitative measure of drought severity, allowing for clear assessment and comparison; can integrate various meteorological and hydrological data, making it comprehensive.	May require extensive historical data for calibration, which can be a limitation in data-scarce regions; interpretation can be complex, as it involves multiple factors.
18		Reconnaissance Drought Index (RDI)	Normalized RDI: ${RDI}_n\left(k\right)={\displaystyle \frac{a_k}{\overline{a_k}}}-1,$ Standardized RDI: ${RDI}_{st}\left(k\right)={\displaystyle \frac{y_k-\overline{y_k}}{{\sigma }_{y_k}}},$ where $a_k={\displaystyle \frac{\sum _{j=1}^k{P}_j}{\sum _{j=1}^k{PET}_j}},$ Pj is precipitation, $y_k=\ln a_k.$	[79]	[80,81]	Combines rainfall and potential evapotranspiration data, providing a holistic view of drought conditions; useful for assessing both short-term and long-term drought impacts on vegetation.	Sensitive to data quality, as inaccuracies in rainfall or evapotranspiration data can affect results; may not capture localized variations in drought conditions effectively.

.

3.2.1. Vegetation Condition

Vegetation condition refers to the health status and growth state of vegetation in a particular area. Vegetation condition-based methods utilize remote sensing data to evaluate and monitor drought by examining the health and vigor of vegetation. In recent decades, drought monitoring based on vegetation conditions has emerged as one of the most widely used technologies. Drought conditions are characterized by increased reflectance in the visible light band—especially in the red-light band—and decreased reflectance in the NIR band. The extent of these changes increases with the severity of the drought. Building on this characteristic, Tucker [82] devised NDVI, which captures the response of vegetation to drought conditions and stands as the earliest index employed for drought monitoring. NDVI serves as the foundation for the development of other drought indices based on vegetation conditions. NDVI is calculated according to the following Equation (1):

NDVI = (NIR − Red)/(NIR + Red)

(1)

Although NDVI has been widely used for environmental and ecological monitoring, it has some limitations. To mitigate the influence of atmospheric conditions, bidirectional reflected radiation, soil and canopy background, as well as saturation issues in high biomass areas, some other indices, such as the Enhanced Vegetation Index (EVI) [83] and Soil-Adjusted Vegetation Index (SAVI [84], have been introduced and widely adopted.

Given the close relationship between vegetation and local geographic climate, as well as vegetation distribution and growth patterns, spatial comparability of drought monitoring directly using these VIs is limited. Consequently, the Anomaly Vegetation Index (AVI) [43] and the Vegetation Condition Index (VCI) [46] have been proposed for drought monitoring using long-term vegetation data. In these methods, historical data are first used to establish a baseline, and deviations from the baseline indicate abnormalcy, such as drought stress. Both exhibit superior spatiotemporal comparability compared to NDVI and are extensively employed in large-scale drought monitoring.

Solar-induced chlorophyll fluorescence (SIF), an optical signal, has been used to quantify hydrological states, thereby helping to characterize vegetative drought [56]. SIF is a key indicator of photosynthetic activity in plants, reflecting the efficiency of light absorption and energy conversion processes [85,86]. SIF serves as a valuable remote sensing tool, enabling researchers to monitor plant health. Approaches used to quantify SIF emission from vegetation top-of-canopy (TOC) radiance are anchored by a simple equation of wavelength $\lambda$ describing the additive contributions of solar-reflected light (r), incident irradiance (E), and SIF radiance (SIF) to the total TOC radiance L [87]:

$$L\left(\lambda \right)={\displaystyle \frac{r\left(\lambda \right)E\left(\lambda \right)}{\pi }}+SIF\left(\lambda \right)$$

(2)

There have been many methods devised to retrieve SIF. Most of them are built on the key assumption that prior knowledge of the spectral shape of all terms of the equation can be leveraged to estimate the unknown terms.

3.2.2. Temperature

Satellite remote sensing has been used to retrieve various temperature measurements, including air temperature, LST, and the temperature of plant leaves and canopies. Thermal remote sensing provides temperature-sensitive channels for detecting various temperatures and their differences. It primarily measures vegetation temperature and LST based on the characteristics of electromagnetic wave radiation and the thermal balance between objects and their environment.

When vegetation experiences water scarcity, in addition to the change in VI, leaf temperature and canopy temperature also increase rapidly due to the water deficit. Based on this physiological characteristic of vegetation, temperature differences are used in drought monitoring. Specifically, reduced water availability decreases the latent heat flux at the leaf surface, leading to a corresponding increase in sensible heat. This results in a larger temperature difference between the foliage and the surrounding air [88,89].

Similar to the VCI Index, Kogan constructed the Temperature Condition Index (TCI) [46] by introducing historical information on temperature, which makes up for the shortcoming of the strong lag of VCI. The TCI method assumes that light intensity, soil texture, crop type, and other factors have little influence on soil moisture and indirectly indicates drought through changes in canopy temperature. In actual monitoring, the TCI value is easily affected by sensors, atmospheric conditions, vegetation types, and other factors, and the monitoring accuracy is sometimes reduced. To overcome the shortcomings of VCI and TCI, the Vegetation Health Index (VHI) combines NDVI and LST information to provide a comprehensive assessment of vegetation health and drought conditions, which are calculated as a weighted average of VCI and TCI [90].

3.2.3. Moisture

The basic principle of using optical remote sensing to assess soil moisture is based on the differences in spectral reflectance caused by changes in surface soil moisture (SSM) [91]. Thermal remote sensing can also be used to retrieve soil moisture information by leveraging the fact that soils with different moisture levels exhibit varying thermal conductivities. This underlies the development and application of thermal inertia, representing the resistance of a material to temperature change. Many researchers have studied and developed this concept [92,93,94,95]. Based on the above principles, two kinds of prevalent methods have been developed for retrieving soil moisture:

(1) Statistical methods based on correlations:

This kind of method combines spectral data, establishing a correlative function between spectral reflectance or reflectance-based indices (including VIs) and soil moisture to estimate the amount of soil moisture. Most of these methods have been developed using single statistics, multivariate analysis, or wavelet analysis. In general, such methods can be summarized as SSM being a function of spectral reflectance or a spectral reflectance-based index, given by

SSM = f(r_i or index),

(3)

where f is a function (similarly hereinafter), r_i is the spectral reflectance at channel i, and the index typically includes spectral reflectance from multiple channels within the optical and thermal band range.

(2) Thermal inertia models.

This class of methods estimates soil moisture by obtaining the thermal inertia of the land surface. Watson et al. [96] first proposed the thermal inertia model and applied it to monitor soil moisture. Price [97] proposed the Apparent Thermal Inertia (ATI) [95] model. The ATI model is often used as a substitute for the actual thermal inertia model because it can closely approximate real thermal inertia [98] while being simpler to calculate. Price’s ATI expression is as follows:

$$\beta =(1-A)/∆T,$$

(4)

where A is the surface albedo, and $∆T$ is the soil temperature difference.

There are also other normalized-difference indices and ratio indices commonly used for analyzing leaf water content. The NDII [99], the MSI [32], and the RMSI [32] include one normalized-difference index and two ratio indices, all of which are based on the local absorption maximum of liquid water at 1240 nm. The NDWI [41] and the SRWI [100] are based on the local absorption minimum at 1650 nm.

3.2.4. Evapotranspiration

Evapotranspiration (ET) is an important component of the water and energy cycle, reflecting mass and energy exchange between ecosystems and the atmosphere [101,102]. The basic principle of remote sensing ET models is to use remote sensing to obtain land surface information, which is then combined with physical process models. This approach estimates ET by calculating surface energy balance or employing empirical relationships.

Evapotranspiration models include some commonly used parameters, such as albedo, emissivity, LST, LAI, and NDVI. These parameters may be obtained through optical and thermal remote sensing techniques and used to estimate net radiation, vegetation transpiration, and soil moisture, thereby effectively estimating the amount of ET. Specifically, these models use remote sensing technology to acquire the components of land surface energy, partition the available energy between vegetation and soil surfaces, and construct the biophysical constraints of ET. Therefore, ET is an integrated variable that involves other variables.

Remote sensing-based ET estimation methods can be categorized into the following groups: (a) [103] energy balance models, (b) [104] surface energy balance models, (c) [105] vegetation index models, (d) [106] two-source energy balance models, (e) Priestley–Taylor [107] models, (f) Penman–Monteith [108] models, and (g) [109] simplified methods. These references represent seminal works and methodologies in each category, providing a comprehensive understanding of remote sensing-based ET estimation methods. A recent comparison of remote sensing ET models was conducted by Bai [110].

Since ET causes the return of most precipitation to the atmosphere, several drought indicators integrating ET as an input variable have been developed, including the WDI [70], ESI [73,111], DSI [76], and RDI [7].

3.2.5. Feature Space

The feature space refers to a multidimensional space constructed from various remote sensing data, such as spectral reflectance, temperature, and VIs. This space effectively differentiates between different surface characteristics. By analyzing data points within the feature space, differences between drought-affected areas and normal areas can be identified. Currently, there are several commonly recognized feature space-based drought indices including PDI/MPDI in the NIR/Red feature space, TVDI in the Ts/NDVI feature space, and VTCI in the LST/FVC feature space.

PDI describes the distribution of SM in the NIR-Red feature space [59]. It is a simple and effective index for drought monitoring but lacks accuracy for vegetation-rich and barren lands. MPDI was introduced to overcome the weaknesses of PDI. It improves upon the PDI by replacing a portion of the vegetation component to better differentiate the effects of vegetation [112] on drought assessment.

TVDI is another commonly used index that incorporates temperature measurements for monitoring drought [113]. TVDI combines LST and NDVI information (LST/NDVI feature space) to estimate soil moisture and drought conditions. It is calculated using the relationship between LST and NDVI, which can provide insights into the status of soil moisture.

Similar to TVDI, VTCI is constructed for estimating soil moisture in the LST and fractional vegetation coverage (LST/FVC) space [114]. TVDI and VTCI represent the synergistic use of optical and thermal observations [115,116,117]. Since a comprehensive review of the progress of the LST/FVC feature space for SSM retrieval has been conducted by Petropoulos et al. and Sun et al. [118,119], readers are encouraged to refer to these critical reviews for more details.

4. LSP Extraction Using Optical and Thermal Remote Sensing

4.1. Fundamentals

Phenology primarily focuses on the timing of events that are significant for organisms, populations, and ecological communities, as well as for biogeochemical and hydrological cycles and ecosystem services. In contrast, LSP specifically refers to the seasonal patterns in vegetated land surfaces, as observed through remote sensing [12,13,120].

The basic principles of remote sensing of phenology rely on the unique spectral response patterns that different plant species and phenological stages exhibit. These variations in spectral reflectance can be observed across the visible, NIR, and SWIR regions of the electromagnetic spectrum. These distinct spectral signatures can be used to identify and track changes in vegetation phenology over time. Therefore, VIs, such as NDVI, EVI, and NDWI, are widely used to quantify vegetation greenness, biomass, and water content. Furthermore, these indices are sensitive to phenological transitions and can be used to detect the timing of key events.

Key phenological metrics, such as SOS, POS, EOS, and LOS, can be derived from the temporal profiles of VIs. These metrics provide valuable information about the timing and duration of phenological events. As shown in Figure 4, phenological parameters can be extracted from the curves of remotely sensed time series data, such as the NDVI and Gross Ecosystem Productivity (GEP).

4.2. Methods

Some methods have been developed to retrieve phenological dates, which can be categorized into three approaches: VIs-based approaches, thermal-based approaches, and modeling approaches based on their data sources and principles.

Table 3. Methodological categories for remote sensing of LSP.

Categories	Descriptions	Strengths	Limitations	Examples
VIs-based Approach	This approach utilizes vegetation indices, such as NDVI, EVI, and LAI, derived from satellite data to track seasonal changes in vegetation greenness and canopy structure.	Widely used, easy to implement, and can capture broad-scale phenological patterns.	Susceptible to atmospheric effects, may not capture fine scale phenological events, and can be influenced by vegetation heterogeneity.	[122,123,124]
Thermal-based Approach	This category of methods leverages LST data from thermal infrared sensors to monitor plant phenology, often in combination with vegetation indices.	Provides information on the thermal environment driving plant development, can capture earlier phenological events.	Requires accurate atmospheric correction, may not work well in regions with complex topography or land cover.	[125,126,127]
Modeling Approach	These methods integrate remote sensing data with process-based or empirical models to simulate and predict plant phenology.	Can incorporate additional environmental drivers, can be used for forecasting and scenario analysis.	Requires detailed parameterization, may be computationally intensive, and can be limited by model assumptions.	[128,129,130]

provides a brief summary. In the following subsections, we will discuss VI time series-based extraction methods and cumulative temperature methods, as they are currently the most commonly used methods.

4.2.1. VI Time Series-Based Methods

A variety of methods have been developed for phenology extraction from VI time series. Essentially, these methods focus on estimating change points in the VI data based on specific assumptions about vegetation dynamics. Some researchers have identified three basic assumptions regarding LSP retrieval [131]. These assumptions are as follows:

Vegetation phenology follows a repetitive seasonal cycle, and VI values vary smoothly with time;

During summer and winter, outliers in VI time series are the result of either cloud cover or atmospheric disturbances, both tending to decrease VI values;

During winter, snow under or temporarily on the canopy may impose a negative bias on the VI signal, since snow has a high visible reflectance and a low NIR reflectance;

In addition, it is very important to smooth the time series of VI data so that these methods can perform well. Zeng et al. provided a summary of common data smoothing methods [132]. According to Atkinson et al. [133], they can be classified into three categories: empirical methods, curve-fitting methods, and data transformations.

4.2.2. Cumulative Temperature Methods

The cumulative temperature method is also known as the growing degree day (GDD) or heat sum approach. This method is based on the principle that plant growth and development are primarily driven by the accumulation of heat over time.

The basic premise of the cumulative temperature method is that plants require a certain amount of heat energy, measured in the form of accumulated temperature units, to reach specific phenological stages. This heat energy is typically quantified as the sum of daily temperatures above a specific base temperature threshold, which is often set to 0 °C or 5 °C, depending on the plant species and the region [14].

The general steps involved in the cumulative temperature method are shown in Figure 5.

The cumulative temperature method has several advantages:

Simplicity and ease of implementation;

Compatibility with various satellite-derived LST data;

Ability to capture the influence of temperature on plant development;

Potential for modeling and predicting phenological events.

However, the method also has limitations, such as the need to consider other environmental factors, such as precipitation and photoperiod, that can also influence plant phenology. Combining the cumulative temperature approach with other remote sensing-based indices or models (e.g., NDVI) can often improve the accuracy and robustness of phenology monitoring.

5. Remote Sensing of Vegetative Drought, Considering Phenological Variability

5.1. Challenges in the Remote Sensing of Vegetative Drought and Phenology

5.1.1. Effect of Vegetation Physiological Characteristics on Drought Detection

During drought periods, different vegetation types exhibit varying degrees of resilience and adaptation strategies. Inherent differences in plant morphology, physiology, and life history strategies among vegetation types lead to divergent drought resistance mechanisms. Typically, drought-tolerant plants such as xerophytic shrubs and grasses exhibit adaptations such as deep root systems, stomatal regulation, and the ability to enter dormancy during water-limited conditions. In contrast, mesic plant types, including many forest trees, often lack these specialized drought avoidance traits and are more vulnerable to the effects of water scarcity [134,135]. Rooting depth and architecture are crucial determinants of a plant’s ability to access soil moisture during drought. Deep-rooted plants, such as woody species, can access water from deeper soil layers, while shallow-rooted herbaceous plants rely more on the diminishing moisture in the upper soil [135,136] layers. The differential rooting patterns lead to contrasting drought responses, with deep-rooted plants exhibiting higher resilience compared to their shallow-rooted counterparts. In addition to root traits, plants employ various physiological mechanisms to cope with drought. Drought-tolerant species often have the ability to restrict transpirational water loss through stomatal regulation, leaf shedding, or the development of more xeromorphic leaf structures [136,137]. These adaptations help maintain plant water status and photosynthetic function under limited soil moisture conditions. The life history strategies and phenological patterns of vegetation types also influence drought responses. For example, annual and ephemeral plants can complete their life cycles before severe drought occurs, whereas perennial plants must endure prolonged periods of water stress [138]. Furthermore, the timing of growth, flowering, and dormancy in relation to the seasonal drought pattern can determine the vulnerability of different vegetation types.

The diverse drought responses observed among vegetation types are a result of the complex interplay between the morphological, physiological, and life history traits of plants. Understanding these differences is crucial for predicting the impacts of drought on ecosystem composition and function, as well as for informing management strategies for sustainable land use and climate change adaptation. Nonetheless, commonly used vegetative drought monitoring indices, such as the NDVI and EVI, often overlook the specific adaptations of various plant types. These indices primarily rely on reflectance measurements in the red and NIR spectra, which can indicate vegetation cover and health. However, they may not accurately reflect the drought responses of different plant species [139]. For example, NDVI has been widely adopted for monitoring vegetation health due to its simplicity and effectiveness in various ecosystems. However, its ability to differentiate between drought-tolerant and drought-sensitive species is limited. High NDVI values can be misleading, as they may indicate healthy vegetation that is, in fact, experiencing physiological stress due to drought [140]. This limitation is particularly evident in ecosystems dominated by deep-rooted xerophytes, which may maintain high NDVI values even while facing significant water stress [141]. Moreover, EVI, which incorporates additional spectral bands to enhance sensitivity to high biomass areas, similarly fails to consider the physiological diversity among vegetation types. While it can improve the detection of changes in vegetation cover, it does not adequately reflect the physiological state of drought-sensitive species, which may show reduced photosynthetic activity despite maintaining biomass [142].

Recent studies have suggested the need for developing VIs that incorporate physiological traits, such as stomatal conductance and water-use efficiency, to provide a more accurate assessment of drought impacts on different vegetation types [143]. For instance, indices that combine spectral data with ground-based measurements of plant physiological parameters could offer a more nuanced understanding of vegetation responses to drought. Furthermore, integrating remote sensing data with ecological modeling can enhance our ability to predict vegetation responses to stress caused by drought. By incorporating species-specific traits and ecological interactions into these models, researchers can develop more robust frameworks to assess the vulnerability of different vegetation types to drought [144].

5.1.2. Effect of Drought on Phenology

The increased frequency and intensity of droughts are anticipated to have a substantial impact on LSP. The effect of drought on plant phenology has attracted growing attention in recent decades, and findings from a diverse range of studies have contributed to a comprehensive understanding of this intricate relationship. Drought can significantly impact plant phenology by altering the timing of crucial developmental stages such as bud burst, flowering, and leaf senescence. Many studies have provided evidence of the advancement or delay of these phenological events in response to drought conditions. Thus, phenology can be a valuable tool for drought monitoring. For instance, water scarcity affects vegetation phenology, which can then indicate the severity [145,146] of a drought.

Several studies have demonstrated that droughts resulted in earlier bud bursts in deciduous tree species, potentially disrupting the synchrony between plant growth and resource availability [147,148,149,150]. Similarly, some studies found that drought-induced water stress delayed flowering in various plant species, affecting their reproductive success and seed production [151,152,153,154]. In fact, the impact of drought on plant phenology is mediated by a range of physiological mechanisms. Drought can decrease soil moisture content, limiting the plant’s water absorption capacity, thereby suppressing physiological processes such as photosynthesis, respiration, and nutrient transport [155]. These physiological changes ultimately lead to the delay or acceleration of plant organ development. Drought is often accompanied by high temperatures, and the synergistic effects of these two factors can cause more severe stress on plants [156]. High temperatures can exacerbate transpiration, further exacerbating water deficits and affecting plant growth and development. Under drought conditions, the content and signaling pathways of plant growth-regulating hormones, such as abscisic acid, can change [157]. The disruption of these hormonal signals can alter phenological rhythms. Drought can promote the reallocation of nutrients, such as carbon and nitrogen, within the plants, changing their distribution patterns among different organs, thereby affecting the growth rates of the organs [158]. Drought may also alter the interactions between plants and their pollinators, seed dispersers, and other biotic agents, thereby affecting phenological processes such as flowering and fruiting [159].

The differences in how phenology responds to drought are not just an ecological curiosity; they have important implications for drought detection and monitoring. Phenological changes can serve as early indicators of drought stress, providing critical information for resource management and agricultural planning. For instance, the timing of flowering and fruiting can directly affect crop yields, making phenological data essential for predicting agricultural outcomes under changing climatic conditions [160]. Recent advances in remote sensing technologies have enabled the integration of phenological data into drought monitoring frameworks. By utilizing satellite-derived phenological metrics, such as the timing of green-up or senescence, researchers can detect changes in vegetation health and water stress across large spatial scales [161]. These remote sensing applications can enhance traditional drought monitoring methods, which typically rely only on meteorological data, by integrating biological responses to water availability [162]. Moreover, phenological data can improve our understanding of ecosystem dynamics and resilience in the face of climate variability. For example, studies have shown that ecosystems with flexible phenological responses are more resilient to drought because they can adapt their growth patterns to make better use [163] of available resources. This adaptability may play a crucial role in maintaining ecosystem functions and services under increasing drought frequency and severity.

5.2. Incorporating Phenological Variability

As reported above, drought monitoring methods based on vegetation growth conditions are influenced by the growth cycle of plants. Both long-term trends in plant phenology and short-term disturbances can impact the accuracy of drought detection. Some drought indices are calculated by comparing the current vegetation growth status to historical growth data from multiple years. This approach helps determine whether vegetation is experiencing drought stress that affects its normal growth. It assumes that the normal growth process and pattern of vegetation are consistent every year. However, with changes in vegetation phenology, this assumption no longer holds, making it challenging to distinguish whether the detection results are due to normal phenological changes in vegetation or due to the effect of drought stress. Therefore, if drought is to be accurately detected, it is necessary to effectively identify and predict the dynamic process of vegetation phenology and exclude this effect from the detection algorithm. Furthermore, understanding drought indices becomes challenging when utilizing a static plant life cycle as a comparative baseline, as these indices are susceptible to interference from dynamic phenological changes. Several drought indices are affected by the variability in vegetation phenology. For example, the use of VIs, such as the NDVI and EVI, relies on the greenness and canopy structure of vegetation. Changes in vegetation phenology can lead to fluctuations in the drought indices derived from these VIs, making it difficult to distinguish between drought-induced stress and natural phenological variations caused by climatic, ecological, and physiological factors [164,165,166,167]. Prominent examples of such indices include AVI, VCI, and VHI. For example, the AVI uses the multi-year average NDVI as a baseline reference, comparing the current NDVI against this average to assess the severity of drought. However, due to evolving phenology, the baseline for a given year is not immutable. Consequently, the disparity between the current NDVI and the global mean NDVI is an inadequate representation of the impact of drought stress on vegetation. Adjustments are necessary to align the effect with genuine phenology. Similarly, drought indices using thermal remote sensing, which rely on the measurement of LST, can be influenced by changes in vegetation phenology. As vegetation undergoes phenological changes, such as leaf emergence or senescence, the surface temperature may vary, impacting the accuracy of drought monitoring through thermal remote sensing. Furthermore, the use of optical remote sensing for drought monitoring can be affected by the timing of phenological events. For example, if a drought occurs during the peak growing season when vegetation is at its maximum greenness, the detection of drought-induced stress may be challenging due to the limited sensitivity of optical sensors to subtle changes in vegetation health during this period.

Recent research has focused extensively on how vegetation phenology characteristics and their variations influence remote sensing for drought monitoring. Several studies have focused on understanding the relationship between vegetation phenology and its influence on the accuracy of remote sensing drought monitoring. For example, Wang and Zhang conducted a comprehensive analysis of the impact of vegetation phenology changes on the performance of remote sensing drought monitoring techniques [19]. Their study highlighted the importance of considering phenological variations in vegetation when interpreting remote sensing data for drought assessment. Smith et al. and Li et al. investigated the impact of vegetation green-up and senescence on the detection of drought using remote sensing data such as MODIS-derived VIs [23,168]. Their findings show that variations in vegetation phenology significantly affected the performance of drought monitoring methods based on VIs. They also suggested that the accuracy of assessing drought conditions can be improved by considering the timing and duration of vegetation growth stages. To mitigate issues arising from phenological variability, some scholars have proposed phenology-adjusted drought indices. For instance, Li et al. introduced the Weighted Vegetation Condition Index (WVCI), a phenology-based index that accounts for varying vegetation sensitivity to drought by assigning specific weights to different phenological stages [18]. In addition, Johnson and Brown explored the use of VIs derived from satellite imagery to monitor drought conditions [169]. Their study revealed that changes in vegetation phenology, such as delayed or early onset of green-up, can serve as indicators of drought stress, thereby enhancing the effectiveness of remote sensing-based drought monitoring. Liu et al. examined the potential of integrating vegetation phenology information into drought monitoring models [20]. By incorporating phenological metrics into remote sensing-based drought indices, they demonstrated improved accuracy in detecting and monitoring drought conditions. Zhang et al. examined the impact of vegetation phenology on thermal-based drought monitoring using Landsat data [21]. The researchers observed that changes in vegetation phenology influenced the relationship between LST and vegetation water stress, highlighting the importance of considering phenological variations in thermal remote sensing-based drought monitoring. Wang et al. further examined how vegetation phenology affects the use of optical remote sensing for monitoring drought conditions [22]. The study demonstrated that the timing of phenological events, such as leaf onset and senescence, could introduce uncertainties in drought monitoring, emphasizing the need to consider phenological variability in such methods.

Overall, there have been many studies that collectively underscore the significance of considering vegetation phenology characteristics and their variations in remote sensing-based drought monitoring. By integrating such information into monitoring models and techniques, more accurate and reliable assessments of drought conditions can be achieved. Furthermore, the inclusion of plant phenology dynamics in drought monitoring frameworks enables a more holistic understanding of ecosystem responses to water stress, thereby improving the effectiveness of drought early warning systems. Although remote sensing-based phenological monitoring has advanced significantly, and plant phenology is increasingly utilized for drought detection, challenges remain in fully harnessing phenology for effective drought monitoring. These include the integration of plant phenology data into drought monitoring systems, the need for improved accuracy in differentiating drought-induced phenological changes from other environmental influences, and the integration of multiple data sources for comprehensive drought assessments. Addressing these challenges is crucial for enhancing the reliability and applicability of plant phenology for drought detection.

5.3. Example Demonstration

To better understand the impact of phenological changes on drought monitoring, and to construct and apply drought indices that consider phenological changes, we utilized the WVCI method proposed by Li et al. [18] to map drought in the northern part of China.

The calculation process for WVCI is as follows:

WVCI_i = mean(5VCI_iave(a) + 7VCI_iave(b) + VCI_iave(c)

where VCI_iave(a), VCI_iave(b), and VCI_iave(c) are the average VCI values for groups (a), (b), and (c) in the specific year i, respectively. The word mean is the mean function. The values of VCI vary from 0 to 100, with lower values indicating poor vegetation growth and greater levels of drought. Essentially, the WVCI accounts for the varying sensitivity of vegetation to drought by assigning specific weights to different phenological stages. Similar to VCI, the values of WVCI range from 0 to 100, where a smaller value indicates poorer vegetation growth conditions and a higher degree of drought. Figure 6 shows the schematic diagram for the calculation of WVCI.

All the phenological metrics were extracted through TIMESAT V3.3 (available from https://web.nateko.lu.se/timesat/timesat.asp) (accessed on 5 October 2024), developed by Jönsson and Eklundh [170]. Figure 7 shows a screenshot of the TIMESAT V3.3 graphical user interface.

Based on the extracted phenological parameters, we obtained the phenology-based WVCI, and the distribution maps for 2010 and 2015 are shown in Figure 8a,b, respectively.

6. Discussion and Conclusions

Drought is a complicated phenomenon influenced by a multitude of factors, including meteorological conditions, soil characteristics, hydrological status, and plant ecosystems. Vegetative drought is specifically defined in terms of the water stress experienced by vegetation, which can be detected through various indicators such as moisture levels, vegetation growth and health, temperature, and evapotranspiration. As a physiological parameter, plant phenology is important in the detection of vegetative drought.

With advances in Earth observation technology, remote sensing has become widely used to detect vegetative drought and extract phenological metrics. This review provides a comprehensive overview of remote sensing approaches for monitoring vegetative drought and phenology, with a focus on satellite-based optical and thermal remote sensing. Although many research teams are active in this field, the complex relationship between vegetative drought and phenology has received little attention to date. As these studies advance in their respective directions, the significance of the interaction between phenology and vegetative drought will become clearer, leading to enhancements in current methods for detecting vegetative drought. We explored this field and presented the common methods used. As a result, we conclude that there is an urgent need to develop improved drought detection methods that consider phenological variability.

We first reviewed the advances in optical and thermal remote sensing for detecting vegetative drought and phenology. Summarizing the existing literature, we divided the commonly used methods for remote sensing of vegetative drought into six categories according to their pertinent variables: moisture, vegetation condition, temperature, ET, feature space, and combination. Additionally, we examined two classic categories of methods: those based on VI time series and cumulative temperature methods, which are used in remote sensing of LSP.

Drought changes the characteristics of vegetation growth periods; thus, monitoring vegetation growth status can help detect vegetative drought. Vegetation condition-based methods are straightforward for detecting vegetative drought. However, vegetation condition-based methods generally exhibit a certain time lag due to the delayed response of the vegetation index to water shortage. In comparison, when responding to drought, vegetation temperature changes occur earlier and are more sensitive than changes in the vegetation index. Soil moisture and ET are not always directly related to vegetation water deficit. Therefore, these methods are typically used to create composite indices for detecting vegetative drought.

In the context of remote sensing of LSP, VI time-based methods are widely used. These methods are easy to implement and can capture broad-scale phenological patterns. Nonetheless, they are susceptible to atmospheric effects, and they may not capture fine-scale phenological events; furthermore, they can be influenced by vegetation heterogeneity. Cumulative temperature methods require accurate atmospheric correction, and they may not work well in regions with complex topography or land cover. However, they can provide information on the thermal environment driving plant development and can capture earlier phenological events.

Following these reviews, we analyzed the current challenges in remote sensing of vegetative drought. This analysis focused on the effects of vegetation physiology, the impact of drought on phenology, and how phenological changes affect drought detection. Additionally, we examined the necessity and feasibility of developing phenology-informed methods in conjunction with existing preliminary research. We also provided a drought analysis example using the WVCI method proposed by other researchers, focusing on the northeastern and Inner Mongolia regions of China.

In the short term, drought significantly impacts vegetative physiology, leading to variations in phenological metrics. These metrics form the foundational basis for several drought indices, such as the AVI and VCI. However, plant phenology is also influenced by various factors, including climatic and environmental conditions, in addition to drought. The interactions between drought and phenology present challenges for the remote sensing of vegetative drought. Incorporating phenological metrics, such as the SOS and EOS, represents a promising approach for developing new detection models that are resilient to phenological variability.

The core objective of these models is to mitigate phenological variability through various strategies, such as the WVCI proposed by Li et al. [18]. Nevertheless, addressing the complexities of phenological variability remains challenging, and further advancements are necessary. Several potential directions for future research include the following:

Extension to Phenology-Tolerant Drought Indices: Incorporate SOS and EOS into existing drought indices (e.g., VCI, TCI) to enhance their drought detection performance. For instance, a delayed SOS may indicate water stress, while an early EOS can signal increased vulnerability to drought. Understanding the effects of phenological variability on drought development will facilitate the refinement of drought detection models;

Longitudinal Analysis: Conduct analyses of long-term datasets to observe trends in phenological metrics and their correlation with drought events, thereby enhancing our understanding of how vegetation responds to changing climatic conditions;

Case Studies and Validation: Implement case studies across diverse ecosystems to validate the effectiveness of incorporating SOS and EOS into drought detection models. Comparing model outputs with ground-truth data can provide valuable insights into the practical implications of these metrics;

AI and Deep Learning Approaches: Utilize machine learning algorithms to analyze the relationship between phenological metrics and drought indices. By training models on historical data, we can improve drought detection and prediction capabilities based on observed phenological changes.

7. Research Prospects

This work uniquely examines the recent advancements in optical and thermal remote sensing of vegetative drought and LSP while exploring the interconnectedness between these two areas—phenology and vegetative drought—which has been relatively overlooked in previous research. In summary, this review provides a comprehensive and in-depth examination of the interconnections among satellite optical and thermal remote sensing, vegetative drought, and LSP. The aim is to advance the understanding and monitoring of these critical environmental phenomena.

Remote sensing techniques, including optical and thermal remote sensing, provide valuable data for detecting early signs of vegetative stress and tracking changes in LSP. These tools enable researchers and managers to better understand the impacts of drought on vegetation and ecosystems, allowing them to make informed decisions about resource management and conservation efforts. The integration of remote sensing data with advanced analytical methods has opened new possibilities for studying the complex relationships among climate, vegetation, and land-surface dynamics. As a result, remote sensing of vegetative drought and LSP continues to play a crucial role in monitoring environmental changes and supporting sustainable land-use practices.

When detecting vegetative drought with remote sensing, it is important to consider both the long-term effects of global climate change on plant phenology and the short-term impacts of extreme weather events, such as drought, on plant phenology. Furthermore, changes in phenology must be taken into account when using remote sensing for drought detection. Variations in phenology can greatly affect the interpretation of VIs, affecting the accuracy of drought monitoring and assessment. Therefore, incorporating phenological considerations into the analysis is crucial for obtaining reliable and comprehensive insights into drought conditions.

The impact of drought on phenology and the subsequent influence of phenological changes on remote sensing-based drought detection present significant challenges and complexities. Consequently, the overlap between remote sensing for drought monitoring and remote sensing for phenology raises several issues that need further exploration and understanding. Addressing these challenges will require interdisciplinary research efforts that integrate expertise in remote sensing, meteorology, ecology, and climate science.

Future research should focus on developing innovative approaches to integrate phenological information into existing drought monitoring systems, improving the accuracy and reliability of remote sensing-based drought assessments. Additionally, there is a need to explore the potential of leveraging advanced machine learning and artificial intelligence techniques to enhance the capability of remote sensing technologies in capturing and interpreting the intricate relationships between drought dynamics and phenological variations. By addressing these research gaps, we can enhance our capacity to monitor and respond to drought events with greater precision and effectiveness, ultimately contributing to a more resilient environment and promoting sustainable development.