Early detection of vegetation stress in Nairobi national park: structural change analysis from 2005 to 2025

Nairobi National Park (NNP), a rare urban wildlife sanctuary bordering Kenya’s capital, is experiencing accelerating habitat destruction due to urban expansion, infrastructure development, and shifting land tenure. This study applies a two-decade satellite-based spatial analysis (2005–2025) to assess vegetation disturbance and ecological thresholds across the park. Using monthly MODIS NDVI data and the BFAST framework, the present study detected abrupt structural changes in vegetation dynamics that traditional linear based trend-based vegetation indices analysis methods failed to capture. The study further compares inflection points of commonly used vegetation indices such as kNDVI, NIRV, and LAI with breakpoint markers to show time lag before change signals are recorded when the indices are used alone. The seasonal-trend model used the split sample technique where part of the data was used for training, with break detection applied to the other end of the timeseries data. Results show that nearly one-third of the park’s vegetation pixels (about 30%) experienced sudden changes in condition between 2005 and 2025. The year 2020 marked the most active period, with 201 pixels showing abrupt shifts and the highest rate of vegetation greening at 76.1%. In contrast, 2018 and 2023 recorded the most intense vegetation decline, with browning rates of 92.7% and 97.2%, respectively. These fluctuations reflect alternating cycles of ecological stress and recovery, with annual change intensity ranging from −0.107 in 2014 (severe decline) to +0.047 in 2016 (moderate recovery). The BFAST method consistently detected short-term vegetation shocks that were missed by standard statistical tools like Mann-Kendall tests and linear regression. In many cases, BFAST identified structural breakpoints up to 3 weeks before visible changes appeared in conventional vegetation indices. This early detection capability highlights BFAST’s value as a diagnostic tool for monitoring rapid ecological shifts and informing early action conservation responses.

1 Introduction

1.1 Background

Nairobi National Park (NNP), established in 1946 as Kenya’s first national park, spans approximately 117 km² and is one of the few protected areas globally located adjacent to a capital city, Nairobi, one of Africa’s fastest-growing urban centres. This unique proximity positions NNP as a critical refuge for biodiversity, hosting endangered species such as black rhinos, lions, giraffes, and over 500 bird species, while supporting essential ecological processes like material cycling, water regulation, and carbon sequestration (Mwangi et al., 2022; Lesilau et al., 2021; Mungai et al., 2025). Vegetation, as a fundamental component of the park’s terrestrial ecosystem, serves as a key indicator of ecological health, reflecting changes driven by both natural and anthropogenic factors (Geng et al., 2019).

Historically, the park’s unfenced southern boundary facilitated seasonal wildlife migration between NNP and the broader Athi-Kapiti ecosystem, ensuring resource access during dry seasons (Lesilau et al., 2021). However, rapid urban expansion, including major infrastructure projects such as the Standard Gauge Railway (SGR) and Southern Bypass, alongside expanding human settlements, has led to progressive habitat fragmentation (Mwangi et al., 2022; Mungai et al., 2025). These developments have obstructed critical migration corridors, notably affecting large mammals like wildebeests, lions, and giraffes, and intensified human-wildlife conflicts, including increased livestock predation and retaliatory killings (Lesilau et al., 2021). For instance, a recent spatial ecology study found that lion home ranges in NNP average only 49 km², significantly smaller than in other protected areas, with movement patterns constrained by human disturbances, including nighttime shifts in activity to avoid human presence and frequent roaming into community lands during wet seasons (Lesilau et al., 2021).

The anthropogenic pressures on NNP have led to significant changes in its vegetation composition and structure, with invasive species proliferation and degradation of critical habitats such as grasslands and riverine woodlands threatening the park’s ecological integrity and its role as a carbon sink and urban green space (Mungai et al., 2025). These changes reflect broader landscape dynamics, including both gradual shifts (e.g., vegetation degradation) and abrupt changes (e.g., habitat conversion due to infrastructure), which require robust monitoring to understand their ecological impacts (Geng et al., 2019; Hui et al., 2024). To effectively monitor these changes, this study employs the Normalized Difference Vegetation Index (NDVI), which provides a straightforward and effective measure of vegetation health and density. NDVI’s simplicity and robustness to minor atmospheric and soil background variations make it a suitable choice for this context, where the vegetation types are relatively diverse but not excessively complex, and more complex indices like SAVI and EVI may not offer significant advantages (Geng et al., 2019; Pandey et al., 2024).

Satellite remote sensing through vegetation indices, e.g., NDVI offers an insightful approach for monitoring landscape and vegetation dynamics due to its high spatial coverage and temporal continuity (Lesilau et al., 2021). The Breaks for Additive Season and Trend (BFAST) algorithm, which decomposes time-series data into trend, seasonal, and remainder components, is particularly suited for detecting both gradual and abrupt changes in vegetation without relying on arbitrary thresholds (Lesilau et al., 2021). For example, BFAST can identify shifts in vegetation cover due to urban expansion or infrastructure development, as well as seasonal variations, providing insights into the timing and magnitude of changes. In the context of NNP, where urban pressures threaten ecological connectivity and habitat quality, such tools are essential for tracking landscape changes and informing conservation strategies to sustain the park’s ecological functions amidst Nairobi’s rapid growth.

1.2 Significance of the park to Nairobi’s population

NNP holds immense ecological, socio-cultural, and economic importance for Kenya, particularly for the urban residents of Nairobi (Mwangi et al., 2022; Lesilau et al., 2021; Mungai et al., 2025). As a unique protected area located within a capital city globally, NNP serves as an ecological buffer against increasing urbanization. The park is home to a diverse range of wildlife, including lions, rhinos, giraffes, and over 500 bird species, inhabiting various habitats such as grasslands, forests, wetlands, and riverine habitats (Mwangi et al., 2022; Ogutu et al., 2013).

Ecologically, the park provides critical ecosystem services, such as carbon sequestration, temperature regulation, and flood mitigation, contributing to Nairobi’s climate resilience (Ogega et al., 2019; Ongaga, 2024). Its location near the central business district makes it an accessible site for research and monitoring of species dynamics and environmental changes, including insect biodiversity (Denlinger, 1980) and wildlife dispersal patterns (Ogutu et al., 2013; Lesilau et al., 2021).

The park serves as a centre for environmental education and heritage, offering opportunities for urban residents and students to engage with nature and contribute to conservation awareness (Gichohi, 2000). However, the park’s semi-fenced structure, while preserving habitat, limits the movement of migratory species, particularly impacting lions with smaller home ranges (Omondi, 1984; Gichohi, 2000). Long-term land-use changes have also reduced the functionality of dispersal areas, especially to the south.

In addition to direct conservation benefits, tourism in NNP stimulates broader economic activity, boosting Nairobi’s reputation as a global eco-tourism destination and supporting livelihoods in surrounding communities. Domestic tourism has increased among urban and peri-urban populations, representing a significant portion of park visitors. NNP also plays a vital role in Kenya’s tourism economy, generating revenue through entry fees, safari tours, and educational experiences, and supporting job creation directly in the park and indirectly in associated sectors such as lodging and transport (Okello et al., 2012; Mwangi et al., 2022).

The park serves as an ecological safeguard for Nairobi by regulating hydrology and improving air quality (Ogega et al., 2019). Its vegetated areas help absorb stormwater and mitigate the urban heat island effect. With increasing concerns about climate change and land degradation, the green infrastructure functions of NNP are becoming more essential for urban resilience.

Despite its value, the park is threatened by fencing, encroachment, and major infrastructure projects like roads and railways (Omondi, 1984; Mwangi et al., 2022; Ogutu et al., 2013). These developments have fragmented habitats and reduced corridor connectivity, particularly to the Athi-Kapiti plains, which have long been a crucial dispersal route for herbivores. The anthropogenic pressures on NNP have led to significant changes in its vegetation composition and structure, with invasive species proliferation and degradation of critical habitats such as grasslands and riverine woodlands threatening the park’s ecological integrity and its role as a carbon sink and urban green space (Mungai et al., 2025).

Therefore, monitoring vegetation dynamics and detecting early signs of vegetation stress is important step for informing conservation strategies and sustaining the park’s ecological functions amidst Nairobi’s rapid growth.

1.3 Statement of the problem

Vegetation indices have been used for a long time to evaluate changes in land use and monitor ecosystem health, with a primary focus on the condition of vegetation and how it responds to environmental factors (Mwangi et al., 2022; Fang, et al., 2018). However, these indices often assume a linear relationship between vegetation and the factors that influence it, ignoring the complexities introduced by rapid human-induced disturbances such as changes in land use, expansion of infrastructure, and degradation of ecosystems (Mendes et al., 2022; Verbesselt et al., 2012; Verbesselt et al., 2010a; He et al., 2024). While traditional vegetation indices are useful for measuring the health of vegetation, they do not account for sudden changes in the trend of, e.g., NDVI caused by external pressures (Geng et al., 2019; DeVries et al., 2015; Dutrieux et al., 2015).

Previous studies on the NNP have mainly concentrated on assessing the health of vegetation without pinpointing when significant environmental changes occurred or if these changes are scientifically linked to urban growth and fragmentation of habitats (Omondi, 1984; Mwangi et al., 2022; Ogutu et al., 2013). Despite the ecological importance of the park, existing research has not yet provided a detailed analysis that can distinguish sudden structural changes from seasonal variations.

This study addresses the forementioned gaps by employing the BFAST algorithm to detect and analyse both abrupt and gradual vegetation changes in NNP using time-series NDVI data. Specifically, it aims to pinpoint the timing and magnitude of significant environmental changes, such as those driven by urban expansion and infrastructure development, and distinguish them from seasonal variations, overcoming the limitations of traditional vegetation indices (Geng et al., 2019). The study offers a detailed temporal and spatial analysis of vegetation dynamics, linking structural changes to specific anthropogenic drivers such as the large infrastructural development projects.

This approach directly responds to the shortcomings of conventional index-based methods, which have struggled to detect abrupt ecological shifts in dynamic urban ecosystems. The resulting insights will strengthen understanding of NNP’s ecological responses to urban pressures and support targeted conservation strategies to mitigate habitat fragmentation and preserve ecological connectivity.

1.4 Objectives of the study

The overarching aim of this study is to assess long-term structural changes and ecosystem fragmentation within NNP over a 20-year period from 2005 to 2025. Unlike previous studies that primarily focused on species distribution, land use, or static habitat classification, this research applies a time-series remote sensing approach to detect both gradual and abrupt shifts in vegetation structure, which serve as key proxies for ecosystem function across spatial and temporal scales.

The study seeks to pinpoint the timing of ecosystem structural shifts, provide quantitative evidence of these transformations, and determine the anthropogenic and environmental drivers that influence vegetation dynamics and habitat integrity over time, features of relevance to interest groups and policymakers. To address this gap, it employs methodical approaches capable of detecting abrupt vegetation changes, thereby enabling a more comprehensive understanding of how human activities and ecological factors shape long-term ecosystem transformations.

The present study integrates advanced remote sensing techniques and geospatial analytics to strengthen data-driven conservation strategies, inform policy decisions, and support sustainable management efforts aimed at mitigating further habitat degradation within the park.

2 Methodology study area and data collection

2.1 Study area

The study focuses on Nairobi National Park, located approximately 7 km from the city centre of Nairobi. The park lies at the intersection of urban expansion and savannah biodiversity just at the eastern part of the city (Figures 1a-c).

Figure 1

Figure 1. (a) shows Map of Kenya’s national parks in Kenya, highlighting Nairobi National Park in red and its location within the country. (b) Land cover map of Nairobi National Park, showing that grassland (70 km²) and shrubland (13 km²) dominate the landscape. Tree cover occupies approximately 11 km². Notably, small-scale farming activities occur within the park, covering an area of about 0.6 km². 1 (c) presents the NPP map overlaid on OpenStreetMap, highlighting road networks in orange and yellow. The new expressway and the Southern Bypass are visible traversing the northern edge of the map.

2.2 Data collection

This study utilized Moderate Resolution Imaging Spectroradiometer (MODIS) data products, specifically, MODIS Terra NDVI composites (MOD13A1 Version 6) at a spatial resolution of 250 m and a 16-day revisit period. The data was sourced from NASA’s Land Processes Distributed Active Archive Centre (LP DAAC). The data was used to analyse long-term vegetation dynamics across Nairobi National Park from 2005 to 2025. The data were selected for their temporal density, radiometric consistency, and ability to mitigate cloud contamination through maximum-value compositing. To ensure spatial consistency with regional datasets, all MODIS HDF files were converted to GeoTIFF format and reprojected from the native sinusoidal grid to the WGS84/UTM Zone 37S coordinate system using the MODIS Reprojection Tool (MRT). Only the NDVI bands and associated quality assurance layers were retained for analysis.

MODIS was selected due to its strong suitability for time-series ecological analysis in cloud-prone environments like Nairobi. Its high revisit frequency and relatively coarse spatial resolution allow for consistent, gap-minimized monitoring of seasonal and interannual vegetation trends, capabilities especially relevant for methods such as Breaks for Additive Season and Trend (BFAST) that rely on continuous, low-noise temporal signals (Almeida et al., 2018; Fang et al., 2018). The 250 m resolution also balances detail and coverage, enabling robust detection of greenness trends while preserving computational efficiency in large spatial and temporal analysis (Geng et al., 2019).

3 Methodology

3.1 Package setup and dependency management

To ensure reproducibility and streamline setup, a modular R environment was established using a custom helper function pkgTest (), which installs and loads necessary libraries. Core dependencies included MODISTools for MODIS data retrieval, terra and raster for spatial data handling, and bfast, zoo, and ggplot2 for time-series analysis and visualization. This setup facilitated platform-independent processing and reproducible research workflows.

3.2 NDVI acquisition and download automation

MODIS Terra NDVI data (MOD13Q1, Version 6; 250 m; 16-day composites) were downloaded for Nairobi National Park using the MODISTools package in R. The product includes red and NIR reflectance bands and a scaled NDVI layer. A 2 × 2 km extraction window was centered within the park to ensure full coverage. Pixel reliability data were also retrieved for quality control. NDVI values, stored as scaled integers (−2000 to +10,000), were converted to raw values (−0.2 to +1.0) using the MODIS-specified scale factor to restore biophysical meaning and support accurate vegetation analysis. Raw NDVI digital numbers were converted using the MODIS-specified scale factor (Equation 1):

$NDV{I}_{true}=\frac{DN}{10,000}(1)$

where:

• DN is the digital number from the MODIS NDVI band,

• $NDV{I}_{true}$ is the scaled vegetation index value.

3.2.1 Supplementary vegetation indices

In addition to NDVI, three supplementary vegetation indices, that is kNDVI, NIRV, and LAI were incorporated to enhance the sensitivity and robustness of vegetation change detection. These indices were selected for their complementary responses to environmental drivers and their ability to isolate climate-induced variability. kNDVI (kernel NDVI) captures nonlinear vegetation responses and is particularly effective in arid and semi-arid landscapes where traditional NDVI may saturate or misrepresent subtle changes (Guo et al., 2024; Sun et al., 2024). NIRV (Near-Infrared Reflectance of Vegetation) integrates NDVI with near-infrared reflectance to better represent photosynthetic activity while minimizing soil background interference (Badgley et al., 2019). LAI (Leaf Area Index) provides a direct measure of canopy structure and biomass and is strongly correlated with hydroclimatic variables such as evapotranspiration and precipitation, making it a reliable indicator for filtering out transient climate effects (Chen et al., 2021; Kipkemoi et al., 2021). These indices were retrieved from MODIS products and processed using similar spatial and temporal parameters as NDVI to ensure comparability across time-series analyses.

3.3 Raster conversion, QA masking, and data cleaning

MODIS data were converted into time-sequenced raster stacks using MODIS data were converted into time-sequenced raster stacks using mt_to_terra () within the terra package in R. To ensure data quality, the pixel reliability layer was used to exclude low-quality observations, retaining only those with reliability scores of 0 (good) and 1 (marginal). Filtering was guided by three criteria: spatially restricting the dataset to pixels within Nairobi National Park (NNP), aligning observations chronologically, and applying a quality mask to remove cloud-contaminated or noisy data. These steps ensured that the resulting NDVI series were clean and suitable for time-series modeling.

3.4 Time-series transformation and visualization

To prepare NDVI data for trend and breakpoint analysis, pixel-level values were converted into a time-series format using a custom wrapper function, timeser (), which extracted relevant data and interpolated dates into decimal year format. The timestamp was computed as Equation 2:

$t=y+\frac{d-1}{\Delta d·N}(2)$

where $t$ is the decimal year, y the calendar year, $d$ the Julian day, $\Delta d$ the temporal frequency, and $N$ the number of steps per year. This transformation ensured compatibility with algorithms like BFAST, which require evenly spaced observations. Visual inspection using ggplot2 and leaflet validated NDVI trends in both time and space. The final series were exported as clean, temporally regular objects for Mann–Kendall trend analysis and structural breakpoint detection.

3.5 Integration of BFAST explorer

This study integrates BFASTspatial and BFAST Explorer to detect and interpret structural changes in vegetation dynamics within Nairobi National Park. BFAST decomposes a time series into trend, seasonal, and residual components, represented as:

${\mathrm{Y}}_{\mathrm{t}}={\mathrm{T}}_{\mathrm{t}}+{\mathrm{S}}_{\mathrm{t}}+{\mathrm{e}}_{\mathrm{t}}\left(\mathrm{t}=1,\dots ,\mathrm{n}\right)(3)$

where ${\mathrm{Y}}_{\mathrm{t}}$ is the observed NDVI, ${\mathrm{T}}_{\mathrm{t}}$ the long-term trend, ${\mathrm{S}}_{\mathrm{t}}$ the seasonal component, and ${\mathrm{e}}_{\mathrm{t}}\left(\mathrm{t}=1,\dots ,\mathrm{n}\right)$ the noise term. Breakpoints correspond to structural changes in either trend or seasonality (Equation 3).

This study applies BFAST algorithms through BFASTspatial to detect and characterize vegetation changes in Nairobi National Park, using MODIS NDVI data spanning the past 2 decades. The integration of BFAST Explorer enhanced the interpretability of detected breakpoints by providing a more nuanced understanding of habitat fragmentation, ecological disturbances, and long-term environmental transformations within Nairobi National Park. To ensure robust trend modeling and reliable breakpoint detection, the study employed a split-sample approach. Specifically, monthly MODIS NDVI data from 2000 to 2013 were used to train the seasonal–trend model, establishing a stable baseline of vegetation dynamics under relatively consistent conditions. Breakpoint detection was then applied to the subsequent period from 2014 to 2025, allowing the model to identify significant structural changes in vegetation patterns that may be linked to urban expansion, infrastructure development, or other anthropogenic pressures. This separation between training and detection periods helped minimize bias and ensured that the model could effectively isolate genuine shifts from seasonal fluctuations or long-term trends.

The trend component is estimated as a piecewise linear function segmented by breakpoints as given in Equation 4:

${\mathrm{T}}_{\mathrm{t}}={\alpha }_{\mathrm{j}}+{\beta }_{\mathrm{j}}\mathrm{t}(4)$

Where ${\alpha }_{\mathrm{j}}$ and ${\beta }_{\mathrm{j}}$ denote the intercept and slope of the $j$-th segment. The magnitude of abrupt change across breakpoints is quantified as shown in Equation 5:

$\text{Magnitude}=\left({\alpha }_{j-1}-{\alpha }_j\right)+\left({\beta }_{j-1}-{\beta }_j\right)\mathrm{t}(5)$

The seasonal component is modeled using harmonic function as shown in Equation 6:

${\mathrm{S}}_{\mathrm{t}}={\displaystyle \sum _{\mathrm{k}=1}^{\mathrm{k}}}\left[{\gamma }_{\mathrm{j},\mathrm{k}}\sin \left(\frac{2\pi \text{kt}}{\mathrm{f}}\right)+{\theta }_{\mathrm{j},\mathrm{k}}\cos \left(\frac{2\pi \text{kt}}{\mathrm{f}}\right)\right]\left({\tau }_{\mathrm{i}-1}^{\#}<\mathrm{t}<{\tau }_{\mathrm{i}}^{\#}\right)(6)$

where $K=3$ harmonic terms and $f=12$ for monthly NDVI. The selection of three harmonics balances model complexity and the need to represent distinct wet and dry season cycles characteristic of the region.

The BFAST procedure iteratively (i) decomposes the series, (ii) applies a MOSUM-based structural change test, and (iii) estimates trend and seasonal parameters using robust M-estimation. Breakpoints are selected using the Bayesian Information Criterion (BIC), ensuring an optimal balance between model fit and parsimony.

Analyses were performed in R using the bfast and bfastSpatial packages. All parameters followed standard defaults except for the segment length parameter $h$, which controls breakpoint sensitivity. Following empirical recommendations and the assumption that meaningful ecological changes occur over periods of ≥2 years, we set $h=\frac{1}{7}$, aligning with Almeida et al. (2018) and Fang et al. (2018).

The inclusion of BFAST Explorer enhances interpretation by supporting interactive visualization of breakpoint timing and magnitude, improving contextual understanding of habitat disturbance and vegetation transitions across the park landscape.

4 Results

4.1 Magnitude and direction of NDVI change

As shown in Figure 2a, which summarizes the timing of NDVI change events, the spatial distribution of change timing follows a clear gradient: blue tones represent earlier years (e.g., 2014–2016), while red tones indicate more recent change events (e.g., 2020–2025). This highlights a clear temporal trend, with later disturbances clustering particularly in the southern and eastern zones of the park.

Figure 2

Figure 2. (a) Time of NDVI change events; (b) Magnitude and spatial distribution of abrupt NDVI changes, 2014–2025. NDVI change values indicate either degradation (negative) or abrupt greening (positive). Magnitudes range from −0.238 to +0.158. Notable changes were observed in 2014, 2015, and 2018.

Figure 2b presents the magnitude and direction of these changes. Here, brown areas reflect negative NDVI changes, suggesting vegetation degradation, whereas green areas indicate positive NDVI changes, signifying greening or vegetation recovery. The observed change magnitudes range from −0.238 to +0.158, with more intense degradation concentrated near the park’s southern boundary.

The temporal distribution of these events varied significantly over the 12-year period. The highest number of change events was recorded in 2020 (201 pixels), followed by 2021 (132 pixels) and 2019 (120 pixels). The earliest and latest years 2014 (20 pixels) and 2025 (3 pixels), showed the fewest disturbances (Table 1). The clustering of NDVI breakpoints between 2018 and 2021 corresponds to a period of heightened human activity and infrastructure development near the park, including the construction phases of the Nairobi Expressway along the eastern boundary (Taylor, 2020).

Table 1

Table 1. A summary of the annual change statistics is presented in Table 1.

NDVI magnitude and direction changes are indicated by either positive or negative change, the negative values indicate degradation, while abrupt greening was indicated by positive values. The change magnitude values varied from −0.238 to +0.158 observed in 2014, 2015, 2018, 2019, 2023, and 2024, whereas net greening was observed in 2016, 2017, 2020, 2021, 2022, and 2025. The annual average ΔNDVI values varied from year to year.

The greatest percentage of sudden browning was observed in years like 2023 (97.2%), 2018 (92.7%), and 2019 (71.7%), indicating extensive deterioration. other hand, greening answers predominated in 2021 (96.9%), 2022 (95.1%), and 2025 (100%) discrete ecological stages characterized by deterioration and recovery. The complete yearly breakdown of magnitude of change is shown in Figure 3 and Table 1.

Figure 3

Figure 3. Spatial distribution of NDVI change magnitudes from 2014 to 2020 in Nairobi National Park’s buffer zones. The map highlights persistent browning in the northwestern tip throughout the period, with intensified browning concentrated in the southern region during 2016, 2017, and 2020.

Figure 3 show clear browning on the northwestern tip in 2014–2020, browning in concentrated more on the southern per in 2016, 2017 and 2020.

4.2 Temporal pattern of change types

To better understand how vegetation has changed over time, the summaries of annual changes are presented in Figure 4 showing the yearly proportions of browning and greening (Figure 4). The results reveal a clear cyclical trend: a period of degradation in 2018–2019, followed by a recovery phase from 2020 to 2022, and early signs of renewed decline in 2023–2024.

Figure 4

Figure 4. Annual proportions of browning and greening from 2018 to 2024: (a) illustrating temporal patterns of vegetation change in Nairobi National Park’s buffer zones, and (b) Z-scores overlay the trends, highlighting statistically significant periods of degradation (2018–2019), recovery (2020–2022), and emerging decline (2023–2024).

The proportion data reveal that 2018 and 2023 were the most browning-dominant years, with over 90% of pixels exhibiting negative NDVI change. In contrast, 2016, 2021, and 2022 recorded exceptionally high greening proportions, exceeding 90%, indicating widespread vegetation recovery. To assess the statistical significance of these observations, we examined the z-scores associated with each year’s NDVI change. The z-scores were compared to a critical threshold of ±1.96, corresponding to a 95% confidence level. Years like 2016 and 2021 exhibited high positive z-scores for greening (z > 1.96), indicating statistically significant vegetation recovery. Conversely, 2018 and 2023 showed strongly negative z-scores for greening (z < −1.96) and high z-scores for browning (z > 1.96), confirming them as years with significant vegetation stress (Figure 5). These results suggest that the observed patterns are unlikely to be due to chance and are instead indicative of real changes in vegetation dynamics.

Figure 5

Figure 5. NDVI trend analysis with structural breakpoints and inflection markers. The plot displays the LOESS-smoothed NDVI time series (gray line) fitted with a linear trend (dark green), alongside BFAST-detected breakpoints (red dashed lines). Black dots mark NDVI inflection points where the vegetation response becomes visually apparent. Curved green arrows connect each breakpoint to its corresponding response, illustrating the consistent lag between structural change and NDVI recognition. Mann–Kendall results (τ = −0.055, p < 1.55 × 10⁻⁸) and Sen’s slope (=0) suggest a statistically significant but weak downward trend overall. The integration of break detection with trend analysis offers a more sensitive and temporally resolved view of ecological change in Nairobi National Park.

The integration of z-scores into the temporal analysis highlights not just the direction of change, but its magnitude relative to long-term variability. This dual perspective underscores the non-linear and episodic nature of vegetation response in protected landscapes.

4.3 Temporal clustering and spatial distribution of breaks

The pixel-wise analysis of NDVI structural breaks using BFAST reveals non-random temporal clustering of vegetation change events within Nairobi National Park (NNP). As shown in Figure 6 and summarized in Table 2, the most intense periods of abrupt change occurred in 2018, 2020–2022, and 2023, highlighting intervals of ecosystem stress interspersed with recovery signals. In particular, the year 2020 stands out with the highest number of pixels (201) experiencing abrupt NDVI changes, a majority of which (>76%) indicate greening, suggesting potential rebound or re-vegetation dynamics following preceding stress periods.

Figure 6

Figure 6. Shows Year-by-year NDVI magnitude change temporal clustering and spatial distribution of breaks between the start of monitoring 2014 to the current year 2025.

Table 2

Table 2. NDVI response timing relative to structural breakpoints detected in NDVI using BFAST.

These breakpoints are not evenly distributed across the park. Spatial mapping of the detected changes indicates a higher concentration of browning events near the park’s eastern boundary, particularly between 2018 and 2023. This aligns closely with known anthropogenic activity, notably the Nairobi Expressway construction corridor, which skirted the eastern edge of NNP starting in mid-2020. In contrast, central and southwestern zones of the park exhibit more frequent greening breaks, especially during 2021 and 2022, potentially reflecting improved vegetation conditions during periods of reduced human interference.

The temporal aggregation of NDVI breakpoints also suggests a cumulative impact of environmental pressures, where years such as 2018 and 2023 show high percentages of negative NDVI change (92.66% and 97.22%, respectively), reflecting widespread vegetation degradation. Conversely, years such as 2016, 2021, and 2022 show predominantly positive changes (over 90%), likely indicating transient ecological recovery.

These temporal clusters and spatial disparities point to the need for localized monitoring strategies. They also illustrate the utility of BFAST in capturing the episodic and spatially differentiated nature of ecosystem responses, which are often masked in conventional trend analyses.

5 Discussion

5.1 Vegetation structural changes and the value of break detection approaches

This study demonstrates the ongoing fragmentation of NNP, with land loss driven largely by infrastructure expansion along its northwestern tip and the southern boundary. Roads, railways, and fencing have disrupted natural dispersal zones, posing threats to wildlife movement and vegetation integrity. Future conservation planning must prioritize ecological connectivity, compatible land use in buffer zones, and alignment with Nairobi’s broader urban development strategies (Ogega et al., 2019; Taylor, 2020). Integrated spatial planning, improved land governance, and stakeholder-inclusive conservation models will be essential to safeguarding the park’s long-term ecological function.

Understanding vegetation dynamics in NNP requires more than detecting simple directional trends; it demands sensitivity to sudden, episodic structural changes that reflect ecosystem stress, degradation, or recovery. Traditional methods such as linear regression and the Mann–Kendall trend test are limited in that they assume stationarity or monotonic progression, thereby missing signals such as short-term collapses or temporary greening pulses (Mendes et al., 2022; Verbesselt et al., 2012; Kipkemoi, 2024). In contrast, the BFAST method decomposes NDVI time series into seasonal, trend, and noise components, making it well-suited for detecting abrupt and non-stationary shifts.

In this study, results from the Mann–Kendall test revealed a weak but statistically significant downward trend in NDVI (τ = −0.055, p < 1.55 × 10⁻⁸), while Sen’s slope estimation returned a value of zero, suggesting little to no sustained directional drift (Figure 6). Visually, the LOESS-smoothed NDVI signal appeared relatively stable over time. However, BFAST uncovered multiple structural breakpoints, including episodes in 2018 and 2023, where canopy declines were both abrupt and ecologically meaningful. These moments would have otherwise remained obscured within trend-based tests, highlighting BFAST’s superior diagnostic sensitivity.

Figure 6 illustrates this contrast. While linear models flatten the NDVI trajectory, BFAST breakpoints and delayed inflection markers (black dots) emphasize a clear pattern: visible vegetation responses consistently lag statistically detected structural change. This further validates the utility of breakpoint methods in anticipating vegetation shifts earlier than conventional indices allow (Wu et al., 2020).

This example of NNP show that by distinguishing gradual from abrupt transitions, BFAST enables conservation managers to detect critical thresholds before they are externally visible, offering actionable lead time for response. In fragmented, peri-urban parks like NNP, where disturbance can be sudden and non-linear, this capability is essential for evidence-based conservation and timely intervention.

5.2 Interpreting temporal and spatial patterns of change

The temporal clustering of abrupt NDVI changes observed between 2018 and 2023 suggests periods of heightened ecological sensitivity. The most extensive disturbance, in terms of pixel count and magnitude, was recorded in 2020, a year coinciding with the peak of Nairobi Expressway construction, which involved heavy machinery, vegetation clearance, and increased human movement near the park’s boundaries. Although the expressway itself did not traverse the park’s core, its proximity, especially near the Ole Sereni Hotel and East Gate area, likely introduced edge effects that impacted vegetation integrity, particularly through dust deposition, noise, and altered drainage.

The high incidence of greening events in 2021 and 2022 (over 95%) may reflect short-term recovery, potentially due to reduced disturbance during the COVID-19 pandemic lockdowns, seasonal rainfall improvements, or internal successional dynamics in disturbed patches. This recovery, however, appears spatially localized, and is subsequently offset by the widespread browning in 2023, suggesting a possible overshoot-rebound cycle or new disturbance regime.

Spatially, the western and southeastern edges of the park exhibit the most consistent browning signals, particularly in proximity to infrastructure corridors. In contrast, central and southern interior zones tend to reflect more stable or positively trending NDVI patterns, potentially due to reduced exposure to edge effects and relatively undisturbed ecological processes.

5.3 Evaluating structural break detection and methodological sensitivity across vegetation indices

Figure 7a presents the LAI time series for Nairobi National Park (NNP), showing both raw observations and LOESS-smoothed trends. Five structural breakpoints, i.e., April 2012 April 2014, July 2016 July 2018, and June 2020, were identified using the BFAST algorithm. These breakpoints represent statistically significant shifts in trend or seasonality that cannot be attributed to noise alone. Because BFAST applies an internal hypothesis-testing framework, each breakpoint marks a validated structural change in vegetation dynamics.

Figure 7

Figure 7. (a) LAI time series (raw values and loess in bold) with BFAST breakpoints; (b) NDVI and commonly used vegetation indices with inflection markers and response delays.

To isolate climatic influences and strengthen causal attribution, Figure 7b compares NDVI, kNDVI, and NIRV, using LOESS smoothers to highlight general trends while retaining the raw NDVI signal in the background. The same BFAST breakpoints are overlaid as red dashed lines, followed by delayed but observable NDVI inflections, marked by bold black dots, occurring days to weeks later. Curved green arrows visually reinforce the temporal lag between ecological disturbance onset and its detectability in traditional indices.

Each vegetation index offers distinct sensitivity to environmental drivers. NDVI broadly reflects canopy greenness but is influenced by seasonal and drought variability (Badgley et al., 2019). kNDVI captures nonlinear vegetation responses and enhances detection under climatic stress, particularly in arid and semi-arid regions (Guo et al., 2024; Ukasha et al., 2022). NIRV integrates NDVI with near-infrared reflectance to better represent photosynthetic activity, reducing soil background interference and improving robustness under variable light and moisture conditions (Badgley et al., 2019). LAI, meanwhile, directly measures canopy structure and biomass, correlating strongly with hydroclimatic variables such as evapotranspiration and precipitation (Chen et al., 2021).

By analyzing convergence and divergence across these indices, especially during known infrastructure development periods, the study isolates climate-driven anomalies and strengthens attribution of BFAST-detected breakpoints to anthropogenic or ecological disturbances.

These findings confirm that while commonly used vegetation indices capture general ecosystem behavior, they do not reliably pinpoint the exact timing of structural shifts (Wu et al., 2020; Pandey et al., 2024). Table 2 further supports this, showing time lags of 18–25 days between each BFAST breakpoint and the corresponding NDVI response. Detecting such subtle but critical transitions is especially important in dynamic landscapes like NNP, where ecological thresholds may be crossed rapidly under shifting land use, hydrological regimes, political pressures, or policy changes (Mwangi et al., 2022).

This pattern demonstrates the efficacy of BFAST for fine-scale vegetation change detection, particularly in ecosystems where non-monotonic or threshold-based dynamics are common. Unlike regression-based trend methods, which rely on continuous slope changes and often smooth over discrete events, BFAST enables the temporal isolation of transient ecological shifts. This advantage aligns with findings from previous studies (e.g., Verbesselt et al., 2010b; De Jong et al., 2011), which argue that slope-based trend analyses can obscure disturbance events and delay detection.

By statistically anchoring structural breaks in a time series, BFAST enables conservation managers to quantify not only whether vegetation is changing, but also when and how abruptly. This diagnostic precision is vital in peri-urban conservation zones like NNP, where management interventions must often respond to complex, rapidly unfolding ecological or land use pressures.

5.4 Management and policy implications

The findings from this study have direct implications for adaptive park management and ecological policy, particularly in the context of Nairobi National Park (NNP). The detection of concentrated browning zones along the park’s northwestern tip and southern border points to areas undergoing abrupt ecological change. These changes are closely associated with proximity to infrastructure developments such as the Nairobi Expressway. This pattern reinforces the concerns raised by Mwangi et al. (2022), who identified infrastructure-park conflict, human-wildlife conflict, and land-use change as critical challenges facing the Kenyan National Park System.

Given these pressures, there is a pressing need to reassess and potentially expand buffer zones around NNP, especially in areas adjacent to urban infrastructure. However, the Kenya Wildlife Service (KWS), which is responsible for managing the park, operates under significant geographic, political, and financial constraints (Mwangi et al., 2022; Ogutu et al., 2013). In this context, tools such as BFAST, which can detect and attribute ecological changes to specific time periods and spatial locations, become particularly valuable. By providing objective, high-resolution evidence of sudden ecological shifts, the workflow can support KWS in making data-driven decisions and in advocating for expanded support and collaboration. The integration of BFAST outputs into management planning can also strengthen partnerships with local communities, non-governmental organizations, and county governments by offering a shared scientific basis for conservation action.

The identification of specific years, namely, 2018, 2020, and 2023 as periods of abrupt change further underscores the utility of BFAST for event-based ecological monitoring. These years should be prioritized for targeted field assessments focusing on vegetation health, species composition, encroachment patterns, and soil conditions. Anchoring management interventions to these temporal breakpoints allows for more responsive and effective conservation strategies, including reforestation, erosion control, and fire risk mitigation not only in Kenya, but also in other contexts with limited real time surveillance budgets.

Moreover, the spatially explicit nature of BFAST outputs enables the design of zoned management plans that align with localized vegetation dynamics. Areas identified as degradation hotspots can be prioritized for active restoration, while more stable zones may be designated for long-term monitoring and ecosystem service valuation. This spatial differentiation enhances the efficiency of resource allocation and can improve stakeholder engagement by linking ecological conditions to tangible management actions.

The study supports the integration of BFAST and similar temporal decomposition techniques into the standard monitoring frameworks for other protected areas in Kenya other global contexts. As anthropogenic pressures intensify, such tools offer a cost-effective and scientifically robust means of tracking ecological resilience. Their adoption can facilitate the development of proactive policies that anticipate and mitigate environmental degradation, rather than merely responding to its consequences.

5.5 Potential for future research

Future research could extend this approach to other protected areas across Africa and globally to examine whether similar early-warning signals of vegetation disturbance can be consistently identified. Applying the methodology to ecosystems such as Serengeti (Tanzania), Kruger (South Africa), Tsavo (Kenya), or Etosha (Namibia), as well as protected savanna–woodland mosaics in the Sahel, would enable comparative assessment of how climate variability, land-use pressures, wildlife movement, and seasonal grazing shape vegetation dynamics. Such cross-site analyses would help differentiate localized disturbances from broader regional ecological trends and test the generalizability of the BFAST-based framework beyond Nairobi National Park.

The ability of this method to detect structural changes several weeks before they become visible in conventional vegetation indices underscores its potential for operational early-action monitoring. Higher-resolution datasets (Sentinel-2, Landsat 8/9, PlanetScope, or UAV imagery) could refine detection of rapid disturbance events such as locust invasions in East Africa, tree die-off from insect pests like the mountain pine beetle and emerald ash borer in North America, and sudden canopy degradation following disease outbreaks. Similarly, applying this approach to assess the spatial extent and recovery trajectories of fire events in remote protected areas could support more timely and targeted ecological management. Integrating BFAST outputs into routine park surveillance workflows would therefore enable proactive habitat protection and faster intervention in response to emerging ecological threats.

6 Conclusion

This study applied the BFAST framework to evaluate vegetation changes using monthly MODIS NDVI data at 1 km spatial resolution. The seasonal–trend model was trained on data from 2000 to 2013, with structural break detection performed for the 2014–2025 period. The results show that approximately 30% of NDVI-observed pixels experienced one or more abrupt vegetation changes during the monitoring period. Breakpoint frequencies varied across years, peaking in 2020 when 201 pixels recorded break events alongside the highest proportion of greening (76.1 percent). The most severe abrupt browning episodes occurred in 2018 and 2023, at 92.7 percent and 97.2 percent respectively. Annual change intensities ranged from −0.107 in 2014 to +0.047 in 2016, a pattern consistent with cycles of ecological stress and recovery. By contrast, 2025 exhibited very few breakpoints, likely due to edge effects associated with the temporal proximity of the final available MODIS NDVI observations (March 2025).

Traditional trend metrics, including Mann–Kendall tests and linear regression-derived temporal trends, did not capture these intra-annual disruptions or structural breaks. While these methods indicated largely monotonic or gradually greening trends across the park, BFAST revealed multiple short-term vegetation shocks that would otherwise remain undetected or detected later in the time series. These findings reaffirm the importance of using breakpoint-oriented approaches when analysing ecosystems affected by short-lived disturbances, human pressures, and climatic variability, particularly when aiming for early action management approaches.

The advantages of BFAST become even clearer when compared with a broader suite of commonly used trend-analysis methods. LOESS and LOWESS smoothing, for example, are useful for visualizing non-linear behaviour (Cleveland, 1979) but lack statistical tests for abrupt change and can smooth over disturbance signals. Non-parametric tests such as Spearman’s Rank Correlation (Spearman, 1961) and the Mann–Kendall test (Mann, 1945; Kendall, 1975) are effective for detecting monotonic trends yet cannot accommodate reversals, multiple-phase trajectories, or sudden shocks. Similarly, the Theil–Sen slope estimator (Theil, 1992; Sen, 1968) provides a robust estimate of a single linear trend but still assumes that this trend remains constant throughout the time series. This assumption is often unsuitable for remote-sensing applications, where each pixel represents its own time series and where multiple neighboring pixels collectively form a mosaic of distinct temporal behaviors (Georganos et al., 2017). Even ordinary linear regression is limited by its assumptions of linearity and stationarity. In contrast, as stated before, BFAST (Verbesselt et al., 2010a; Verbesselt et al., 2010b) decomposes the time series into trend and seasonal components and statistically identifies structural breaks in each. This enables the detection of abrupt, non-linear, and non-stationary changes that these other approaches either mask or overlook. As demonstrated in this study, BFAST is particularly suitable for remote-sensing applications in disturbance-prone ecosystems such as NNP, where vegetation dynamics frequently exhibit short-lived shocks and complex recovery patterns.

Data availability statement

Publicly available datasets were analyzed in this study. This data can be found here: https://www.earthdata.nasa.gov/data/instruments/modis.

Author contributions

IK: Conceptualization, Data curation, Formal Analysis, Investigation, Methodology, Project administration, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review and editing.

Funding

The author(s) declared that financial support was not received for this work and/or its publication.

Acknowledgements

The author gratefully acknowledges the initial guidance of Prof. Katerina Michaelides, who contributed to the early brainstorming of ideas on nonlinear monitoring of habitat changes in the Horn of Africa during the author’s PhD studies at the University of Bristol, United Kingdom.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

References

Almeida, A. E., Menini, N., Verbesselt, J., and Torres, R. S. (2018). “BFAST explorer: an effective tool for time series analysis,” in IGARSS 2018 proceedings - IEEE international geoscience and remote sensing symposium, 4913–4916. doi:10.1109/IGARSS.2018.8517877

CrossRef Full Text | Google Scholar

Badgley, G., Anderegg, L. D., Berry, J. A., and Field, C. B. (2019). Terrestrial gross primary production: using NIRV to scale from site to globe. Glob. Change Biology 25 (11), 3731–3740. doi:10.1111/gcb.14729

PubMed Abstract | CrossRef Full Text | Google Scholar

Chen, D., Zhang, P., Seftigen, K., Ou, T., Giese, M., Barthel, R., et al. (2021). Hydroclimate changes over Sweden in the twentieth and twenty-first centuries: a millennium perspective. Geogr. Ann. Ser. A. 103 (2), 103–131.

CrossRef Full Text | Google Scholar

Cleveland, W. S. (1979). Robust locally weighted regression and smoothing scatterplots. J. Am. Statistical Association 74 (368), 829–836. doi:10.1080/01621459.1979.10481038

CrossRef Full Text | Google Scholar

De Jong, R., de Bruin, S., de Wit, A., Schaepman, M. E., and Dent, D. L. (2011). Analysis of monotonic greening and browning trends from global NDVI time-series. Remote Sens. Environ. 115 (2), 692–702.

CrossRef Full Text | Google Scholar

Denlinger, D. L. (1980). Seasonal and annual variation of insect abundance in the Nairobi national park, Kenya. Biotropica 12 (2), 100–106. doi:10.2307/2387725

CrossRef Full Text | Google Scholar

DeVries, B., Verbesselt, J., Kooistra, L., and Herold, M. (2015). Tracking disturbance-regrowth dynamics in tropical forests using satellite time series. Remote Sens. Environ. 160, 144–155. doi:10.1016/j.rse.2015.08.020

CrossRef Full Text | Google Scholar

Dutrieux, L. P., Verbesselt, J., Kooistra, L., and Herold, M. (2015). Monitoring Forest Cover Loss Using Multiple Data Streams, a Case Study of a Tropical Dry Forest in Bolivia. ISPRS J. Photogramm. Remote Sens.. doi:10.1016/j.isprsjprs.2015.03.015

CrossRef Full Text | Google Scholar

Fang, X., Zhu, Q., Ren, L., Chen, H., Wang, K., and Peng, C. (2018). Large-scale detection of vegetation dynamics and their potential drivers using MODIS images and BFAST: a case study in Quebec, Canada. Remote Sens. Environ. 206, 391–402. doi:10.1016/j.rse.2017.11.017

CrossRef Full Text | Google Scholar

Geng, L., Che, T., Wang, X., and Wang, H. (2019). Detecting spatiotemporal changes in vegetation with the BFAST model in the qilian Mountain region during 2000–2017. Remote Sens. 11 (2), 103. doi:10.3390/rs11020103

CrossRef Full Text | Google Scholar

Georganos, S., Abdi, A. M., Tenenbaum, D. E., and Kalogirou, S. (2017). Examining the NDVI-Rainfall relationship in the semi-arid Sahel using geographically weighted regression. J. Arid Environ. 146, 64–74. doi:10.1016/j.jaridenv.2017.06.004

CrossRef Full Text | Google Scholar

Gichohi, H. (2000). “Functional relationships between parks and agricultural areas in East Africa: the case of Nairobi national park,” in Wildlife conservation by sustainable use. Editors H. H. T. Prins, J. G. Grootenhuis, and T. T. Dolan (Springer), 141–167.

Google Scholar

Guo, B., Zhang, R., Lu, M., Xu, M., Liu, P., and Wang, L. (2024). A new large-scale monitoring index of desertification based on kernel normalized difference vegetation index and feature space model. Remote Sens. 16 (10), 1771. doi:10.3390/rs16101771

CrossRef Full Text | Google Scholar

He, L., Shan, Y. Q., Liu, C., Cao, H., Liu, X. N., and Guo, Y. (2024). Prediction of bedload transport inside vegetation canopies with natural morphology. J. Hydrodynamics 36 (3), 556–569. doi:10.1007/s42241-024-0033-7

CrossRef Full Text | Google Scholar

Hui, Y., Tang, Y., Yang, Q., and Mochida, A. (2024). Numerical study on influence of surface vegetation on aerodynamics of high-rise buildings. Sustain. Cities Soc. 107, 105407. doi:10.1016/j.scs.2024.105407

CrossRef Full Text | Google Scholar

Kendall, M. G. (1975). Rank correlation methods. Griffin.

Google Scholar

Kipkemoi, I. (2024). Impacts of climate and climate change on water and vegetation dynamics in Horn Of Africa drylands. Doctoral dissertation, University of Bristol.

Google Scholar

Kipkemoi, I., Michaelides, K., Rosolem, R., and Singer, M. B. (2021). Climatic expression of rainfall on soil moisture dynamics in drylands. Hydrology Earth Syst. Sci. Discuss. 2021, 1–24. doi:10.5194/hess-2021-48

CrossRef Full Text | Google Scholar

Lesilau, F., Verschueren, S., van’t Zelfde, M., Musters, K. C. J. M., de Snoo, G. R., and de Iongh, H. H. (2021). Spatial ecology of lions in a small, semi-fenced park surrounded by dense human populations: the case study of Nairobi national park, Kenya. Glob. Ecol. Conservation 26, e01449. doi:10.1016/j.gecco.2021.e01449

CrossRef Full Text | Google Scholar

Mann, H. B. (1945). Nonparametric tests against trend. Econ. J. Econometric Society 13, 245–259. doi:10.2307/1907187

CrossRef Full Text | Google Scholar

Mendes, M. P., Rodriguez-Galiano, V., and Aragones, D. (2022). Evaluating the BFAST method to detect and characterise changing trends in water time series: a case study on the impact of droughts on the mediterranean climate. Sci. Total Environ. 846, 157428. doi:10.1016/j.scitotenv.2022.157428

PubMed Abstract | CrossRef Full Text | Google Scholar

Mungai, I. M., Gichuki, N., Sigana, D. A., Agwanda, B., Chiyo, P., Obanda, V., et al. (2025). Drivers of rodent community structure in an Urban National Park, Kenya. PloS One 20 (4), e0321659.

PubMed Abstract | CrossRef Full Text | Google Scholar

Mwangi, F., Zhang, Q., and Wang, H. (2022). Development challenges and management strategies on the Kenyan national park system: a case of Nairobi national park. Int. J. Geoheritage Parks 10 (1), 16–26. doi:10.1016/j.ijgeop.2022.02.003

CrossRef Full Text | Google Scholar

Ogega, O. M., Wanjohi, H. N., and Mbugua, J. (2019). “Exploring the future of Nairobi national park in a changing climate and urban growth,” in The geography of climate change adaptation in urban Africa. Editors P. Elias, and R. Ajibade (Springer International Publishing), 249–272. doi:10.2174/1874839201307010011

CrossRef Full Text | Google Scholar

Ogutu, J. O., Owen-Smith, N., Piepho, H. P., Said, M. Y., Kifugo, S. C., Reid, R. S., et al. (2013). Changing wildlife populations in Nairobi national park and adjoining athi-kaputiei plains: collapse of the migratory wildebeest. Open Conservation Biol. J. 7 (1), 11–26.

Google Scholar

Okello, M. M., Kenana, L., and Kieti, D. (2012). Factors influencing domestic tourism for urban and semiurban populations around Nairobi national park, Kenya. Tour. Anal. 17 (1), 79–89. doi:10.3727/108354212x13330406124214

CrossRef Full Text | Google Scholar

Omondi, P. (1984). The impact on Nairobi national park of changes in land use in adjacent areas. Doctoral dissertation, University of Nairobi.

Google Scholar

Ongaga, C. O., Makokha, M., Obiero, K., Kipkemoi, I., and Diang’a, J. (2024). Urbanization and hydrological dynamics: a 22-year assessment of impervious surface changes and runoff in an urban watershed. Front. Water 6, 1455763. doi:10.3389/frwa.2024.1455763

CrossRef Full Text | Google Scholar

Pandey, A., Mondal, A., Guha, S., Upadhyay, P. K., and Kundu, S. (2024). Comparing the seasonal relationship of land surface temperature with vegetation indices and other land surface indices. Geol. Ecol. Landscapes 9, 1–17. doi:10.1080/24749508.2024.2392391

CrossRef Full Text | Google Scholar

Sen, P. K. (1968). Estimates of the regression coefficient based on Kendall's tau. J. Am. Statistical Association 63 (324), 1379–1389. doi:10.2307/2285891

CrossRef Full Text | Google Scholar

Spearman, C. (1961). The proof and measurement of association between two things. Stud. Individual Differences Search Intelligence., 45–58. doi:10.1037/11491-005

CrossRef Full Text | Google Scholar

Sun, H., Ma, X., Liu, Y., Zhou, G., Ding, J., Lu, L., et al. (2024). A new multiangle method for estimating fractional biocrust coverage from Sentinel-2 data in arid areas. IEEE Trans. Geoscience Remote Sens. 62, 1–15. doi:10.1109/tgrs.2024.3361249

CrossRef Full Text | Google Scholar

Taylor, I. (2020). Kenya’s new lunatic express: the standard gauge railway. Afr. Stud. Q. 19 (34), 29–52. Available online at: https://journals.flvc.org/ASQ/article/view/136004.

Google Scholar

Theil, H. (1992). “A rank-invariant method of linear and polynomial regression analysis,” in Henri Theil’s contributions to economics and econometrics: econometric theory and methodology (Netherlands: Springer), 345–381.

Google Scholar

Ukasha, M., Ramirez, J. A., and Niemann, J. D. (2022). Temporal variations of NDVI and LAI and interactions with hydroclimatic variables in a large and agro-ecologically diverse region. J. Geophys. Res. Biogeosciences 127 (4), e2021JG006395. doi:10.1029/2021jg006395

CrossRef Full Text | Google Scholar

Verbesselt, J., Hyndman, R., Newnham, G., and Culvenor, D. (2010a). Detecting trend and seasonal changes in satellite image time series. Remote Sens. Environ. 114 (1), 106–115. doi:10.1016/j.rse.2009.08.014

CrossRef Full Text | Google Scholar

Verbesselt, J., Hyndman, R., Zeileis, A., and Culvenor, D. (2010b). Phenological change detection while accounting for abrupt and gradual trends in satellite image time series. Remote Sens. Environ. 114 (12), 2970–2980. doi:10.1016/j.rse.2010.08.003

CrossRef Full Text | Google Scholar

Verbesselt, J., Zeileis, A., and Herold, M. (2012). Near real-time disturbance detection using satellite image time series. Remote Sens. Environ. 123, 98–108. doi:10.1016/j.rse.2012.02.022

CrossRef Full Text | Google Scholar

Wu, L., Li, Z., Liu, X., Zhu, L., Tang, Y., Zhang, B., et al. (2020). Multi-type forest change detection using BFAST and monthly landsat time series for monitoring spatiotemporal dynamics of forests in subtropical wetland. Remote Sens. 12 (2), 341. doi:10.3390/rs12020341

CrossRef Full Text | Google Scholar