ORIGINAL RESEARCH article

Front. Environ. Sci., 25 September 2026

Sec. Environmental Informatics and Remote Sensing

Volume 14 - 2026 | https://doi.org/10.3389/fenvs.2026.1889444

Climatic factors influencing vegetation net primary productivity in semi-arid savanna reserves of Limpopo Province, South Africa

  • Department of Environmental Sciences, University of South Africa, Florida Science Campus, Johannesburg, South Africa

Abstract

Savanna ecosystems are important components of the terrestrial carbon cycle, but their vegetation productivity is highly sensitive to climatic variability, especially in semi-arid environments. This study assessed the spatiotemporal dynamics and climatic drivers of vegetation net primary productivity (NPP) across semi-arid savanna reserves in Limpopo Province, South Africa, from 2001 to 2025. Landsat-derived Enhanced Vegetation Index (EVI) was analyzed to characterize long-term vegetation greenness, while MODIS products were used to assess annual NPP and seasonal productivity. Climatic variables, including precipitation, soil moisture, potential evapotranspiration, solar radiation, and temperature, were obtained from CHIRPS, South African Weather Service observations, and ERA5-Land datasets. Geospatial processing was performed using Google Earth Engine and ArcGIS Pro, while Theil–Sen, Mann–Kendall, coefficient of variation, correlation analysis, and Geo-detector were implemented in Google Colab. Results showed a positive annual mean NPP trend of 4.94 g C m-2 yr-1, despite strong interannual variability. NPP trends were mostly directionally positive but largely statistically insignificant, while 86.72% of the landscape showed high or relatively high stability. Precipitation showed strong interannual variability with a weak upward trend, while temperature increased gradually over the study period. Correlation analysis based on ecological-season climate summaries showed that NPP was positively associated with precipitation and soil moisture and negatively associated with solar radiation, PET, and temperature, highlighting the dominant role of growing-season moisture availability. Temporal-window analysis further indicated stronger recent climate–NPP sensitivity, especially to soil moisture, precipitation, PET, and solar radiation. Geo-detector results revealed strong interactions among climatic and topographic variables, notably slope–precipitation, slope–PET, and slope–temperature. These findings suggest that vegetation productivity in protected semi-arid savanna reserves is influenced by both moisture availability and spatial environmental gradients, offering valuable insights for climate-adaptation planning and savanna ecosystem management.

1 Introduction

Climate change is intensifying ecological stress and reshaping terrestrial ecosystem processes (Calvin et al., 2023). Rising temperatures, changing precipitation regimes, and increasing climate variability are altering vegetation functioning, carbon exchange, and ecosystem resilience across terrestrial ecosystems, especially in climate-sensitive semi-arid environments (Smith and Boers, 2023). Understanding how vegetation productivity responds to climatic variability is, therefore, crucial for assessing ecosystem sensitivity and aiding climate adaptation strategies.

Savanna ecosystems are vital components of the terrestrial carbon cycle, supporting biodiversity and carbon storage across extensive tropical and subtropical regions. Complex interactions among climatic factors, soils, herbivory, and local environmental conditions shape their vegetation structure and productivity (Li et al., 2024). In semi-arid savannas, vegetation growth is strongly constrained by water availability, making productivity highly sensitive to climatic variability (Smith and Boers, 2023). Changes in rainfall, temperature, and solar radiation can influence photosynthesis, evapotranspiration, biomass accumulation, and ecosystem stability (Huang et al., 2020; Wang and Zhang, 2024). Prolonged dry conditions may further suppress vegetation productivity and reduce ecosystem resilience, particularly in water-limited landscapes (Yang et al., 2022). Understanding the spatiotemporal dynamics of vegetation productivity and its climatic drivers is crucial for evaluating how ecosystems respond to a changing climate.

Vegetation regulates the exchange of energy, water, and carbon between the land surface and the atmosphere, playing a key role in terrestrial ecosystem functioning (Piao et al., 2020). Net Primary Productivity (NPP) is widely used to evaluate vegetation productivity, ecosystem condition, and terrestrial carbon sequestration because it reflects the carbon fixed by plants after autotrophic respiration (Gang et al., 2013; Yang et al., 2020). Field-based measurements offer detailed data on vegetation productivity and biomass, but they are time-consuming, costly, and spatially limited, making them difficult to apply consistently across large heterogeneous landscapes (Coops, 2015; Wallis et al., 2019). This limitation is especially significant in heterogeneous savanna reserves, where vegetation structure and productivity vary greatly across areas. Geospatial applications offer an effective means of monitoring long-term changes in NPP.

Advances in satellite remote sensing and geospatial technologies have improved the monitoring of climate and vegetation dynamics across broad spatial and temporal scales.

(van Schalkwyk et al., 2022). MODIS NPP products are particularly useful for long-term ecosystem monitoring because they provide annual, spatially continuous estimates of vegetation productivity. The MOD17A3HGF product estimates annual GPP and NPP using a light-use-efficiency framework and has been widely applied in terrestrial productivity studies (Running et al., 2004). Although MODIS NPP captures broad productivity patterns, uncertainties may occur in low-productivity or heterogeneous landscapes (Neumann et al., 2015; Turner et al., 2004). Geo-detector analysis offers a valuable method for identifying the relative influence of environmental factors on vegetation productivity (Wang et al., 2010). Unlike traditional correlation techniques, Geo-detector can evaluate spatial stratified heterogeneity and quantify how individual and combined factors account for variations in NPP (Kumar et al., 2025). This makes it well-suited for analyzing the joint effects of climate and topography in diverse semi-arid savanna landscapes.

Limpopo Province contains extensive areas of the Savanna Biome, which is the largest biome in South Africa (Mucina and Rutherford, 2006). This study focuses on three savanna reserves in the province: Atherstone, Makuya, and Messina Nature Reserves. These reserves represent distinct semi-arid savanna environments (Mucina and Rutherford, 2006). They are characterized by strong rainfall seasonality, significant interannual rainfall variability, and close links between climate and vegetation productivity (Maluleke et al., 2024). Previous studies have demonstrated that climate variability significantly affects vegetation condition and rangeland function in Limpopo, emphasizing the province’s vulnerability to climate change (Scheiter et al., 2018). The near-natural environments within protected areas make these reserves valuable for studying how climate influences vegetation productivity.

Despite increasing interest in vegetation productivity dynamics, the climatic controls on NPP in protected semi-arid savanna ecosystems of southern Africa remain insufficiently documented. Many existing studies have examined vegetation productivity on large regional scales, where local ecological differences can be obscured, or in landscapes where land-use effects can complicate the interpretation of climate–vegetation relationships (Ugbaje et al., 2017; Von Maltitz et al., 2024). Protected savanna reserves in Limpopo Province offer valuable near-natural environments for studying long-term vegetation–climate relationships. This study examines NPP dynamics across Atherstone, Makuya, and Messina Nature Reserves from 2001 to 2025, using correlation and Geo-detector analysis to evaluate the relationships and the explanatory power of precipitation, temperature, solar radiation, PET, and soil moisture. The results are expected to enhance understanding of climate-related patterns of vegetation productivity in semi-arid savanna landscapes.

2 Materials and methods

2.1 Study area

The study focused on three semi-arid savanna reserves in Limpopo Province, South Africa: Atherstone Nature Reserve, Messina Nature Reserve, and Makuya Nature Reserve (Figure 1). Atherstone Nature Reserve is located at approximately 24.493°S, 26.760°E and covers about 23,306 hectares, while Messina Nature Reserve is at 22.408°S, 30.070°E and spans roughly 4,975 ha. Makuya Nature Reserve is located at 22.546°S, 30.980°E and covers around 13,203 ha. These reserves are part of the Savanna Biome and represent distinct savanna bioregions: Atherstone in the Central Bushveld, Messina in the Mopane region, and Makuya in the Lowveld bioregion (Mucina and Rutherford, 2006).

FIGURE 1

Mean annual rainfall ranges from 400 to 500 mm and is mostly concentrated during the austral summer (October–March), while winters are mild and dry. The region’s rainfall is driven by the seasonal movement of the Intertropical Convergence Zone and is strongly modulated by the El Niño–Southern Oscillation, with El Niño events often linked to drier conditions (Kruger and Shongwe, 2004; van Schalkwyk et al., 2022). Temperatures are generally high, frequently exceeding 30 °C in summer. The topography consists of gently undulating plains with scattered low hills, and soils range from sandy to clay-rich types, thereby influencing vegetation distribution (Mucina and Rutherford, 2006). Figure 2 illustrates the spatial distribution of topographic and climatic variables.

FIGURE 2

Topographically, the reserves feature gently undulating plains with scattered low hills, and soils vary from sandy to clay-rich substrates. These environmental gradients affect vegetation, water, and productivity. Vegetation includes bushveld and savanna types, with woody shrubs, scattered trees, and seasonal grasses typical of semi-arid savannas (Mucina and Rutherford, 2006). The variation in bioregions, rainfall seasonality, semi-arid climate, and protected status make these reserves ideal for studying climatic influences on Net Primary Productivity (NPP) in protected savanna ecosystems with reduced direct land-use disturbance.

2.2 Data sources and processing

This study employed a multisource geospatial workflow to evaluate the spatiotemporal dynamics of vegetation productivity and its climatic drivers across Atherstone, Makuya, and Messina Nature Reserves (Table 1). Annual Net Primary Productivity (NPP) was derived from the MODIS MOD17A3HGF product for 2001–2025 and used for analyzing long-term trends, stability, and spatial variation of NPP. For intra-annual and ecological-season analyses, MODIS MOD17A2HGF 8-day PsnNet data were aggregated to monthly and seasonal productivity values (Running and Zhao, 2021).

TABLE 1

DatasetProductSpatial resolutionUnits
Annual NPPMODIS MOD17A3HGF500 mg C m-2 yr-1
Seasonal productivityMODIS MOD17A2HGF500 mg C m-2 month-1
EVILandsat surface reflectance30 mUnitless
PrecipitationCHIRPS v2.00.05° (∼5 km)mm
TemperatureERA5-land∼9 km°C
Soil moistureERA5-land∼9 kmm3 m-3
PETERA5-land∼9 kmmm
Solar radiationERA5-land∼9 kmW m-2
DEMSRTM30 mm

Datasets used in the study.

The Enhanced Vegetation Index (EVI) was derived from Landsat surface reflectance imagery spanning 2001–2025 and served as a supplementary indicator of vegetation greenness and canopy condition. Landsat images were filtered by date and study area, cloud-contaminated pixels were masked, and annual EVI composites were produced for each reserve. EVI was used to support interpretation of long-term vegetation conditions, while MODIS NPP was used as the primary measure of productivity.

Climatic variables included precipitation, temperature, solar radiation, PET, and soil moisture. Precipitation was obtained from CHIRPS v2.0, while temperature, solar radiation, PET, and soil moisture were derived from ERA5-Land. Elevation was obtained from the Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM), and slope was derived from the DEM. All raster datasets were clipped to the reserve boundaries and analyzed spatially to support temporal, spatial, correlation, and Geo-detector analyses.

Spatial harmonization was implemented during extraction and tabular integration, rather than by downscaling all datasets to a new raster resolution. MODIS productivity datasets were extracted at their native 500 m scale and used as the reference layers for productivity analysis. Climatic and topographic variables were extracted using common reserve boundaries and spatial sampling units. The resulting datasets were integrated in Google Colab for correlation and Geo-detector analyses. Coarser climate datasets were used as spatially aligned explanatory inputs and were not interpreted as fine-scale climate surfaces; therefore, spatial results were interpreted at the MODIS NPP scale.

Google Earth Engine (GEE) was used for data access, preprocessing, monthly and annual aggregation, zonal statistics, spatial extraction, and data export (Gorelick et al., 2017). Raster and tabular outputs were exported for further processing and analysis. ArcGIS Pro was used for spatial clipping, raster classification, map production, and class-area calculations. Selected-year NPP maps, Sen’s slope maps, Mann–Kendall trend significance maps, and coefficient of variation maps were produced in ArcGIS Pro. Climatic variables were re-aggregated into ecological seasons, defined from July 1 of the preceding year to June 30 of the target year, to account for the austral summer rainfall pattern. Precipitation and PET were summed over each ecological season, while temperature, soil moisture, and solar radiation were averaged. Solar radiation was standardized and retained as mean flux density in W m-2 for all analyses to ensure consistency. The methodological workflow is summarized in Figure 3.

FIGURE 3

2.3 Statistical analysis

2.3.1 Theil–Sen and Mann–Kendall trend analysis

Temporal trends in NPP were analyzed using the Theil–Sen slope estimator and the Mann–Kendall test for trend significance. These non-parametric methods are suitable for long-term ecological and climate time-series analysis because they do not assume normally distributed data and are less affected by outliers (Jiang et al., 2020; Wei et al., 2022).

The Theil–Sen slope was calculated as the median of all pairwise slopes (Equation 1):where is the Theil–Sen slope, and are NPP values in years and , and is the number of years in the study period. A positive slope indicates an increasing trend in NPP, while a negative slope indicates a decreasing trend.

The Mann–Kendall statistic was calculated using Equations 2, 3:where:

The standardized test statistic Z was calculated using Equation 4:

The significance of the trend was evaluated using a two-sided test. When , the trend was considered statistically significant at the selected significance level; otherwise, the trend was considered insignificant.

2.3.2 Stability analysis

The coefficient of variation (CV) was used to evaluate the temporal stability of NPP from 2001 to 2025. CV measures the relative variability of a time series and is commonly used to assess the stability or volatility of vegetation productivity (Wei et al., 2022). The coefficient of variation was calculated using Equation 5:

NPP stability was classified into five classes: high stability (CV < 0.05), relatively high stability (0.05 ≤ CV < 0.10), moderate volatility (0.10 ≤ CV < 0.15), relatively high volatility (0.15 ≤ CV < 0.20), and high volatility (CV ≥ 0.20). Lower CV values indicate greater temporal stability, while higher CV values indicate stronger interannual variability.

2.3.3 Correlation and geo-detector analysis

Pearson correlation analysis was used to assess relationships between seasonal productivity and corresponding climatic variables. The ecological seasons were aligned with the austral summer growing season. Correlation significance was evaluated at p < 0.05. The Pearson correlation coefficient was determined using Equation 6:where r is the Pearson correlation coefficient, xᵢ is the climatic variable value in year i, yᵢ is the NPP value in year i, and x̄ and ȳ are their respective mean values. Positive r values indicate a direct relationship, while negative values indicate an inverse relationship. To assess changes in climate sensitivity, climate–NPP correlations were compared between early (2002–2013) and recent (2014–2025) ecological seasons. Changes in absolute Pearson correlation strength were used to indicate whether NPP became more or less sensitive to each climatic variable.

Geo-detector analysis was performed in Google Colab to evaluate the explanatory power and interaction effects of climatic and topographic variables on ecological-season productivity. The factor detector was used because it does not assume a linear relationship between the dependent and independent variables, making it suitable for identifying spatially stratified environmental effects (Wang et al., 2010; Wang and Xu, 2017). Continuous variables were classified into discrete strata before Geo-detector analysis. Elevation, slope, precipitation, temperature, soil moisture, solar radiation, and PET were all discretized into five strata using quantile classification. The five-strata scheme was selected to provide sufficient class separation while maintaining a balanced number of observations within each stratum. Quantile classification was used to reduce the risk of sparse classes and allows comparison among variables with different value ranges. The Geo-detector factor detector is expressed in Equation 7:where q represents the explanatory power of a driving factor and ranges from 0 to 1, a higher q value indicates a stronger contribution to NPP spatial variation. Nh and N are the number of units in stratum h and the entire study area, respectively, while σh2 and σ2 are the variances of NPP within stratum h and across the full study area.

Climatic variables with higher q-values were interpreted as having a stronger influence on the spatial variability of NPP. Pearson correlation results were used to describe the direction of climate–NPP relationships, while Geo-detector results were used to assess the relative importance and interaction effects of climatic and topographic drivers.

3 Results and discussion

3.1 Spatial and temporal trends of EVI from 2001 to 2025

The annual mean EVI showed clear spatial and temporal variation across Atherstone, Messina, and Makuya Nature Reserves from 2001 to 2025 (Figure 4). The time-series pattern reveals strong year-to-year fluctuations, with periods of decline followed by recovery. Despite this variability, the overall trend was upward, indicating a gradual improvement in vegetation greenness across the reserves during the study period. Mean EVI increased from 0.161 to 0.274 in Atherstone, from 0.213 to 0.269 in Makuya, and from 0.118 to 0.172 in Messina.

FIGURE 4

The spatial distribution of EVI also varied significantly among the reserves and across different years. By 2025, higher EVI values were more evident in parts of Atherstone, Makuya, and Messina, indicating localized increase in vegetation greenness. However, low and fragmented EVI values continued in some regions, showing that vegetation greenness remained spatially uneven. This pattern is consistent with the heterogeneous nature of semi-arid savanna environments, where vegetation conditions are influenced by rainfall availability, soil properties, topography, and vegetation structure (Mucina and Rutherford, 2006).

The interannual EVI pattern showed noticeable declines in some years and significant peaks in 2017, 2021, and 2025. These fluctuations suggest that vegetation greenness responded to changing climatic and local environmental conditions rather than increasing smoothly and linearly. Overall, the results indicate increased vegetation greenness across the reserves, but the response was spatially uneven and temporally variable. This greening may also reflect bush encroachment, alien plant invasion, or shifts in vegetation composition rather than uniform improvement in vegetation conditions.

3.2 Spatial and temporal variability of NPP

3.2.1 Interannual variation in NPP

The annual mean NPP across the semi-arid savanna reserves showed a fluctuating but positive directional trend from 2001 to 2025, with a growth rate of 4.94 g C m-2 yr-1 (Figure 5A). However, the goodness-of-fit of the linear trend was low (R2 = 0.120), indicating that time explained only a small proportion of the annual NPP variability. This shows that year-to-year fluctuations were more prominent than the long-term trend. Therefore, the increase reflects a general upward tendency rather than a steady or uniform rise in productivity. Lower NPP values were recorded in 2002, 2015, and 2024, while higher values were recorded in 2014, 2017, 2021, and 2025. The sharp decline from 2014 to 2015 and the strong increase from 2024 to 2025 further demonstrate the sensitivity of productivity to short-term climatic fluctuations. Such residual variability may reflect interannual rainfall variability and broader climate oscillations, including ENSO-related wet and dry phases, which are known to influence rainfall patterns in the region. Overall, the results indicate that vegetation productivity increased directionally over the study period, but annual NPP variability appears to have been more strongly influenced by interannual climatic fluctuations than by the fitted linear time trend.

FIGURE 5

3.2.2 Spatiotemporal pattern and characteristics of NPP

The spatial distribution of annual mean NPP showed clear heterogeneity across Atherstone, Messina, and Makuya Nature Reserves in 2001, 2005, 2015, and 2025 (Figures 5b–d). Productivity was unevenly distributed within and among the three reserves, indicating that vegetation growth varied spatially over time. This pattern is expected in semi-arid savannas, where vegetation productivity is influenced by rainfall availability, soil conditions, topography, and local ecological settings (Mucina and Rutherford, 2006). Between 2001 and 2025, higher NPP values became more visible in parts of the study area, suggesting localized increases in vegetation productivity. However, lower-productivity areas persisted, showing that productivity gains were not spatially uniform. These results indicate that vegetation productivity changed over time but remained strongly heterogeneous across the study reserves.

3.2.3 Trend analysis for NPP

The fitted Sen’s slope of NPP ranged from −39.17 to 43.54 g C m-2 yr-1, indicating both increasing and decreasing productivity trends across the savanna reserves from 2001 to 2025 (Figure 6A). Directionally positive slope values covered most of the study area, indicating a widespread but mostly weak increase in vegetation productivity. Spatially, stronger increases were more evident in parts of Atherstone and Makuya, while weaker increases were more common in Messina Nature Reserve.

FIGURE 6

Trend significance analysis showed that NPP changes were dominated by directionally positive, but mostly insignificant trends (Figure 6B; Table 2). The largest proportion of the study area fell within the insignificant increase category, accounting for 73.44% of the total area. In contrast, weakly significant, significant, and extremely significant increases accounted for 14.45%, 9.08%, and 1.43%, respectively. Thus, while the direction of change was predominantly positive, statistically stronger evidence of increasing NPP was confined to a smaller proportion of the landscape. Decreasing trends were limited in extent, with all reduction classes together accounting for less than 2% of the study area. These findings indicate that NPP dynamics across Atherstone, Makuya, and Messina Nature Reserves were characterized mainly by a widespread positive tendency, but not by a uniformly significant greening trend. Therefore, the observed increase should be interpreted as gradual, spatially uneven, and largely statistically weak rather than as a strong or consistent improvement across the entire protected savanna landscape.

TABLE 2

SignificanceTrendZ valueArea%
Extremely significant reduction<0Z < −2.5862.520.06
Significant reduction<0−2.58 ≤ Z < −1.9648.410.04
Weakly significant reduction<0−1.96 ≤ Z < −1.6535.110.03
Insignificant reduction<0−1.65 ≤ Z < 01,6191.47
Unchanged0Z = 00.920.00
Insignificant increase>00 < Z ≤ 1.6580,71373.44
Weakly significant increase>01.65 < Z ≤ 1.9615,87914.45
Significant increase>01.96 < Z ≤ 2.589,9749.08
Extremely significant increase>0Z > 2.581,5681.43

NPP trend significance statistics.

3.2.4 Monthly variation of vegetation productivity

Monthly productivity showed a clear seasonal pattern across Atherstone, Makuya, and Messina Nature Reserves (Figure 7A). Productivity was highest from January to March, corresponding with the austral summer growing season and the main rainfall period shown in Figure 7B. Makuya recorded the highest NPP during this period, increasing from 95 g C m-2 month-1 in January to 98 g C m-2 month-1 in February, before slightly declining to 95 g C m-2 month-1 in March. Atherstone also showed high early-season productivity, rising from 62 to 65 g C m-2 month-1, while Messina ranged from 52 to 55 g C m-2 month-1 over the same period.

FIGURE 7

The highest mean monthly NPP Supplementary Table S1 occurred in February (72.038 g C m-2 month-1), followed by March (70.551 g C m-2 month-1) and January (69.466 g C m-2 month-1). In contrast, October recorded the lowest mean NPP (17.840 g C m-2 month-1; 3.31%), reflecting the late dry-season minimum before rainfall-driven recovery began toward November and December. This seasonal pattern shows that vegetation productivity was highest during the austral summer growing season, declined through the dry season, and began to recover toward November and December.

3.3 coefficient of variation (CV) index for NPP

The coefficient of variation (CV) showed clear spatial differences in the temporal stability of NPP across the protected savanna reserves from 2001 to 2025 (Table 3). Most of the study area was classified as stable, with 53.51% in the high-stability class and 33.21% in the relatively high-stability class. Together, these two classes accounted for 86.72% of the total area, indicating that NPP remained relatively consistent across much of the landscape during the study period. Areas of moderate volatility accounted for 11.99%, while relatively high volatility covered only 1.30% of the study area. High volatility was almost absent, representing only 0.002%. This distribution indicates that strong interannual fluctuations in NPP were confined to very small areas, while most reserves maintained low variability in vegetation productivity. These results suggest that, despite the year-to-year fluctuations observed in NPP value, vegetation productivity across most of the protected savanna landscape remained relatively stable over the long term. The dominance of high- and relatively high-stability classes indicates a generally consistent productivity pattern, although localized areas of moderate or higher variability remain.

TABLE 3

CV rangeStability classAreaPercentage (%)
CV < 0.05High stability59,15453.51
0.05 ≤ CV < 0.10Relatively high stability36,71233.21
0.10 ≤ CV < 0.15Moderate volatility13,25211.99
0.15 ≤ CV < 0.20Relatively high volatility1,4321.30
CV ≥ 0.20High volatility0.230.002

Stability analysis of NPP based on the CV from 2001 to 2025.

3.4 Climatic factors and their relationship with NPP

3.4.1 Temporal patterns of climatic factors

Mean precipitation and temperature showed contrasting temporal patterns across the study area from 2001 to 2025 (Figure 8). Precipitation varied strongly among years, with wetter and drier years occurring in clusters rather than in regular alternation. The fitted trend showed a weak increase of 4.25 mm yr-1, indicating a slight upward tendency over the study period. Higher precipitation values were observed in 2021 and 2025, while lower values were recorded around 2002, 2006, and 2015. Mean temperature also showed a gradual increasing trend, with a fitted slope of 0.021 °C yr-1. Although both variables fluctuated over time, the results indicate marked interannual rainfall variability together with gradual warming. These climatic patterns provide useful context for explaining the annual fluctuations observed in vegetation productivity.

FIGURE 8

Mean monthly precipitation was highest during the summer rainfall season, particularly from November to February, with peak rainfall occurring around January. Rainfall declined sharply from March onward and remained very low during the dry winter months, especially from June to August. This confirms a strongly seasonal rainfall regime, where most moisture input occurs during the summer growing season. Mean monthly temperature followed a contrasting seasonal cycle. Temperatures were highest during the summer months, particularly from October to March, and lowest during the winter months of June and July. The coincidence of high rainfall and warm temperatures during summer indicates that this period represents the main growing season for vegetation productivity. In contrast, the winter period was characterized by low rainfall and cooler temperatures, conditions less favorable to sustained vegetation growth.

3.4.2 Influence and relative contribution of climate variations on NPP

Correlation analysis was used to examine the relationship between vegetation productivity and climatic variables across the combined protected savanna reserves (Table 4). The results showed that NPP was more strongly associated with moisture-related variables than with thermal variables. Precipitation had the strongest positive relationship with NPP (r = 0.8741; p < 0.001), followed by soil moisture (r = 0.6834; p < 0.001). In contrast, NPP was negatively correlated with solar radiation (r = −0.5730; p = 0.0034), PET (r = −0.5417; p = 0.0063), and temperature (r = −0.3222; p = 0.1247). These results suggest that vegetation productivity was primarily linked to growing-season moisture availability, while higher evaporative demand, stronger radiation, and warmer conditions were generally associated with lower NPP.

TABLE 4

RelationshipCorrelation coefficient (r)p-valueDirection
NPP-soil moisture0.6834<0.001Positive
NPP-precipitation0.8741<0.001Positive
NPP-solar radiation−0.57300.0034Negative
NPP-temperature−0.32220.1247Negative, non-significant
NPP-PET−0.54170.0063Negative

Correlation between NPP and climatic variables across the reserves.

Climate–NPP sensitivity was compared between early (2002–2013) and recent (2014–2025) ecological-season periods (Supplementary Table S2). Correlation strength increased in the recent period for all climatic variables, with the largest increases observed for soil moisture, solar radiation, and PET. Precipitation remained strongly positive in both periods, increasing from r = 0.7481 to r = 0.9259. Overall, NPP became more closely linked to growing-season moisture availability and atmospheric water demand during the recent period. These patterns indicate that moisture-related variables, particularly precipitation and soil moisture, dominated NPP variability, while PET and solar radiation intensified atmospheric water-demand stress.

The monthly lag analysis identified January as the strongest climatic response period (Table 5). Same-month January precipitation and soil moisture were positively correlated with NPP (r = 0.795 and r = 0.731, respectively; p < 0.05), whereas January temperature and solar radiation were negatively correlated with NPP (r = −0.680 and r = −0.759, respectively; p < 0.05). Earlier-month effects were weaker and less consistent. November precipitation showed a moderate positive relationship with NPP (r = 0.496; p < 0.05), while most other lagged relationships were not significant. This suggests that NPP responded most strongly to current or near-current growing-season moisture conditions.

TABLE 5

LagMonthPrecipitationTemperatureSoil moistureSolar radiation
0January0.795*−0.680*0.731*−0.759*
1December0.229−0.2000.115−0.378
2November0.496*0.283−0.135−0.096
3October−0.0680.167−0.072−0.037
4September−0.0660.410−0.3800.210
5August−0.503*−0.140−0.3390.095

Monthly lag correlations between NPP and climatic variables.

3.4.3 Spatial correlation patterns

Spatial correlation showed that precipitation had a predominantly positive association with NPP across the study area (Table 6). Significant positive precipitation-NPP correlations covered 78.92% of the area, while an additional 21.01% showed insignificant positive correlations. Negative precipitation-NPP correlations were negligible. This confirms that rainfall was a major spatial control on NPP variation across the reserves. Temperature showed the opposite pattern. Significant negative temperature-NPP correlations covered 89.26% of the area, with a further 10.73% showing insignificant negative correlations. Positive temperature-NPP relationships were almost absent. This indicates that higher temperatures were associated with reduced NPP.

TABLE 6

VariableSignificant positive (%)Insignificant positive (%)Significant negative (%)Insignificant negative (%)
Temperature0.000.0189.2610.73
Precipitation78.9221.010.000.08

Spatial correlation area proportions between NPP and selected climatic variables.

3.4.4 Relative contribution and factor interaction detector

The Geo-detector factor results showed clear differences in the relative explanatory power of the selected variables (Figure 9a). Elevation had the highest mean q-value (0.813), followed by solar radiation (0.669), temperature (0.644), precipitation (0.512), PET (0.492), slope (0.451), and soil moisture (0.366). The factor interaction detector was applied to assess how paired variables improved the explanation of NPP variation relative to individual variables (Figures 9b–g). Across the selected years, the interaction q-statistics varied, indicating that the explanatory strength of paired variables was not uniform over time. Several interactions showed higher q-values than the corresponding single-factor values, confirming that combinations of variables better explained NPP variation than isolated variables alone.

FIGURE 9

In 2001 and 2005, the strongest interactions were mainly associated with elevation, slope, precipitation, temperature, and PET. In 2001, high q-statistics were observed for elevation–slope, slope–precipitation, elevation–precipitation, and slope–PET. In 2005, elevation–slope and slope–precipitation remained strong, while interactions involving slope with temperature and PET also showed high explanatory power. These results indicate that slope-related interactions were already prominent in the early part of the analysis period.

The interaction pattern changed in 2010, when several variable pairs showed lower q-statistics than in 2001 and 2005. However, slope–PET, elevation–precipitation, elevation–slope, slope–temperature, and slope–solar radiation remained among the strongest interactions. In 2015, pairwise explanatory power increased again, with high q-statistics for elevation–precipitation, elevation–slope, slope–temperature, slope–PET, and slope–solar radiation. This shows that interaction strength was not stable through time, but slope-related combinations repeatedly appeared among the dominant interactions.

In 2020, interaction strength declined for several variable pairs, particularly those involving temperature, soil moisture, solar radiation, and PET. Despite this decline, slope–precipitation, elevation–slope, slope–PET, slope–temperature, and slope–solar radiation remained relatively strong. By 2025, interaction strength increased again, with slope–precipitation reaching the highest q-statistic, followed by elevation–slope, slope–PET, slope–temperature, and elevation–precipitation. Overall, the selected variables interacted differently across years, but interactions involving slope were consistently among the strongest pairwise effects.

Although these interactions show that climatic and topographic variables jointly explain NPP variation, the observed vegetation patterns should also be considered within the broader ecological context of the study reserves. Reserve-level evidence indicates that rainfall variability, bush encroachment, overgrazing, fire management, and carrying-capacity constraints may influence vegetation condition in parts of the study area (Kruger, 2014; Kruger and Manager Biodiversity, 2025; Ysterberg Enviro-Veld and Game Management Services, 2016; Ysterberg Enviro-Veld and Game Management Services, 2021). For instance, in Atherstone, elephant impacts on woody vegetation have also been reported (Seloana et al., 2017), while in Makuya, wildlife movement from the adjoining Kruger system may further contribute to herbivore pressure (Limpopo Wildlife Resorts, 2026). These factors may shape woody cover, grass-layer condition, and productivity signals alongside climate.

The increase in vegetation greenness and productivity might also indicate broader ecological processes. Rising atmospheric CO2 can boost photosynthesis and water-use efficiency, especially in warm, water-limited environments, thereby increasing vegetation cover even when rainfall does not increase consistently (Donohue et al., 2013). In savannas, this process may favour woody plants and contribute to woody thickening or bush encroachment, which can increase greenness and carbon uptake without necessarily improving the grass-layer condition (Devine et al., 2017). Therefore, part of the observed increase in EVI and NPP may reflect changes in woody vegetation structure, in addition to climate-driven productivity responses. Although MODIS NPP products are useful for long-term monitoring, uncertainties remain in heterogeneous savanna landscapes because mixed woody–grass cover and coarse spatial resolution can affect productivity estimates (Turner et al., 2006).

4 Conclusion

This study examined the spatiotemporal dynamics of vegetation productivity and its climatic influences across Atherstone, Makuya, and Messina Nature Reserves in Limpopo Province, South Africa, using multisource geospatial data and statistical analyses. Landsat-derived EVI indicated an overall increase in vegetation greenness from 2001 to 2025, with mean EVI rising in all three reserves. However, the increase was uneven across space and showed strong interannual variability, suggesting that vegetation conditions did not improve uniformly throughout the protected savanna landscape.

MODIS-derived NPP showed a positive directional trend from 2001 to 2025, with fitted annual increase of 4.94 g C m-2 yr-1. However, the low goodness-of-fit (R2 = 0.120) indicates that the linear time trend explained only a small proportion of annual NPP variability. This suggests that interannual climatic fluctuations, rather than a steady linear increase, dominated the productivity signal. Spatially, NPP remained uneven across and within the reserves, with some areas showing more visible higher productivity by 2025, while lower productivity zones persisted elsewhere. Sen’s slope and Mann–Kendall analyses showed that directionally positive trends were widespread, but most increases were statistically insignificant. This indicates that the greening signal was spatially uneven and largely weak, with statistically stronger positive trends limited to a small part of the study area rather than a consistent regional change.

The monthly productivity pattern showed clear seasonality, with the highest productivity occurring from January to March and the lowest values around September and October. This seasonal pattern indicates that vegetation productivity and carbon uptake were strongest during the wet summer growing period and declined during the dry season. The coefficient of variation analysis showed that 86.72% of the study area fell into the high-stability and relatively high-stability classes, indicating that most of the protected savanna landscape maintained relatively stable productivity despite year-to-year fluctuations.

Climatic analysis showed that precipitation and temperature followed contrasting temporal patterns. Precipitation exhibited strong interannual variability with a weak upward trend, while temperature showed a gradual increase. The correlation analysis based on ecological-season climate summaries showed that NPP was positively associated with precipitation and soil moisture and negatively associated with solar radiation, PET, and temperature, highlighting the dominant role of growing-season moisture availability in vegetation productivity. The temporal-window comparison further suggests that NPP became more strongly coupled to climatic variability in the recent period, particularly to soil moisture, precipitation, PET, and solar radiation.

The Geo-detector results showed that combinations of variables better explained NPP variation than individual factors alone. While elevation and slope were included as supporting topographic variables, the interaction results showed that slope-related combinations with precipitation, PET, temperature, and solar radiation repeatedly produced strong explanatory effects. These findings demonstrate that NPP dynamics in the selected semi-arid savanna reserves were shaped by both direct climatic relationships and spatially structured interactions among climatic and topographic variables. Overall, the study provides evidence that vegetation productivity in these protected savanna reserves is strongly linked to moisture availability, warming conditions, seasonal climate variability, and spatial environmental gradients.

Statements

Data availability statement

The original contributions presented in the study are included in the article/Supplementary Material; further inquiries can be directed to the corresponding author.

Author contributions

YM: Conceptualization, Data curation, Formal Analysis, Investigation, Methodology, Project administration, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review and editing. SN: Funding acquisition, Methodology, Project administration, Resources, Supervision, Validation, Writing – review and editing.

Funding

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

Acknowledgments

The authors extend their sincere gratitude to the Limpopo Department of Economic Development, Environment and Tourism for granting permission to conduct this research in the study reserves.

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 used in the creation of this manuscript. Generative AI tools were used only for grammar correction and clarity improvement.

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.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fenvs.2026.1889444/full#supplementary-material

References

  • 1

    CalvinK.DasguptaD.KrinnerG.MukherjiA.ThorneP. W.TrisosC.et al (2023). IPCC, 2023: Climate Change 2023: Synthesis Report, Summary for Policymakers. Contribution of Working Groups I, II and III to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. 10.59327/IPCC/AR6-9789291691647.001

  • 2

    CoopsN. C. (2015). Characterizing forest growth and productivity using remotely sensed data. Curr. For. Rep.1 (3), 195–205. 10.1007/s40725-015-0020-x

  • 3

    DevineA. P.McDonaldR. A.QuaifeT.MacleanI. M. D. (2017). Determinants of woody encroachment and cover in African savannas. Oecologia183 (4), 939–951. 10.1007/s00442-017-3807-6

  • 4

    DonohueR. J.RoderickM. L.McVicarT. R.FarquharG. D. (2013). Impact of CO2 fertilization on maximum foliage cover across the globe’s warm, arid environments. Geophys. Res. Lett.40 (12), 3031–3035. 10.1002/grl.50563

  • 5

    GangC.ZhouW.LiJ.ChenY.MuS.RenJ.et al (2013). Assessing the spatiotemporal variation in distribution, extent and NPP of terrestrial ecosystems in response to climate change from 1911 to 2000. PLoS ONE8 (11), e80394. 10.1371/journal.pone.0080394

  • 6

    GorelickN.HancherM.DixonM.IlyushchenkoS.ThauS.MooreR. (2017). Google earth engine: planetary-scale geospatial analysis for everyone. Remote Sensing of Envir.202, 18–27. 10.1016/j.rse.2017.06.031

  • 7

    HuangQ.ZhangF.ZhangQ.OuH.JinY. (2020). Quantitative assessment of the impact of human activities on terrestrial net primary productivity in the yangtze river Delta. Sustain. Switz.12 (4), 1697. 10.3390/su12041697

  • 8

    JiangY. L.GuoJ.PengQ.GuanY.ZhangY.ZhangR. (2020). The effects of climate factors and human activities on net primary productivity in Xinjiang. Int. J. Biometeorol.64 (5), 765–777. 10.1007/s00484-020-01866-4

  • 9

    KrugerM. J. (2014). “Ysterberg enviro-veld and game management services 2007/03696/23 CC ysterberg enviro-veld and game management services 2007/03696/23 CC August 2015 biodiversity: department of finance,” in The Description: Appointment of a Service Provider to Do Annual Veld Condition Assessments (VCA) on State Owned Nature Reserves in Limpopo for the Period.

  • 10

    KrugerJ.Manager BiodiversityS. (2025). Ysterberg enviro-veld and game management services 2007/03696/23 CC description: appointment of a service provider to do annual veld condition assessments (VCA) on state owned nature reserves in Limpopo for the period November 2024 to November 2027 progress report for the grass monitoring project.

  • 11

    KrugerA. C.ShongweS. (2004). Temperature trends in South Africa: 1960-2003. Int. J. Climatol.24 (15), 1929–1945. 10.1002/joc.1096

  • 12

    KumarS.KarkalaJ.AryaJ.ChaturvediR. K. (2025). Synergistic analysis of climate and anthropogenic impacts on NPP dynamics in India’s Western ghats using geodetector. Sci. Rep.15 (1), 38252. 10.1038/s41598-025-22061-8

  • 13

    LiX.WesselsK.ArmstonJ.DuncansonL.UrbazaevM.NaidooL.et al (2024). Evaluation of GEDI footprint level biomass models in southern African savannas using airborne LiDAR and field measurements. Sci. Remote Sens.10, 100161. 10.1016/j.srs.2024.100161

  • 14

    Limpopo Wildlife Resorts (2026). Makuya Nature Reserve. Available online at: https://www.lwr.gov.za/lodges/resorts/makuya?utm (Available, June 29, 2026).

  • 15

    MalulekeP.MoeletsiM. E.TsuboM. (2024). Analysis of climate variability and its implications on rangelands in the Limpopo province. Climate12 (1), 2. 10.3390/cli12010002

  • 16

    MucinaL.RutherfordM. C. (2006). The Vegetation of South Africa, Lesotho and Swaziland. Pretoria: South African National Biodiversity Institute.

  • 17

    NeumannM.ZhaoM.KindermannG.HasenauerH. (2015). Comparing MODIS net primary production estimates with terrestrial national forest inventory data in Austria. Remote Sens.7 (4), 3878–3906. 10.3390/rs70403878

  • 18

    PiaoS.WangX.ParkT.ChenC.LianX.HeY.et al (2020). Characteristics, drivers and feedbacks of global greening. Nat. Rev. Earth Environ.1 (1), 14–27. 10.1038/s43017-019-0001-x

  • 19

    RunningS. W.ZhaoM. (2021). User’s guide daily GPP and annual NPP (MOD17A2H/A3H) and year-end gap-filled (MOD17A2HGF/A3HGF) products NASA Earth observing system MODIS land algorithm (for collection 6.1). Available online at: https://landweb.modaps.eosdis.nasa.gov/QA_WWW/forPage/MODIS_C61_Land_. (Accessed August 8, 2026).

  • 20

    RunningS. W.NemaniR. R.HeinschF. A.ZhaoM.ReevesM.HashimotoH. (2004). A continuous satellite-derived measure of global terrestrial primary production. BioScience54 (6), 547. 10.1641/0006-3568(2004)054[0547:acsmog]2.0.co;2

  • 21

    ScheiterS.GaillardC.MartensC.ErasmusB. F. N.PfeifferM. (2018). How vulnerable are ecosystems in the Limpopo province to climate change?South Afr. J. Bot.116, 86–95. 10.1016/j.sajb.2018.02.394

  • 22

    SeloanaM.Q.S.W.K.M.J. P.JJ. J. (2017). Elephant damage to Sclerocarya birrea on different landscapes. Int. J. Biodivers. Conservation9 (4), 97–106. 10.5897/ijbc2015.0912

  • 23

    SmithT.BoersN. (2023). Global vegetation resilience linked to water availability and variability. Nat. Commun.14 (1), 498. 10.1038/s41467-023-36207-7

  • 24

    TurnerD. P.OllingerS.SmithM. L.KrankinaO.GregoryM. (2004). Scaling net primary production to a MODIS footprint in support of Earth observing system product validation. Int. J. Remote Sens.25 (10), 1961–1979. 10.1080/0143116031000150013

  • 25

    TurnerD. P.RittsW. D.CohenW. B.GowerS. T.RunningS. W.ZhaoM.et al (2006). Evaluation of MODIS NPP and GPP products across multiple biomes. Remote Sens. Environ.102 (3–4), 282–292. 10.1016/j.rse.2006.02.017

  • 26

    UgbajeS. U.OdehI. O. A.BishopT. F. A.LiJ. (2017). Assessing the spatio-temporal variability of vegetation productivity in Africa: quantifying the relative roles of climate variability and human activities. Int. J. Digital Earth10 (9), 879–900. 10.1080/17538947.2016.1265017

  • 27

    van SchalkwykL.BlameyR. C.DysonL. L.ReasonC. J. C. (2022). A climatology of drylines in the interior of subtropical Southern Africa. J. Clim.35 (19), 2809–2828. 10.1175/JCLI-D-21-1005.1

  • 28

    Ysterberg Enviro-Veld and Game Management Services (2021). Executive summary: veld condition assessments on state-owned nature reserves in Limpopo.

  • 29

    Ysterberg Enviro-Veld and Game Management Services (2016). Veld condition assessment on atherstone collaborative nature reserve.

  • 30

    Von MaltitzG. P.MidgleyG. F.VeitchJ.BrümmerC.RötterP.et al (2024). Ecological Studies 248 sustainability of southern African ecosystems under global changeSci. for Manag. and Policy Interventions. Ecological Studies, 248. Cham: Springer. 10.1007/978-3-031-10948-5

  • 31

    WallisC. I. B.HomeierJ.PeñaJ.BrandlR.FarwigN.BendixJ. (2019). Modeling tropical montane forest biomass, productivity and canopy traits with multispectral remote sensing data. Remote Sens. Environ.225, 77–92. 10.1016/j.rse.2019.02.021

  • 32

    WangJ. F.XuC. D. (2017). Geodetector: principle and prospective. Acta Geogr. Sin.72(1), 116–134. 10.11821/dlxb201701010

  • 33

    WangT.ZhangW. (2024). Analysis of the characteristics and mechanism of spatial and temporal changes in net primary productivity (NPP) in northwest Liaoning Province. Appl. Sci. Switz.14 (15), 6762. 10.3390/app14156762

  • 34

    WangJ. F.LiX. H.ChristakosG.LiaoY. L.ZhangT.GuX.et al (2010). Geographical detectors-based health risk assessment and its application in the neural tube defects study of the heshun region, China. Int. J. Geogr. Inf. Sci.24 (1), 107–127. 10.1080/13658810802443457

  • 35

    WeiX.YangJ.LuoP.LinL.LinK.GuanJ. (2022). Assessment of the variation and influencing factors of vegetation NPP and carbon sink capacity under different natural conditions. Ecol. Indic.138, 108834. 10.1016/j.ecolind.2022.108834

  • 36

    YangH.HuD.XuH.ZhongX. (2020). Assessing the spatiotemporal variation of NPP and its response to driving factors in Anhui province, China. Environ. Sci. Pollut. Res.27 (13), 14915–14932. 10.1007/s11356-020-08006-w

  • 37

    YangF.WangJ.ZhangC.LiJ.XieH.ZhuogeZ. (2022). The impact of human activities on net primary productivity in a grassland open-pit mine: the case study of the shengli mining area in Inner Mongolia, China. Land11 (5), 743. 10.3390/land11050743

Summary

Keywords

climatic factors, geo-detector, google earth engine (GEE), limpopo province, net primary productivity (NPP), savanna ecosystem

Citation

Mshelia YS and Nkosi SE (2026) Climatic factors influencing vegetation net primary productivity in semi-arid savanna reserves of Limpopo Province, South Africa. Front. Environ. Sci. 14:1889444. doi: 10.3389/fenvs.2026.1889444

Received

23 May 2026

Revised

18 July 2026

Accepted

20 July 2026

Published

25 September 2026

Volume

14 - 2026

Edited by

Yong Q Tian, Central Michigan University, United States

Reviewed by

Graham Von Maltitz, South African National Biodiversity Institute, South Africa

Suneel Kumar Duvvuri, Government College Autonomous Rajahmundry, India

Updates

Copyright

*Correspondence: Yoksa Salmamza Mshelia,

Disclaimer

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.

Outline

Figures

Cite article

Copy to clipboard


Export citation file


Share article

Article metrics