Seasonal Drought Dynamics in Kenya: Remote Sensing and Combined Indices for Climate Risk Planning

Abstract

Drought is a pervasive and intensifying climate hazard with profound implications for food security, water availability, and socioeconomic stability, particularly in sub-Saharan Africa. In Kenya, where over 80% of the landmass comprises arid and semi-arid lands (ASALs), recurrent droughts have become a critical threat to agricultural productivity and climate resilience. This study presents a comprehensive spatiotemporal analysis of seasonal drought dynamics in Kenya for June–July–August–September (JJAS) from 2000 to 2024, leveraging remote sensing-based drought indices and geospatial analysis for climate risk planning. Using the Standardized Precipitation Evapotranspiration Index (SPEI), Vegetation Condition Index (VCI), Soil Moisture Anomaly (SMA), and Fraction of Absorbed Photosynthetically Active Radiation (FAPAR) anomaly, a Combined Drought Indicator (CDI) was developed to assess drought severity, persistence, and impact across Kenya’s four climatological seasons. Data were processed using Google Earth Engine and visualized through GIS platforms to produce high-resolution drought maps disaggregated by county and land-use class. The results revealed a marked intensification of drought conditions, with Alert and Warning classifications expanding significantly in ASALs, particularly in Garissa, Kitui, Marsabit, and Tana River. The drought persistence analysis revealed chronic exposure in drought conditions in northeastern and southeastern counties, while cropland exposure increased by over 100% while rangeland vulnerability rose nearly 56-fold. Population exposure to drought also rose sharply, underscoring the socioeconomic risks associated with climate-induced water stress. The study provides an operational framework for integrating remote sensing into early warning systems and policy planning, aligning with global climate adaptation goals and national resilience strategies. The findings advocate for proactive, data-driven drought management and localized adaptation interventions in Kenya’s most vulnerable regions.

1. Introduction

Drought is among the most complex and widespread natural disasters globally, and its impacts are particularly severe in sub-Saharan Africa, where it has become a recurrent threat to sustainable development. In this region, droughts disrupt food systems, water availability, and socioeconomic stability, often leading to humanitarian crises. Kenya, as a representative case in Eastern Africa, is especially vulnerable due to its high dependence on rain-fed agriculture and its exposure to climate variability and land degradation [1]. Recurrent drought events have not only intensified in frequency and severity in recent decades, but they have also become increasingly unpredictable, challenging conventional monitoring and response mechanisms [2].

More than 80% of Kenya’s landmass falls within arid and semi-arid lands (ASALs), inhabited by pastoral and agropastoral communities whose livelihoods are particularly sensitive to rainfall variability [3]. These regions often experience below-average rainfall and are ecologically fragile, with limited adaptive capacity. Compounding this vulnerability are land-use changes, deforestation, unsustainable water abstraction, and population growth. Such pressures increase exposure to climate risks and decrease ecosystem resilience, making drought not just an environmental hazard but a deeply embedded socioeconomic challenge [4]. Kenya’s climate is characterized by a bimodal rainfall regime, with the long rains (March to May (MAM)) and the short rains (October to December (OND)) being the primary agricultural seasons. The failure of these rains can lead to widespread crop failure, livestock deaths, water scarcity, increased malnutrition, and disease outbreaks [5]. The OND season, in particular, has become increasingly erratic in recent years, with failed or suppressed rainfall events causing widespread disruption. This variability is often linked to broader climatic systems such as the El Niño-Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD), which influence regional precipitation anomalies [6]. Conversely, the January–February (JFM) and June–September (JJAS) periods, although typically dry or marginally wet, provide important windows for evaluating soil moisture recovery, vegetation stress, and hydrological trends post-rainy season [7].

Historically, Kenya has relied on ground-based meteorological observations and early warning bulletins provided by institutions such as the Kenya Meteorological Department (KMD) and the National Drought Management Authority (NDMA). While these systems have improved over time, they still face challenges in spatial coverage, timeliness, and integration with local knowledge systems. The sparsity of meteorological stations in ASALs, data gaps, and bureaucratic delays in dissemination hinder rapid response and planning, especially at the sub-county and community levels [8].

In light of these limitations, remote sensing technologies and geospatial analytics are increasingly being adopted as alternative or complementary approaches for drought detection, monitoring, and mapping. Satellite-derived indicators provide spatially continuous and temporally frequent observations that are critical for understanding the onset, duration, and severity of drought events across large and data-scarce regions. Among these, the Vegetation Condition Index (VCI), derived from Moderate Resolution Imaging Spectroradiometer (MODIS) NDVI, is widely used to assess vegetation health and biomass anomalies, serving as a proxy for agricultural and ecological drought [7]. Meanwhile, the Standardized Precipitation Evapotranspiration Index (SPEI), which standardizes precipitation and evaporation deviations over selected time scales, is effective in capturing meteorological drought and has become a global standard due to its adaptability and statistical robustness [9]. With global warming, atmospheric evaporative demand (PET) is becoming an important driver of drought severity.

The use of cloud-based platforms such as Google Earth Engine (GEE) now enables large-scale processing of Earth observation data with enhanced computational efficiency. GEE’s ability to integrate multiple datasets such as MODIS, CHIRPS, and Landsat, alongside user-defined algorithms, allows for scalable drought monitoring across regions, seasons, and years [10]. Integrating SPEI and VCI allows for a multidimensional assessment of drought that accounts for both cause (rainfall deficit) and effect (vegetation stress), making it particularly suitable for heterogeneous environments like Kenya. Previous studies have demonstrated the effectiveness of this integration in East Africa. For instance, [11] used Climate Hazards Group InfraRed Precipitation with Station (CHIRPS) and NDVI-derived indices to validate drought risk zones across Ethiopia, while [12] emphasized the role of remote sensing in improving drought early warning systems in the Horn of Africa. However, these efforts have largely focused on annual or regional scales, with limited disaggregation across agro-ecological zones or seasons. To address these gaps, this study applies a Combined Drought Indicator (CDI) approach, which integrates SPEI and VCI within a unified framework, offering a more comprehensive, seasonally and spatially resolved assessment of drought.

Given the evolving risks posed by climate change, drought planning in Kenya must move beyond reactive crisis response to proactive, evidence-based risk management. Seasonal drought mapping serves this purpose by illuminating when and where water deficits are likely to occur, enabling anticipatory actions such as pre-positioning of relief supplies, adjustment of planting calendars, and conservation of grazing lands [13]. Moreover, the creation of drought atlases based on SPEI and VCI data supports integration into national contingency plans, county integrated development plans (CIDPs), and regional food security strategies. This gap is particularly concerning given that seasonal drought characteristics vary across Kenya’s counties due to differences in altitude, land cover, and local climate regimes [14]. Counties such as Turkana, Garissa, Wajir, and Tana River face recurrent early-year droughts, while Kitui, Makueni, and Machakos often experience late-year rainfall deficits.

This study seeks to generate high-resolution, seasonal drought occurrence maps for Kenya using SPEI and VCI derived from CHIRPS rainfall data and MODIS NDVI, respectively, for the period 2000–2024. It directly contributes to the global climate adaptation agenda and aligns with frameworks such as the Sendai Framework for Disaster Risk Reduction, the Paris Agreement on Climate Change, and the Sustainable Development Goals (SDGs), particularly SDG 13 on climate action and SDG 2 on food security. The resulting seasonal drought atlas serves as a decision-support tool for policymakers, humanitarian agencies, researchers, and community stakeholders in building resilience against future droughts.

2. Study Area

Kenya, located in Eastern Africa between latitudes 5° N and 5° S and longitudes 34°E and 42°E, is a country of diverse topography and climate, making it particularly susceptible to hydrological extremes such as droughts. It spans approximately 582,646 km² and consists of varied landscapes ranging from the humid coastal region to the arid and semi-arid lands that cover about 80% of the country [14]. These ASAL regions are especially vulnerable to droughts due to low and erratic rainfall, high evapotranspiration rates, and limited access to adaptive infrastructure. Kenya’s climate is primarily influenced by the Intertropical Convergence Zone (ITCZ), which drives a bimodal rainfall pattern in most parts of the country with short rains (October to December) and long rains (March to May) [15]. However, the northern and eastern ASAL counties, such as Turkana, Wajir, Mandera, and Garissa, often receive less than 500 mm of rainfall annually, making them hotspots for recurrent drought events [16,17].

Administratively, Kenya is divided into 47 counties, many of which have developed localized climate adaptation plans to address frequent climatic shocks. Despite the presence of national early warning systems, such as those coordinated by the National Drought Management Authority (NDMA), data gaps persist in understanding spatial and seasonal drought variability at local scales [18].

Kenya’s climatic variability is closely tied to its agro-ecological zones, which range from the humid highlands with annual rainfall above 1200 mm to the dry lowlands and arid northern plains that often receive less than 400 mm. The agro-ecological zones vary by rainfall and agricultural potential. Zone I is humid with high rainfall and excellent farming conditions. Zone II is sub-humid, receiving moderate rainfall and supporting medium to high agricultural productivity. Zone III is semi-arid with low to moderate rainfall and limited farming potential. Zone IV is arid, characterized by low rainfall. Zone V is very arid with minimal agricultural viability. Zone VII is transitional, lying between arid and desert climates [18]. Temperature regimes also vary significantly, from the cool highland areas such as the Aberdares and Mount Kenya, where mean annual temperatures can fall below 15 °C, to the hot and dry lowlands in Turkana and Garissa, where daytime temperatures often exceed 35 °C. These gradients in rainfall and temperature strongly influence water availability, soil moisture, vegetation cover, and ultimately the exposure and sensitivity of different regions to drought (Figure 1).

Agro-ecological zoning further illustrates Kenya’s vulnerability to drought, characterized by fragile ecosystems and limited agricultural potential. In contrast, highland and medium-potential zones in counties such as Kericho, Nyeri, and Bungoma provide more stable agricultural outputs due to reliable rainfall and favorable temperatures. The dependence of rural households on rain-fed agriculture and livestock within the ASALs underscores the relevance of mapping drought occurrence, since prolonged dry spells directly threaten livelihoods, food security, and ecosystem resilience [19].

3. Methodology

3.1. Overview

This study employs an integrated remote sensing and geospatial approach to monitor and map seasonal drought occurrence across Kenya using multi-source satellite-derived drought indices. The methodology combines precipitation anomalies, soil moisture dynamics, and vegetation condition indicators into a harmonized framework for assessing drought severity and persistence. The analysis focused on the June–September (JJAS) season for the period 2000–2024 (2000, 2010, 2020, and 2024), which were selected as temporal milestones to illustrate long-term drought shifts under increasing climate stress.

Although this proof-of-concept relies on four time steps, the approach is scalable to a continuous multi-decadal time series, which would enhance trend detection and attribution of drought patterns to climate variability and change in future work. The methodological process involved the acquisition and preprocessing of precipitation, evapotranspiration, vegetation, and soil moisture datasets, the computation of standardized drought indices (SPEI, VCI, SMA, and FAPAR anomaly), their integration into a Combined Drought Indicator (CDI) using a four-class classification scheme (Normal, Watch, Warning, and Alert), validation of CDI results against National Drought Management Authority (NDMA) bulletins, and spatial as well as statistical analyses at pixel and county levels to detect severity, persistence, and sectoral exposure to drought.

3.2. Data Sources

3.2.1. Satellite-Derived Datasets

The data sources employed in this study integrate both satellite-derived and ancillary datasets to ensure robust spatial-temporal drought analysis across Kenya. The primary remote sensing inputs include CHIRPS, MODIS NDVI, soil moisture datasets, and FAPAR-derived products. The Climate Hazards Group InfraRed Precipitation with Station Data (CHIRPS) provides precipitation data with a spatial resolution of 0.05° and a daily-to-monthly temporal scale. This dataset has been rigorously validated across East Africa and is widely utilized for computing precipitation-based drought indices such as the Standardized Precipitation Evapotranspiration Index (SPEI) due to its accuracy and historical coverage [16]. Temperature Data were acquired from MODIS (MOD11A2) land surface temperature and ERA5 reanalysis datasets. The PET Estimation was derived from Potential evapotranspiration computed using Penman–Monteith (FAO56 method).

In parallel, vegetation conditions were assessed using the Moderate Resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) data, specifically the MOD13Q1 product at a 250-m spatial resolution. This dataset was used to derive the Vegetation Condition Index (VCI), which effectively captures plant stress responses to moisture availability and is a widely adopted proxy in drought monitoring frameworks [17]. Additionally, root-zone soil moisture anomalies were extracted from the LISFLOOD hydrological model developed by the Joint Research Centre (JRC), which provides temporally consistent and physically based estimates of soil water content [19]. To complement these indicators, the study also utilized the Fraction of Absorbed Photosynthetically Active Radiation (FAPAR) anomaly, a satellite-based metric that quantifies vegetation photosynthetic efficiency and biomass response under drought conditions [20].

3.2.2. Administrative Boundaries

For spatial disaggregation and contextualization, administrative boundary data for Kenya’s counties were obtained from the International Livestock Research Institute (ILRI) and the Kenya National Bureau of Statistics (KNBS). These shapefiles facilitated precise mapping of drought severity across administrative units. To validate the satellite-derived indicators and ensure alignment with on-the-ground impacts, the study incorporated drought bulletins and situational reports from the National Drought Management Authority (NDMA). These reports served as ground truth references to assess the consistency of remotely sensed signals with observed drought occurrences, thereby enhancing the credibility and operational relevance of the analysis. A summary of all datasets used in the study is presented in

Table 1. A summary of the datasets used.

Dataset	Variable(s)	Spatial Resolution	Temporal Resolution	Period Covered	Source/Provider	Application in Study
CHIRPS v2	Precipitation	0.05° (~5 km)	Daily, monthly	1981–present	Climate Hazards Group, UC Santa Barbara	Computation of SPEI (precipitation input)
ERA5-Land	Temperature, wind speed, relative humidity, radiation	0.1° (~9 km)	Hourly, aggregated to monthly	1981–present	ECMWF Copernicus Climate Data Store	PET estimation for SPEI (Penman–Monteith)
MODIS MOD11A2	Land Surface Temperature (LST)	1 km	8-day composites	2000–present	NASA LP DAAC	Auxiliary input for PET
MODIS MOD13Q1	NDVI	250 m	16-day composites	2000–present	NASA LP DAAC	Derivation of VCI
LISFLOOD (JRC)	Root-zone soil moisture index	~5 km	Daily to monthly	1990–present	Joint Research Centre (JRC)	Soil Moisture Anomaly (SMA)
MODIS FAPAR (MCD15A3H)	Fraction of Absorbed PAR	500 m	8-day composites	2000–present	NASA LP DAAC	FAPAR anomaly computation
Administrative Boundaries	County-level shapefiles	Vector (administrative units)	Static	2019	ILRI, KNBS	Spatial aggregation and disaggregation of drought metrics
NDMA Bulletins	Observed drought conditions	County-level qualitative reports	Monthly	2000–2024	Kenya National Drought Management Authority	Ground-truthing and validation of remote sensing indicators

.

3.3. Drought Indices

3.3.1. Standardized Precipitation Evapotranspiration Index (SPEI)

The Standardized Precipitation Evapotranspiration Index (SPEI) was computed at the 3-month scale (SPEI-3) for the JJAS season. SPEI integrates both rainfall supply and atmospheric evaporative demand by standardizing the climatic water balance (precipitation minus PET) over a given accumulation period [21]. This makes it particularly useful in assessing droughts that are exacerbated by rising temperatures and other climate drivers.

For this study, CHIRPS daily precipitation data were aggregated into monthly totals, while meteorological variables for potential evapotranspiration (PET), including air temperature, wind speed, relative humidity, and surface net radiation, were obtained from ERA5-Land reanalysis. PET was calculated using the FAO-56 Penman–Monteith method, which provides a robust physically based estimation of atmospheric evaporative demand. All computations were carried out over the common climatological period of 2000, 2010, 2020, and 2024, ensuring consistency with the SPEI analysis.

SPEI uses the climatic water balance (D), defined as precipitation (P) minus potential evapotranspiration (PET), fitted to a log-logistic probability distribution and standardized (Equations (1) and (2)):

D = P − PET

(1)

$$\mathrm{S}\mathrm{P}\mathrm{E}\mathrm{I}=(\mathrm{D}\mathrm{i}-\overline{\mathrm{D}})/\mathrm{\sigma }\mathrm{D}$$

(2)

where Di is the climatic water balance for the ith accumulation period; $\overline{\mathrm{D}}$ is the long-term mean of the water balance, and σD is the standard deviation of the water balance. This allows capturing both precipitation deficits and increased evaporative demand.

PET was computed using the FAO-56 Penman–Monteith equation (Equation (3)):

PET = [0.408Δ(R_n − G) + γ(900/(T + 273))u2(e_s − e_a)]/[Δ + γ(1 + 0.34u2)]

(3)

where PET = potential evapotranspiration (mm day⁻¹), Δ = slope of the saturation vapor pressure curve (kPa °C⁻¹), R_n = net radiation at the crop surface (MJ m⁻² day⁻¹), G = soil heat flux density (MJ m⁻² day⁻¹), γ = psychrometric constant (kPa °C⁻¹), T = mean daily air temperature at 2 m height (°C).

u2 = wind speed at 2 m height (m s⁻¹), e_s = saturation vapor pressure (kPa), and e_a = actual vapor pressure (kPa). This physically based equation ensures robust estimation of atmospheric evaporative demand.

3.3.2. Vegetation Condition Index (VCI)

The VCI is derived from the MODIS NDVI time series, normalized relative to its historical maximum and minimum for each pixel to account for inter-annual variability [7]. VCI helps detect vegetation anomalies caused by drought, providing an indirect yet spatially rich view of ecological drought effects (Equation (4)).

$$NDVI={\displaystyle \frac{{NDVI}_i-{NDVI}_{min}}{{NDVI}_{max}-{NDVI}_{min}}}\times 100$$

(4)

where NDVI_i is the NDVI in a given period, and NDVI_min and NDVI_max represent the historical minimum and maximum NDVI values for the same period. A VCI < 35% was considered indicative of drought-stressed vegetation.

3.3.3. Soil Moisture Anomaly (SMA) and FAPAR Anomaly

SMA, obtained from the LISFLOOD’s soil water index, represents deviations in volumetric soil moisture from its long-term mean. It is expressed in standard deviations, with negative anomalies (typically <−1) signaling soil moisture deficits that correspond with agricultural drought phases [22]. FAPAR anomalies were computed by subtracting long-term climatological means from observed values in each 10-day composite (dekad). Persistent negative anomalies across consecutive dekads suggest ongoing vegetative stress linked to water scarcity, serving as an early proxy for drought impact on biomass productivity [20]. To account for vegetation response lag, FAPAR and VCI anomalies were aligned with SPEI values from the preceding month and dekad composites. This ensured that delayed vegetation stress, typically manifesting weeks after rainfall deficits, was captured in the CDI framework.

3.4. Combined Drought Indicator (CDI)

The Combined Drought Indicator (CDI) was applied as the central framework for synthesizing multiple drought indices into a unified classification system. Following the approach of [23], the CDI integrates the Standardized Precipitation Evapotranspiration Index (SPEI), Soil Moisture Anomaly (SMA), and the Fraction of Absorbed Photosynthetically Active Radiation (FAPAR) anomaly to capture the interaction between meteorological, soil, and vegetation responses to water stress. The integration process is based on conditional rules that assess whether precipitation deficits are translating into soil and vegetation stress, thereby distinguishing between early-warning conditions and fully developed drought events. For this study, drought severity was classified into four operational categories that are widely used in drought early warning systems: Normal, Watch, Warning, and Alert (

Table 2. Classification Scheme for Combined Drought Indicator (CDI). Criteria integrating SPEI, Soil Moisture Anomaly (SMA), and FAPAR Anomaly [23].

CDI Class	Standardized Precipitation Evapotranspiration Index (SPEI)	Soil Moisture Anomaly (SMA)	FAPAR Anomaly
Normal	SPEI ≥ −1	SMA ≥ −1	FAPAR ≥ −1
Watch	SPEI < −1	SMA ≥ −1 and/or FAPAR ≥ −1	Early warning; precipitation deficit detected, but limited impact on vegetation or soil moisture.
Warning	SPEI < −1	SMA < −1 or FAPAR < −1	One non-precipitation indicator shows stress, indicating emerging drought effects in vegetation or soil.
Alert	SPEI < −1.5	SMA < −1 and FAPAR < −1	All indicators show stress, active, and widespread drought conditions.

).

This harmonized four-class scheme was consistently applied across both the methodological framework and the results presentation, ensuring comparability and alignment with operational drought monitoring systems such as those used by the National Drought Management Authority (NDMA) in Kenya and other global early warning platforms. By adopting this classification, the CDI offers a transparent and policy-relevant tool for identifying drought onset, progression, and severity, enabling evidence-based planning and response at multiple scales.

To validate the remotely sensed CDI outputs, we compared them against historical drought reports from the National Drought Management Authority (NDMA) spanning 2000–2024. These reports provide county-level drought classifications (Normal, Alert, Alarm, Emergency, Recovery), which were reclassified into equivalent CDI categories (Normal, Watch, Warning, Alert). Cross-comparison was conducted using contingency tables to assess agreement between CDI-derived and NDMA-reported drought status. This validation step was critical in testing the operational reliability of CDI for drought early warning in Kenya.

In this study, all three CDI components, SPEI, SMA, and FAPAR anomaly, were assigned equal weights following [12]. The equal-weight approach is widely applied in operational drought monitoring frameworks when index quality is comparable across meteorological, hydrological, and vegetation dimensions. The decision avoids overemphasizing any single component and ensures that CDI classifications reflect compound stress conditions rather than meteorological anomalies alone. Equal weighting is also consistent with implementation standards in Copernicus Global Drought Observatory and regional drought monitoring systems in East Africa [24].

3.5. Tools and Workflow

The methodological workflow for this study integrated advanced geospatial and statistical tools to ensure efficient, reproducible, and accurate drought assessment and mapping. The entire process was structured around two major platforms: Google Earth Engine (GEE) for satellite data processing and extraction, and desktop-based GIS and programming tools for visualization, spatial analytics, and statistical computations.

Google Earth Engine (GEE) served as the central cloud-computing environment for all remote sensing-based data analysis. GEE’s ability to handle large-scale spatiotemporal datasets made it ideal for the computation of seasonal drought indicators. Specifically, GEE was used to preprocess and filter CHIRPS rainfall data and MODIS NDVI datasets to match the climatological focus of the study. Seasonal composites for the June–September (JJAS) periods in 2000, 2010, 2020, and 2024 were generated to enable consistent temporal comparisons. The Standardized Precipitation Evapotranspiration Index (SPEI-3) was computed directly on GEE for each target year and integrated with MODIS-derived NDVI to calculate the Vegetation Condition Index (VCI). It aggregated precipitation and PET over the JJAS 3-month window for each of the study years (2000, 2010, 2020, and 2024). All drought indices were classified into severity levels based on established thresholds and exported as cloud-optimized GeoTIFFs for further analysis. A water balance series was then created by calculating the difference between precipitation and PET. For each pixel, a log-logistic probability distribution was fitted to the long-term water balance series, and cumulative probabilities were transformed into standardized z-scores, yielding SPEI values.

Post-processing and spatial analysis of the drought layers were conducted using open-source and proprietary GIS tools. QGIS version 3.28 was utilized for spatial overlays with county boundaries, zonal statistics to quantify drought impact by administrative unit, and thematic map development. The processed layers were also subjected to vector-based extraction to isolate cropland and rangeland-specific drought patterns. In parallel, Python (v3.11) and R (v4.3) programming environments were employed for statistical evaluation of the drought metrics. This included time-series plotting, trend analysis, and correlation assessment between SPEI, VCI, and the derived Combined Drought Indicator (CDI) classes. Optional use of ArcGIS Pro 3.4 supported cartographic refinement and layout design for publication-quality map figures, particularly for comparative analysis across the four study years. These tools collectively enabled a robust and multidisciplinary approach to seasonal drought mapping in Kenya.

An annual JJAS CDI composite was generated for each year from 2000–2024 by computing CDI values for each pixel and selecting the maximum drought severity class within the JJAS season. This process produced a consistent temporal dataset suitable for both spatial frequency analysis and Mann–Kendall trend detection (Figure 2).

Trend Analysis Using Mann–Kendall and Sen’s Slope

Long-term temporal trends in drought conditions (2000–2024) were assessed using the non-parametric Mann–Kendall (MK) test and Sen’s slope estimator [24,25]. The MK test evaluates monotonic trends without assuming data normality, making it suitable for hydroclimatic variables such as SPEI, SMA, and CDI time series. Sen’s slope quantifies the magnitude of change by estimating the median rate of increase or decrease per year.

CDI time series for each county and pixel were extracted from the annual JJAS composites (2000–2024), producing 25 data points per pixel. Positive Sen’s slope values indicate a trend toward wetter conditions, while negative slopes represent intensifying drought conditions. Statistical significance was evaluated at α = 0.05 using two-tailed testing. These methods improve drought characterization by identifying long-term directional changes that might be obscured by interannual variability.

3.6. Seasonal and Interannual Analysis

Drought mapping was undertaken for the JJAS seasons of 2000, 2010, 2020, and 2024 to capture inter-decadal variations in drought conditions. For each of these years, CDI-based maps were generated to show the geographic distribution and severity of drought, highlighting persistent hotspots, newly emerging drought-affected areas, and regions showing signs of recovery. Beyond these temporal snapshots, a continuous CDI time series covering the full 2000–2024 period was developed at the pixel level (1 km) to quantify the frequency and persistence of drought events. Frequency was expressed as the number of JJAS seasons in which each pixel experienced Watch, Warning, or Alert conditions, while persistence was measured as the proportion of seasons a pixel remained under drought stress, thus identifying zones of chronic exposure. Long-term trends in drought occurrence were further evaluated using the non-parametric Mann–Kendall test together with Sen’s slope estimator to establish whether directional changes in drought patterns were statistically significant. At the county level, zonal statistics were applied to summarize drought severity and persistence across administrative boundaries.

Impact Variables

To link drought conditions with socioeconomic and ecological systems, three impact variables were included: population, croplands, and rangelands. Population exposure was derived from the 2024 WorldPop gridded population dataset at ~1 km resolution. Population counts were spatially overlaid on CDI maps to estimate the number of people falling into different drought severity classes. Croplands were extracted from the ESA Climate Change Initiative (CCI) Land Cover product (300 m resolution), while Rangelands were mapped using MODIS MCD12Q1 land cover data. For each impact variable, exposure was quantified by calculating the proportion of pixels classified under Watch, Alert, or Warning CDI categories for each analysis year (2000, 2010, 2020, and 2024). This allowed assessment of sectoral vulnerabilities to drought stress.

The percentage increase in vulnerability was computed relative to baseline exposure in the year 2000, explaining the large proportional increase observed in rangeland systems. Drought persistence was calculated as the number of JJAS seasons in which a county pixel was classified under Alert or worse categories across the study period (2000–2024), producing a persistence heatmap. This metric highlights areas facing chronic stress due to recurrent droughts.

4. Results

4.1. Individual Drought Indices

The individual drought indices revealed distinct yet complementary patterns of hydroclimatic stress across the study period, providing a multidimensional understanding of drought dynamics. The figures presented for SPEI, SMA, and VCI focus on 2000 and 2024, representing extreme drought and relatively wetter reference years, respectively, to illustrate decadal contrasts in hydroclimatic conditions. The SPEI (Figure 3) highlighted pronounced rainfall deficits in 2000, indicative of a widespread meteorological drought event that disrupted precipitation regimes across much of the basin. In contrast, 2024 displayed comparatively wetter conditions, though the improvement was not uniform, suggesting interannual variability and the influence of shifting large-scale climatic drivers. The Soil Moisture Anomaly (SMA) (Figure 4) further captured the persistence of below-average soil water availability during 2000, which was particularly severe in cultivated and semi-arid zones where evapotranspiration demands exceeded precipitation inputs. Although 2024 showed partial recovery in soil water content, localized deficits remained evident, particularly in regions dependent on rain-fed agriculture, reflecting the lagged hydrological response to precipitation anomalies. Vegetation Condition Anomaly (VCA) (Figure 5) emphasized the ecological consequences of these meteorological and hydrological extremes, with strong negative anomalies in 2000 corresponding to spatially coherent patterns of vegetation stress and degradation. The alignment of vegetation anomalies with areas of low SPEI and SMA underscored the cascading effects of climate variability on ecosystems, highlighting the sensitivity of vegetation health to coupled deficits in both rainfall and soil moisture. Together, these three independent indicators provide a robust characterization of drought occurrence and intensity, illustrating the multifaceted pathways through which climate variability propagates into ecological and agricultural impacts.

4.2. Combine Drought Indicator for Kenya

Figure 6 presents the 24-year spatial distribution of drought severity across Kenya for the June–September period in the years 2000, 2010, 2020, and 2024, based on the CDI. The CDI integrates multiple drought-relevant variables to classify conditions into four categories: Normal, Watch, Warning, and Alert. In 2000 and 2010, the majority of the country remained under Normal or Watch conditions, with only isolated pockets, primarily in the ASALs, experiencing Warning or Alert levels. However, by 2020, there was a marked intensification of drought severity, with a substantial expansion of Warning zones and the emergence of widespread Alert conditions, particularly in southeastern and eastern Kenya. By 2024, the Alert category dominates large swaths of the country, indicating a critical escalation in drought frequency and intensity. This progression suggests a systemic shift in Kenya’s hydroclimatic regime, likely driven by compounding effects of climate change, land degradation, and anthropogenic water stress. These spatial patterns underscore the growing vulnerability of key agricultural and pastoral regions, necessitating urgent policy interventions in drought preparedness, water resource management, and climate adaptation strategies.

4.3. Spatial-Temporal Dynamics of Drought Conditions

Figure 7 illustrates the relative changes in CDI-based drought severity classes across Kenya during the JJAS season for the years 2000, 2010, 2020, and 2024. The maps classify changes as improvements (green), no change (gray), or degradations (yellow to orange), measured with respect to shifts in CDI categories.

In 2000, the central and eastern parts of Kenya, particularly regions within Machakos, Kitui, and parts of Embu and Meru counties, exhibited pronounced CDI degradations, with extensive areas moving from Normal to Watch or from Watch to Warning. By 2010, modest CDI improvements were observed in western Kenya (e.g., Kakamega and Bungoma counties) and in the middle parts of the country. However, degradation persisted in the southeastern lowlands and in the northern ASAL counties. The 2020 map reveals a resurgence of drought stress, with widespread transitions into Warning and Alert categories across Garissa, Tana River, and Isiolo. The 2024 projection suggests a mixed outlook: while some recovery is anticipated in the western and central highlands (e.g., Kericho, Nyeri), severe degradations into Alert conditions remain entrenched in the eastern and coastal regions.

4.4. Drought Persistence Analysis

4.4.1. Drought Persistence

Figure 8 presents a drought persistence heatmap for Kenya spanning the period 2000–2024, offering a spatially explicit assessment of long-term drought exposure. Persistence was calculated as the proportion of time steps in which a location was classified under Warning or Alert conditions relative to the total time steps. The heatmap utilizes a color gradient from yellow (low persistence) to red (high persistence) to quantify frequency and duration.

The analysis reveals distinct regional disparities, with northeastern and southeastern Kenya exhibiting the most severe and sustained drought conditions. In the northeast, Marsabit and parts of Wajir and Garissa counties consistently show high drought persistence. Similarly, southeastern counties such as Tana River, Kitui, and Makueni demonstrate prolonged exposure. The coastal region, notably Kilifi and Lamu, also displays elevated drought persistence. In contrast, central Kenya (e.g., Nyeri, Kiambu, Murang’a) shows relatively low persistence, while western Kenya near Lake Victoria exhibits moderate persistence likely buffered by the lake’s microclimatic influence.

4.4.2. Validation with NDMA Bulletins

The CDI outputs showed strong agreement with NDMA county-level drought bulletins, particularly in identifying severe drought episodes such as those of 2009 and 2017 in northern and eastern ASAL counties. Agreement exceeded 70% for Alert and Warning classifications, confirming the operational relevance of the CDI within Kenya’s drought monitoring framework. This is consistent with previous studies demonstrating that monthly NDMA county drought bulletins provide a robust operational benchmark for validating drought indices against reported impacts, especially in the context of remote-sensing-based indicators [26].

4.5. Drought Impact Assessment Across Population and Land Use Classes (2000–2024)

Quantitative assessment of drought occurrence was undertaken, and the results in

Table 3. Comparison of Drought Impact Indicators Across CDI Categories in Kenya (2000 vs. 2024).

CDI Class	Year	Population Affected	Pop (%)	Area Affected (km2)	Area (%)	Cropland Affected (km2)	Cropland (%)	Rangeland Affected (km2)	Rangeland (%)
Alert	2000	284,469	0.78	163,463.43	0.84	284.80	0.74	10.13	0.02
	2024	330,655	1.46	320,135.88	1.63	586.53	1.54	586.15	1.30
Warning	2000	136,477	0.37	155,842.88	0.80	145.61	0.39	313.88	0.69
	2024	373,078	1.64	239,017.94	1.21	792.70	2.08	427.65	0.94
Watch	2000	3,048,781	8.30	2,288,604.68	11.66	2911.26	7.66	4978.57	10.97
	2024	2,342,630	10.31	3,394,559.03	17.30	3628.56	9.55	6388.24	14.09
Normal	2000	33,307,974	90.73	17,070,867.32	86.97	34,746.50	91.38	40,142.44	88.53
	2024	19,720,462	86.85	15,748,229.54	80.25	33,111.80	87.09	38,084.14	83.99

show the spatial and sectoral impacts of the Combined Drought Indicator (CDI) classes. The classes include Alert, Warning, Watch, and Normal, for the years 2000 and 2024, covering four key variables: total population affected, geographical area impacted (km²), cropland area, and rangeland area. Each indicator was expressed both in absolute terms and as a percentage of the total.

4.5.1. Population Exposure to Drought Conditions

The comparative analysis of drought impact indicators across CDI categories in Kenya between 2000 and 2024 highlights notable shifts in vulnerability (Figure 9). Results show an increase in affected population and land area in the Alert, Warning, and Watch categories, with population exposure rising from 0.78% to 1.46% in Alert and from 8.30% to 10.31% in Watch conditions. Similarly, land area under drought expanded, especially in the Watch category from 11.66% to 17.30%. Cropland and rangeland exposure also intensified, with cropland affected in Warning rising from 0.39% to 2.08%, while rangeland exposure in Watch increased from 10.97% to 14.09% (Table 3). Conversely, the proportion of population and area under Normal conditions declined, reflecting an increase in drought severity over time.

4.5.2. Geographic Extent of Drought

Geospatial analysis revealed that in 2000, Alert conditions affected approximately 163,463 km² (0.84%), while Warning conditions spanned 155,843 km² (0.80%). Watch zones covered 2,288,605 km² (11.66%), with the vast majority of 88.71% classified under Normal conditions (17,412,284 km²). In 2024, the spatial footprint of drought had expanded substantially. The Alert class now covers 320,136 km² (1.63%), nearly doubling its extent from 2000. The Warning category also showed an increase, reaching 239,018 km² (1.22%). The most notable increase was in the Watch category, which grew to 3,394,559 km², equivalent to 17.30% of the national territory. Consequently, land under Normal conditions contracted to 79.85% (15,993,194 km²), indicating progressive spatial encroachment of drought-prone areas.

4.5.3. Cropland Vulnerability

Agricultural land analysis shows that in 2000, croplands under Alert and Warning were minimal at 285 km² (0.75%) and 146 km² (0.39%), respectively. About 2911 km² (7.66%) were categorized under Watch, while a substantial 92.52% (35,441 km²) remained under Normal moisture conditions. By 2024, cropland exposure had risen significantly. The Alert cropland had more than doubled to 587 km² (1.54%), while the Warning class cropland expanded to 793 km² (2.08%). Cropland under Watch increased to 3628 km², equivalent to 9.55% of total cropland area. Consequently, Normal cropland declined to 86.83% (33,774 km²), representing a relative drop of over 5.7 percentage points. These findings suggest an increasing agricultural drought hazard with potential implications for food security.

4.5.4. Rangeland Susceptibility

Kenya’s rangelands, vital for pastoral livelihoods, showed a consistent but notable vulnerability. In 2000, only 10.13 km² of rangeland were under Alert conditions, representing 0.02% of the total. The Warning and Watch categories affected 314.88 km² (0.69%) and 4979 km² (10.98%), respectively. The vast majority of 90.28% (40,945 km²) remained under Normal conditions. By 2024, rangelands under Alert expanded dramatically to 586.15 km² (1.29%), while Warning areas contracted slightly to 428.65 km² (0.94%). However, Watch conditions increased to 6388 km², constituting 14.09% of total rangelands. Consequently, the Normal rangeland area decreased to 82.58% (37,789 km²), showing a net decline of nearly 7.7 percentage points, and reinforcing concerns over the growing drought stress on grazing resources.

These findings reveal a clear trend toward increased spatial coverage and population exposure to drought conditions in Kenya between 2000 and 2024 (

Table 4. Summary of the temporal comparison between 2000 and 2024.

Category	2000 (km2/%)	2024 (km2/%)	Relative Change
Population–Alert	284,469 (0.78%)	330,655 (1.46%)	+86%
Population–Watch	3,048,781 (8.30%)	2,342,630 (10.31%)	+24%
Area–Alert	163,463 km2 (0.84%)	320,136 km2 (1.63%)	+96%
Cropland–Alert	285 km2 (0.75%)	586 km2 (1.54%)	+106%
Rangeland–Alert	10.13 km2 (0.02%)	586.15 km2 (1.29%)	+5,687%

), with croplands and rangelands showing amplified vulnerability. The outputs provide essential data-driven evidence for prioritizing drought preparedness strategies, early warning dissemination, and sectoral interventions for agricultural and pastoral systems.

4.6. Trend Analysis: Mann–Kendall and Sen’s Slope Results

The Mann–Kendall (MK) test and Sen’s slope estimator revealed statistically significant (p < 0.05) increasing drought trends across large parts of Kenya between 2000 and 2024 (

Table 5. Summary of Mann–Kendall (MK) Test and Sen’s Slope Trend Analysis of CDI (2000–2024).

Region/County Group	Mann–Kendall Trend	Statistical Significance (p-Value)	Sen’s Slope Range (CDI Units year−1)	Interpretation of Trend	Key Implications
ASAL counties (Garissa, Wajir, Mandera, Kitui, Tana River, Marsabit)	Increasing drought severity	p < 0.05	–0.03 to –0.05	Strong negative trend indicating progressive intensification of drought conditions	Implies a full CDI class shift (e.g., Normal → Watch → Warning) within ~20 years
Eastern Kenya (broader ASAL belt)	Increasing drought severity	p < 0.05	Predominantly negative	Persistent long-term drying	Confirms drought intensification is systematic rather than episodic
Central Kenya	Neutral to weakly positive	p ≥ 0.05	Near zero to slightly positive	Stable or marginally improving conditions	Limited evidence of drought intensification
Western Kenya (Kericho, Kisii, Vihiga, Nyamira)	Neutral to weakly positive	p ≥ 0.05	Near zero to slightly positive	Hydrological stability	Reflects relatively resilient hydroclimatic conditions
National pattern	Spatially heterogeneous	Mixed	ASAL-dominated negative slopes	Strong regional contrast	Trends consistent with NDMA historical reports and climate diagnostics

). Negative Sen’s slope values dominated the ASALs, particularly in Garissa, Wajir, Mandera, Kitui, Tana River, and Marsabit, indicating a progressive shift toward more severe drought classifications over the 25-year period.

Central and western counties, including Kericho, Kisii, Vihiga, and Nyamira, exhibited neutral or weakly positive slopes, suggesting stable or marginally improving hydrological conditions. The ASAL belt showed the strongest negative CDI slopes, with Sen’s slope magnitudes reaching –0.03 to –0.05 CDI units per year, equivalent to a full-class shift (e.g., Normal → Watch → Warning) within two decades. These findings align with NDMA historical drought reports and corroborate regional climate diagnostics linking long-term drying trends to enhanced warming, increased evaporative demand, and recurrent rainfall deficits associated with IOD and ENSO variability [16,27]. The MK/Sen analysis provides robust statistical confirmation that drought intensification is not episodic but part of a persistent long-term trend in eastern Kenya.

5. Discussion

The findings of this study provide robust evidence of intensifying drought conditions in Kenya, underscoring significant spatial expansion, temporal persistence, and sectoral vulnerability associated with climate-induced water scarcity. The analysis, anchored on the CDI, harmonized to four standard categories (Normal, Watch, Warning, Alert), confirms not only the intensification of drought events in Kenya over the past two decades but also reveals a structural transformation in hydroclimatic patterns that has substantial implications for sustainable development and disaster risk management in the country.

Beyond the CDI synthesis, the independent drought indicators themselves provide valuable insights into the mechanisms of drought onset and impact. The SPEI effectively captured rainfall variability, with the strong negative anomalies in 2000 confirming the occurrence of a pronounced meteorological drought, consistent with earlier regional assessments of rainfall failure linked to ENSO and Indian Ocean Dipole variability [27]. However, the SPEI alone was insufficient to explain the severity of ecosystem and livelihood impacts, underscoring the need for complementary measures. The SMA revealed how precipitation deficits translated into hydrological stress, particularly in agricultural and semi-arid regions, where low soil water availability exacerbated vulnerability in rain-fed systems. This finding echoes work by [28], who emphasized soil moisture as a critical mediator between meteorological forcing and agricultural drought. The VCA further highlighted the ecological consequences of combined meteorological and hydrological deficits, with strong vegetation stress in 2000 aligning spatially with areas of low SPEI and SMA. Similar vegetation responses to coupled water deficits have been reported in the Horn of Africa [29], reaffirming the ecological sensitivity of rangelands and croplands to prolonged drought. Taken together, the three indicators illustrate complementary dimensions of drought processes: meteorological triggers, soil water deficits, and ecological responses, while also revealing the limitations of single-indicator approaches when used in isolation.

The observed transition from predominantly Normal and Watch conditions in the early 2000s to widespread Alert and Warning classifications in 2020 and 2024 aligns with global trends of increasing drought frequency and severity under anthropogenic climate change. This progression is particularly pronounced in the ASALs of Kenya, where high CDI scores were consistently recorded. These findings corroborate earlier assessments by [2,16], which documented recurring rainfall failure in Kenya’s eastern and northeastern counties due to disruptions in the Indian Ocean Dipole (IOD) and El Niño–Southern Oscillation (ENSO) patterns.

The CDI-based drought maps (Figure 6) reveal a critical increase in the spatial extent of severe drought conditions, particularly in southeastern Kenya, with Tana River, Kitui, and Garissa counties experiencing persistent Alert conditions. This aligns with the work of [17], who highlighted that these counties face compounding risks due to fragile ecosystems, land degradation, and poor access to water resources. The increased exposure of these regions to extreme drought conditions calls into question the adequacy of existing water harvesting and storage infrastructure, which remains underdeveloped despite repeated drought episodes.

The spatial-temporal degradation maps (Figure 7) add another layer of insight by highlighting temporal shifts in CDI classes, where degradation represents a transition into more severe categories (e.g., Normal → Watch, Watch → Warning, or Warning → Alert). Counties such as Makueni and Kajiado exhibit consistent class degradations across the decades, suggesting chronic stress and limited natural or policy-induced recovery. On the contrary, counties in the western highlands and around Lake Victoria show intermittent improvements, indicating some buffering from localized hydrological and microclimatic advantages [13].

The drought persistence heatmap (Figure 8) provides an integrated perspective on the cumulative drought experience across Kenya from 2000 to 2024. Persistent hotspots in Marsabit, Turkana, and Wajir emphasize the chronic nature of drought in the northern ASAL belt. These counties are socioeconomically reliant on pastoralism, which is acutely sensitive to rangeland degradation, water scarcity, and animal mortality during droughts [12]. The escalation in drought persistence in these zones reaffirms their classification as humanitarian risk epicenters, echoing alerts issued by NDMA and Famine Early Warning Systems Network (FEWS NET) during past drought emergencies. This spatial patterning underscores the need for region-specific drought mitigation strategies. High-persistence zones require targeted interventions such as drought-resilient agriculture, enhanced water harvesting infrastructure, and early warning systems. The heatmap serves as a critical tool for informing national drought policy, guiding resource allocation, and prioritizing resilience-building efforts in Kenya’s most vulnerable regions.

A critical dimension revealed through this study is the increasing population’s exposure to drought risk. Between 2000 and 2024, the population under Alert conditions nearly doubled, while those under Watch conditions increased by over 24%. This increase mirrors national demographic growth and urban-rural shifts, but more importantly, indicates a failure to diversify livelihoods or relocate vulnerable populations to less drought-prone regions.

Equally concerning is the rising vulnerability of Kenya’s croplands and rangelands. The analysis shows that cropland under Alert conditions more than doubled between 2000 and 2024, from 279 km² to 575 km². The expansion of Watch and Warning croplands suggests deteriorating rainfall reliability and soil moisture retention, critical variables for rainfed agriculture. This aligns with SPEI-3 patterns showing negative deviations during key agricultural months (JJAS), consistent with similar assessments in the Horn of Africa [11]. Additionally, rangelands—which support Kenya’s vast pastoral economy—saw a 56-fold increase in area under Alert drought classification. This dramatic escalation, also noted by [6] points to ecosystem degradation, bush encroachment, and declining forage quality, all of which threaten the viability of pastoralism as a sustainable livelihood.

The multi-indicator approach used in this study strengthens the scientific robustness of the results. By integrating SPEI, SMA, and FAPAR anomalies, the CDI reflects both the meteorological trigger (rainfall anomalies) and the ecological response (vegetation and soil moisture deficits). The SPEI-3 index is highly sensitive to agricultural drought, while FAPAR anomalies effectively capture delayed responses in plant productivity [20]. This methodological triangulation offers a more comprehensive picture of drought development and recovery than any single index alone. The harmonization into four operational CDI categories, validated against NDMA drought bulletins, further strengthens its relevance for Kenya’s early warning system. Moreover, the spatial harmonization of drought maps using GEE and QGIS enabled high-resolution analysis at the county level, which is crucial for decentralizing drought response and integrating early warning into local development planning. Counties can now leverage the drought severity and persistence outputs to calibrate their County Integrated Development Plans (CIDPs), Disaster Risk Reduction (DRR) strategies, and resource allocation frameworks.

Comparative analysis with regional and global studies further situates these findings within a broader drought-monitoring discourse. The four-class CDI aligns with international drought early warning practices [23]. For example, similar CDI-based approaches in Ethiopia and the Sahel have been shown to improve drought early warning compared to rainfall-only indices [30]. Our findings of increased drought persistence in northern Kenya parallel trends observed across the Horn of Africa, where prolonged dry spells have intensified pastoral vulnerability [31]. At the same time, the county-level granularity of this study provides a finer-scale contribution than many regional analyses, making the outputs more actionable for localized planning. Importantly, the CDI drought maps developed here can be directly incorporated into NDMA’s early warning bulletins, county-level climate adaptation plans, and NGO drought response programming, thereby bridging the gap between scientific innovation and operational drought management in Kenya.

6. Conclusions

This study provides a comprehensive spatiotemporal assessment of drought occurrence and severity in Kenya by integrating multiple remote sensing indices, SPEI, VCI, SMA, and FAPAR, into the Combined Drought Indicator (CDI). By focusing on the JJAS seasons of 2000, 2010, 2020, and 2024, the analysis captures both historical and projected drought dynamics at the subnational scale, offering new insights into Kenya’s evolving hydroclimatic risks. The CDI improves upon single-indicator approaches by simultaneously accounting for meteorological deficits, soil moisture anomalies, and vegetation stress, thereby reducing uncertainty and enhancing policy relevance.

The results reveal a clear escalation in the frequency, intensity, and persistence of droughts over the past two decades, particularly in arid and semi-arid regions such as Garissa, Kitui, Marsabit, and Tana River. Drought persistence analysis identifies northeastern and southeastern counties as chronic hotspots, while the increasing exposure of croplands, rangelands, and human populations highlights growing threats to food security and livelihoods. These findings align with broader regional trends in the Horn of Africa and emphasize the inadequacy of reactive, short-term drought responses. By disaggregating drought occurrence across seasons and administrative units, the CDI outputs provide actionable evidence for NDMA, county governments, and NGOs to strengthen early warning systems, climate adaptation plans, and resource allocation frameworks.

Looking forward, future work should integrate socioeconomic exposure metrics, household vulnerability data, and climate projections to better capture the human and economic dimensions of drought risk. Linking CDI outputs with poverty, livelihood, and migration datasets would enable more holistic drought risk assessments, while embedding downscaled climate model projections could help anticipate long-term shifts in drought regimes under climate change. Such extensions would further enhance the utility of CDI-based monitoring as both a scientific tool and a decision-support framework, supporting Kenya’s transition toward proactive, resilience-centered drought management.