A systematic search was conducted across major academic databases, including Google Scholar, Web of Science, and Scopus, using the keywords “nitrous oxide emissions”, “cropping systems”, “emission factors”, and “Sub-Saharan Africa”. These platforms were chosen for their extensive collections of peer-reviewed articles, ensuring access to high-quality research. Unlike previous meta-analyses, our approach explicitly included cropping systems, fertilizer use, and soil properties, broadening the scope of the search. We adopted the Intergovernmental Panel of Climate Change (IPCC, 2006a) definition of cropland, which includes arable and tillable land, rice fields, and agroforestry systems not classified as forest. From the initial screening, 1,410 studies were selected (Figure 1), of which 39 met the criteria for inclusion in the final analysis. A further subset of 20 studies was retained for fertilizer-specific comparisons, as these provided explicit fertilizer treatment data. Articles were systematically excluded due to the following reasons: i) Lack of focus on N₂O emissions–many papers mentioned nitrogen cycling, soil health, or greenhouse gases broadly but did not report specific N₂O flux measurements; ii) Absence of in situ measurements–some studies modeled emissions or relied on literature reviews rather than direct field measurements; iii) Study type exclusions–laboratory incubation studies, management-focused studies whose primary aim was to test or evaluate specific management interventions designed to reduce N₂O emissions, rather than to measure baseline or typical emissions under standard agricultural practices, meta-analyses, non-peer-reviewed HTML documents and geographic mismatch.

After applying these filters, 39 studies spanning 13 countries remained (covering the years 2004–2025), all of which either reported in situ annual N₂O emissions or provided sufficient data for unit conversion-based estimation (see Annex 2). These studies covered diverse cropping systems and soil types, with data spanning from 1999 to 2025 (no date restrictions were applied to maximize inclusion). As climate data was not explicitly reported in many studies, mean annual temperature (MAT) was sourced from WorldClim 2.1 (Fick and Hijmans, 2017) and mean annual precipitation (MAP) from CHIRPS (Funk et al., 2015) with an average for the year range of 2000–2020.

The extracted and compiled dataset included:

N₂O fluxes and associated emission values.

For a focused analysis on the effect of N fertilizer application on N₂O emissions, a subset of studies was extracted from the full review database. Studies were included in this subset if they explicitly reported fertilizer-induced N₂O fluxes or emission factors, or if they presented treatment comparisons between fertilized and unfertilized plots. This subset was used to evaluate emission responses to different nitrogen input levels. This selection resulted in 20 studies with comprehensive data (see Annex 3).

Data on annual N₂O emissions were standardized to kg N₂O-N ha^-1 yr^-1 to facilitate cross-study comparisons. Extreme high values were inspected and retained unless they were clearly attributable to measurement or reporting errors. To address skewness, variance-robust statistical approaches were applied. In most cases, means were reported because they capture the full contribution of episodic pulses that often dominate annual N₂O totals in SSA and are consistent with inventory-oriented reporting (IPCC, 2006b), In some cases, medians were used, for example, when sample sizes were very small or when distributions were highly skewed to provide a robust summary less influenced by outliers. Where studies reported annual totals, we used those values directly. For studies in multi-cropping systems that provided season level measurements for all seasons, seasonal fluxes were summed to obtain an annual total. For studies that reported only a single season in regions known to have multiple cropping seasons, we treated those observations as seasonal. Where possible, data were aggregated by region, crop type, or fertilizer treatment to identify broad trends. Averages, medians, and ranges of emission estimates were calculated to highlight variability across studies. Statistical analyses were conducted to explore relationships between N₂O emissions and key drivers such as fertilizer application rates, soil properties, and climatic conditions. To determine the most suitable data-fitting models, we evaluated a range of approaches, including linear, nonlinear, natural log, and logarithmic models. Model selection was based primarily on the Akaike Information Criterion (AIC), which provides a robust framework for comparing models by balancing goodness of fit with model complexity and penalizing additional parameters. Statistical significance was set at a 5% critical level. For the regression analyses of N₂O fluxes across treatments, model assumptions of homoscedasticity and normality were evaluated using residual diagnostics. Where the raw data exhibited skewness or non-constant variance, log or natural log transformations were applied to stabilize the variance and improve normality. The best-fitting model for the dataset was selected based on statistical robustness and interpretability in the context of the research question. All statistical analyses were performed using the R software to ensure the robustness of the results (R Core Team, 2024).

Our estimate of SSA’s average total cropland N₂O emission is considerable, estimated at 0.24 ± 0.028 Tg N₂O-N yr^-1 based on a total cropland area of 322.63 million ha across nine countries that reported emission factors (Annex 4). This constitutes up to 20% of cropland global emissions, estimates which have increased roughly 1-3 times in the last decades, ranging from 0.4–1.4 Tg N₂O-N yr^-1 in the 1960s to 1.3–3.3 Tg N₂O-N yr^-1 in the 2010s (Tian et al., 2019; Wang et al., 2020; Xu et al., 2020). In our analysis, the average area-weighted cropland N₂O emission was estimated at 0.73 ± 0.09 kg N₂O-N ha^-1 yr^-1 with values ranging from 0 to 2.76 kg N₂O-N ha^-1 yr^-1 across all site-year combinations. This estimate excludes vegetable systems, which often exhibit disproportionately high emissions due to intensive fertilization and greenhouse cultivation, despite covering relatively small land areas. This compared to 1.24 kg N₂O-N ha⁻¹ yr⁻¹ the average total N₂O emissions reported for SSA (Tian et al., 2023). While high-emitting regions globally, such as the North China Plain, Ganges Basin, Saudi Arabia, and parts of Western Europe report emissions exceeding 4 kg N₂O-N ha^-1 yr^-1 due to fertilizer inputs that go above 350 kg N ha^-1 yr^-1 in some regions, SSA’s agricultural systems are characterized by much lower N inputs, typically under 50 kg N ha^-1 yr^-1 (Masso et al., 2017; Elrys et al., 2020). Despite the relatively low synthetic N inputs and direct emissions, SSA’s total N₂O emissions remain notable due to the region’s vast cropland area. Emission Factors (EFs) across SSA are highly variable, ranging from 0.01% to 4.1% and observed emissions (including vegetables) span from as low as 0.01 up to 113.4 kg N₂O-N ha⁻¹ yr⁻¹ reflecting the diversity of agro-ecological conditions and management practices.

The 39 studies compiled in this review span several countries in SSA (Figure 2), with Kenya having the highest representation, contributing 18 studies in 16 locations (Table 1). Measurement durations ranged from as short as 21 days to as long as 3 years. A significant portion (63%) of studies measured annual N₂O fluxes, typically spanning 12 months or more, with an average duration of 16.4 months. Seasonal flux studies (32.6%) focused on growing seasons, ranging from 2 to 9 months, with an average duration of 5.5 months. Short-duration studies (4.3%) were relatively uncommon, lasting only a few weeks to under 2 months (Annex 2). Between 1999 and 2008, studies tended to be short-term in nature. From 2009 onwards, there was a clear shift toward multi-year studies. After 2020, studies increasingly cover 12 months and more, likely due to advancements in research methodologies and automation, as well as greater funding availability.

Sampling frequency varied extensively across studies. The most common frequency was weekly, reported in 6 studies, followed by daily measurements in another 6 studies. A few studies report measurements twice per week (2 studies), and some are event-based, with measurements taken following specific agricultural events such as fertilization (4 studies). Monthly measurements are less common, with 3 studies using this frequency. From a sub-sample of N₂O emissions from maize we sought to assess if there was any effect of length of measurements. Figure 3 illustrates how the duration of N₂O emission measurements influence the reported emission estimates. Studies categorized by study duration reveal clear differences in mean emissions (kg N₂O-N ha⁻¹ yr⁻¹).

Studies that measured annual fluxes tend to report higher and more variable emissions than seasonal or short-term studies, likely because they capture key off-season pulses or episodic rainfall-triggered events that shorter studies miss. In contrast, seasonal and short-duration studies, which often cover only the cropping period or a few months, show lower median emissions and reduced variability. A Kruskal–Wallis test comparing emissions between short-term/seasonal studies (≤1 season) and long-term studies (≥1 year) indicated a trend toward higher emissions in long-term studies (χ² = 3.24, df = 1, p = 0.072), though the difference was not statistically significant. While our dataset does not allow us to quantify the extent of potential underestimation due to the limited number of replicates, the contrast between growing-season and annual emissions observed in other studies (Ortiz-Gonzalo et al., 2018; Shumba et al., 2023; Zheng et al., 2023) suggests that restricting measurements to short timeframes may risk missing emissions occurring outside peak periods. Therefore, longer-term monitoring could help improve understanding of the full emission dynamics, especially in rainfed systems with high temporal variability. A limitation of our analysis is that annualization did not account for differences between single and double cropping systems. As a result, comparisons of annualized emissions across countries may be biased, with multi-season systems appearing to emit more simply due to the additional fertilization events. Future work should separate seasonal emissions or use yield-scaled metrics to allow for more balanced comparisons across contrasting cropping systems.

We analyzed emissions across five fertilizer application rate categories (Unfertilized control or (0 kg N ha^-1), <50 (less than 50 but more than 0), 51–100, 101–150, >150 kg N ha^-1), and observed a clear trend of increasing median annual N₂O emissions (median ∼0.1 kg N₂O–N ha^-1 yr^-1) to ≤50 kg N ha^-1 (∼0.2), 51–100 kg N ha^-1 (∼0.4), and 101–150 kg N ha^-1 (∼0.5) and mean with higher N input levels, peaking in the 101–150 kg N ha^-1 range (∼1.11 kg N₂O-N ha⁻¹ yr⁻¹; EF ≈ 0.9%) (Figure 4; Supplementary Figure S1). Higher application rates (>150 kg N ha^-1) are sparsely represented. Interestingly, mean emissions in the >150 kg ha^-1 group appear low (∼0.59 kg N₂O-N ha⁻¹ yr⁻¹) but are however based on only six data points (Supplementary Figure S1), and may reflect limited representation, site-specific dynamics, or possibly as shown in one study, the shorter length of the study (3-month period). Emissions under control conditions (no added N) averaged 0.19 kg N₂O-N ha⁻¹ yr⁻¹. This baseline still shows appreciable fluxes, possibly due to N deposition, soil mineralization, or rewetting events. These trends suggest that fertilizer-induced emissions are generally proportional to application rates, and that NUEs may decline beyond 100 kg N ha^-1, aligning with recent discussions around optimal fertilization thresholds in SSA (Dimkpa et al., 2023; Falconnier et al., 2023).

Our findings support the notion that moderate N application (≤100 kg/ha) may strike a balance between minimizing emissions and maintaining yield gains. Atakora et al. (2019) further found diminishing agronomic efficiency beyond 100 kg N ha^-1, suggesting a plateau in yield response despite increased emissions. Meanwhile, our finding that emissions do not increase exponentially at high rates differs slightly from global meta-analyses which suggest nonlinear responses at very high N levels (e.g., exponential increases above crop demand thresholds). This difference may reflect the limited number of high-N (>150 N ha^-1) field studies available for SSA, or it may reflect different microbial processes and physicochemical properties in tropical soils. Although some SSA assessments (Shumba et al., 2023; Tully et al., 2023) report exponential increases in N₂O emissions at higher nitrogen application rates, our synthesis reveals a predominantly linear response across the observed dataset (Figure 4; Supplementary Figure S2). Both Tully et al. (2023) and Shumba et al. (2023) found that application rates around 50 kg N ha^-1 resulted in relatively low emissions while maintaining maize productivity. Similarly, Hickman et al. (2015) demonstrated substantial yield gains under fertilizer rates ≤100 kg N ha^-1 in Western Kenya without proportionate increases in N₂O emissions. These findings are consistent with earlier studies showing that synchronizing N inputs with crop N demand can minimize emissions while sustaining yields (McSwiney, 2005). It is important to note, however, that these relationships can vary depending on crop growth stage, nitrogen uptake dynamics, and environmental conditions (Hickman et al., 2015; Maier et al., 2022).

Our second key result is the observation of high variability in emissions within each fertilizer rate category, particularly between 101 and 150 kg N ha^-1, emphasizing that fertilizer alone cannot fully explain N₂O fluxes. Site-level variability, likely driven by interactions among soil type, moisture, climate, and management practices modulates the emissions response. For instance, pulses of N₂O in control treatments (Barton et al., 2008; Hickman et al., 2014; Zheng et al., 2019) highlight the influence of soil organic matter turnover and rainfall dynamics, particularly during rewetting events (Davidson, 2009). Supplementary Figure S2 illustrates how the relationship between fertilizer rate and emissions varies by country, with Ghana and Nigeria showing the strongest treatment-emission signals. This supports the need for country-specific emissions factors or models, especially where climate and soil types differ. In some countries, limited or highly clustered data may obscure trends altogether, reinforcing the importance of regionalized data collection.

Due to limited availability of unfertilized control measurements in many studies, we estimate emission factors (EFs) as the ratio of total N₂O emissions to applied N fertilizer, assuming minimal contribution from soil N mining or other background sources. This approach is consistent with findings by Della Chiesa et al. (2019), who showed that although background emissions can represent a substantial proportion of cropland N₂O fluxes, fertilized treatments still emit more than background levels. The resulting EFs vary broadly, from as low as 0.05% in low-input systems up to over 4% in intensively fertilized fields, with a median around 0.4%–0.7%. Although the number of replicates is limited in several countries, most of the available studies-particularly from Kenya, Tanzania and Zimbabwe report Efs below the IPCC default of 1% (Millar et al., 2004; Chikowo et al., 2007; Chapuis-Lardy et al., 2009; Hickman et al., 2014; Rosenstock et al., 2016; Atakora et al., 2019; Musuya et al., 2019; Zheng et al., 2019; Bigaignon et al., 2020; Macharia et al., 2020; Wachiye et al., 2020; Mosongo et al., 2022; Lemarpe et al., 2023; Alasinrin et al., 2025)), suggesting that this default may overestimate typical emissions in these systems.

Our synthesis indicates that N₂O emissions generally increase with fertilizer application rate across the observed range (50–100 kg N ha^-1) and varies considerably by country (Supplementary Figure S2). Overall, emission factors appear lower than global default estimates, suggesting that modest fertilizer additions in SSA (50–100 kg N ha^-1) generally produce relatively small N₂O increases compared to similar rates elsewhere. In fact, at rates above 150 kg N ha^-1, some studies reported emissions equal to or even lower than those in the 101–150 kg range (Figure 4), suggesting that yield-scaled emissions may improve with higher fertilization as crop uptake offsets absolute fluxes. While data beyond 150 kg N ha^-1 are limited in our dataset, the available evidence within the 50–150 kg N ha^-1 range suggests reasonably predictable linear responses in most contexts, though the risk of disproportionate emission increases remains under certain soil and climatic conditions that favor nitrification and denitrification. Taken together, our findings suggest that agricultural intensification in SSA could proceed with moderate N application rates (∼50–100 kg N ha^-1) as a potential baseline, avoiding excessive inputs beyond crop requirements. This recommendation is based on observed trends in the reviewed studies and should be considered as guidance rather than a definitive prescription.

Although the number of available studies varies across crop types, our synthesis provides important descriptive insights into how N₂O emissions differ among key agricultural systems in SSA. These patterns, while not intended as formal statistical comparisons, help contextualize the diversity of emission dynamics in the region and identify priority systems for future measurement efforts. Across the dataset, staple crops such as maize, sorghum, and other cereals generally exhibit moderate emissions, with maize-the most extensively studied crop-ranging between 0.1 and 4.29 kg N₂O -N ha^-1 yr^-1. Despite their wide cultivation, emissions in these systems are highly skewed, with median values (∼0.5 kg N₂O -N ha^-1 yr^-1) far below observed maxima, suggesting localized hotspots driven by site-specific management.

Sorghum and traditional cereals remain sparsely studied, though Brummer et al. (2008) reported emissions up to 1.42 g N₂O -N ha^-1 h^-1 for sorghum under fertilized conditions in Burkina Faso. In contrast, perennial crops such as coffee (1–1.9 kg N₂O -N ha^-1 yr^-1) and tea (0.4–3.9 kg N₂O -N ha^-1 yr^-1) show lower and less variable emissions. Vegetable systems, however, stand out as the most significant sources of cropland N₂O emissions in SSA, far exceeding other crop types in both intensity and variability. Emissions from these systems ranged from 0.01 to 113.4 kg N₂O -N ha^-1 yr^-1, often linked to high fertilizer inputs (25–750 kg N ha^-1) and intensive irrigation, likely under commercial production. Similar findings have been reported across SSA (Predotova et al., 2010; Lompo et al., 2012; Masaka et al., 2014; Rosenstock et al., 2016; Kurgat et al., 2018). While vegetable production is essential for dietary diversity, frequent irrigation and heavy fertilization promote denitrification under oxygen-limited conditions, thereby amplifying N₂O losses (Qasim et al., 2021).

Beyond crop type, N₂O emissions exhibited clear but heterogeneous associations with soil texture (Figure 5). In this first order synthesis, fine-textured soils showed both higher and more variable emissions relative to coarse and medium textures, with several observations exceeding 1 kg N₂O–N ha^-1 yr^-1, particularly under monocropping systems. This pattern is consistent with enhanced denitrification in clay-rich soils under conditions of elevated water-filled pore space (WFPS) and restricted oxygen diffusion (Pelster et al., 2012; Zhu et al., 2020; Hickman et al., 2021). Medium-textured soils, such as loams, exhibited consistently low emissions across cropping systems with minimal variability suggesting a potentially greater capacity to buffer moisture and nitrogen dynamics. Coarse-textured soils (e.g., Lixisols, Inceptisols) generally showed low to moderate emissions, although variability was evident, including relatively higher emissions under monocropping in some cases.

These patterns align with mechanistic understanding whereby coarse soils tend to be carbon-and water-limited, while fine textured soils may become nitrogen- and oxygen-limited under saturation (Pelster et al., 2012). Although data availability remains uneven across soil texture classes, these results highlight the importance of soil physical properties in shaping N₂O emission patterns in SSA and provide a useful basis for future targeted measurements and model development in SSA.

To evaluate the representativeness of study sites to the SSA region, we analyzed the distribution of temperature (Mean Annual Temperature; MAT) and rainfall (Mean Annual Precipitation; MAP) from the reviewed studies against regional climatic averages. We compared the distribution of MAT across two datasets: one representing the entire SSA region (WorldClim gridded dataset) and another specific to the study locations (Figure 6). This comparison helps contextualize the findings and assess their generalizability across the diverse environmental conditions in SSA. The pathways of N₂O production vary significantly across different climatic zones due to differences in temperature and precipitation.

MAT from gridded datasets exhibits a bimodal distribution, with a dominant mode at 19.4 °C and a secondary peak near 27 °C–28 °C (Figure 6, top panel), indicating a predominance of cooler highland or subtropical sites and fewer hotter lowland or semi-arid locations. Field reported MAT is also bimodal, but with notable differences (Figure 6, bottom panel). The dominant mode is higher, at 22.1 °C, and the second peak (∼28 °C–29 °C) is less prominent than in the MAT data. The observed distribution appears shifted toward warmer temperatures, and the mean and median values are slightly higher than those of the gridded data. Despite overall agreement in the primary temperature range (≈21 °C–23 °C), both datasets underrepresent sites with MAT >25 °C, highlighting a spatial bias toward moderate-temperature regions, where studies are concentrated despite evidence of higher N₂O emissions in warmer environments (Rochette et al., 2016).

During the growing season, high-intensity rainfall often leads to increased emissions by creating anaerobic conditions that enhance denitrification. Conversely, extended dry periods suppress emissions by limiting microbial activity and reducing substrate availability but conversely increase emission pulses as the substrate becomes available to microbes in a rewetting flush (Leitner et al., 2017). To evaluate how well field studies represent actual growing season rainfall patterns across SSA, we compared mean rainfall values from 16 study locations (Annex 5) for which we could geolocate sites to the NAME_2 or NAME_3 administrative level, using CHIRPS gridded estimates (∼5 km resolution). We extracted CHIRPS rainfall estimates specifically for the growing season months between 2000 and 2020, relevant to each study location. Our comparison revealed a consistent pattern: field-reported rainfall values were generally higher than corresponding gridded data for each location, with some sites showing differences of several hundred millimeters (Figure 7). This is consistent with previous studies showing that point measurements capture localized, high-intensity convective events that gridded satellite products tend to smooth out (Dinku et al., 2010; Maidment et al., 2017; Sun et al., 2018). In addition, discrepancies may arise from reporting practices: field studies often report rainfall for the exact experimental period, whereas gridded datasets reflect standardized seasonal totals.

It is also possible that gridded data may underestimate site-specific rainfall availability particularly during extreme climate years (e.g., drought in 2022-2023 or El Nino driven excess rainfall in 2024–2025). These mismatches are critical for N₂O emissions research, as rainfall governs soil moisture dynamics and associated nitrification and denitrification processes. Furthermore, studies suggest that increasing rainfall variability under climate change may further reduce the representativeness of gridded products (Kotir, 2011; Waha et al., 2013). These discrepancies reflect spatial averaging in satellite-derived data versus localized microclimatic conditions captured by field observations, underscoring the importance of integrating site-level measurements into N₂O emission modeling frameworks.

Rainfall trends were assessed at 14 study sites with ≥20 years of data (Annex 6) to evaluate climate-driven changes in precipitation. Most sites showed increasing growing season rainfall over the past 2 decades, with more than half indicating rising trends at the 5% and 1% significance levels (p < 0.05; Supplementary Figure S3). This shift toward wetter conditions could boost agricultural productivity but also increases the possibility of higher N₂O emissions due to elevated soil moisture, increasing denitrification potential and/or increasing leaching of N from soils followed by downstream environmental impacts (Omotoso and Omotayo, 2024). For example, the Dano station in Burkina Faso exhibited a significant positive slope of approximately +1.5 mm/year, translating to a 30 mm increase in annual rainfall over two decades-sufficient to trigger N₂O pulses under saturated soil conditions. Conversely, only two sites, South Africa’s Ehlanzeni and Zimbabwe’s Chitungwiza, reported decreasing rainfall trends, indicating specific climate concerns.

Despite limited studies in data-poor regions, many demonstrate strong documentation and methodological transparency, including chamber details, calibration procedures, and flux calculation equations, enhancing data usability. A key strength of these studies is their grounding in local contexts and collaboration with African research institutions such as IITA (International Institute of Tropical Agriculture), CIMMYT (International Maize and Wheat Improvement Center), ILRI (International Livestock Research Institute) and national systems, reflecting a commitment to capacity building and long-term research infrastructure in SSA (Merbold et al., 2021). Many studies report spatial coordinates and climate variables—especially rainfall and temperature—critical for interpreting N₂O emissions given biases in gridded datasets, highlighting the value of initiatives like Tahmo (2025).

Key gaps remain. Many studies omit important drivers such as soil moisture, micrometeorological data, soil texture, organic matter, and microbial activity (Davidson and Kanter, 2014). Soil moisture, in particular, is rarely directly measured and is often only inferred from seasonal rainfall patterns, through statistical correlations with N₂O emissions or reported as water-filled pore space (WFPS) in a limited number of studies. For example, WFPS values reported ranged from 5% to 10% (Tully et al., 2023) to over 70% (Baggs et al., 2006), illustrating site-to-site variation and the limited availability of continuous measurements. Fertilizer management details-type, timing, and application-are often missing, limiting cross-study comparisons and assessment of fertilizer-induced emissions (Shcherbak et al., 2014).

Standardized protocols for chamber measurements, soil properties, and detailed fertilizer reporting (Butterbach-Bahl et al., 2016; de Klein et al., 2020), inclusion of zero-N plots, yield-and input-scaled metrics, and long-term data sharing through FAIR (Findable, Accessible, Interoperable, Reusable)-compliant repositories and proper metadata documentation (Fluxnet, 2025) would improve data quality and comparability.

Kenya has led N₂O research in SSA, with studies in 16 of 19 years (2000–2018) covering diverse maize systems and following rigorous protocols, providing a strong basis for model calibration. Hotter regions (MAT >25 °C), such as the Sahel, northeastern Kenya, eastern Ethiopia, and Sudan, remain underrepresented, risking underestimation of emissions. Future work should target these zones to capture SSA’s full climatic diversity.

Beyond quantifying N₂O emissions, several studies tested mitigation practices relevant to local farming, including fertilizer rate variations, organic amendments, integrated soil fertility management, and residue and tillage management. Tillage consistently increased N₂O emissions, with conventional and minimum tillage showing higher fluxes than no-tillage (Baggs et al., 2006; Vilakazi et al., 2021; Alasinrin et al., 2025). Maize stover addition, while not improving yields, also elevated emissions, illustrating trade-offs in organic residue management (Zheng et al., 2019).

Integrated soil fertility management combining organic and mineral inputs showed that rainfall timing and soil moisture strongly influence N₂O emissions (Kimetu et al., 2007; Githongo et al., 2022). Despite these contributions, key gaps include: (1) reliance on seasonal totals or small-plot studies, (2) limited on-farm testing-only a few, like Wanyama et al. (2018), incorporate on-farm or participatory testing, and (3) few measurements of underlying mechanisms such as mineral N pools, isotopic data or microbial activity.

In this section, we bring together the findings of this study to outline future research priorities for the SSA region.