The variety of natural resources and climate conditions found in smallholding systems has resulted in a wide range of land use systems (1). Consequently, farming systems vary greatly depending on socioeconomic and agro-ecological environments, which ultimately affect system management practices in a given community (2–4). Various farming systems exist in smallholder settings in sub-Saharan Africa (SSA) based on cropping systems. Perennial crops, crop-livestock mixed farming, maize-mixed farming, root and tuber farming, and cereal-root crops are among the major farming systems in SSA (1, 4). However, crop-livestock mixed farming is the most common farming system for smallholders who produce for subsistence, accounting for roughly two-thirds of smallholder farmers’ livelihoods in Sub-Saharan Africa (SSA) (5). Crop-livestock mixed farming is preferred due to the interdependence of livestock and crops, which results in nutrient recycling within the system (6).

The challenge of nutrient cycling in mixed crop-livestock farming is that, quantities of animal manure produced within farmsteads in SSA are smaller due to the small number of livestock kept per household, unlike in developed countries where manure is considered a problematic waste (7). As a result, smallholder farmers are forced to use little manure, which does not meet nutrition crop requirements; thus, crop production in this region relies heavily on natural soil fertility (8, 9). It is estimated that 60–80% of household income is generated at the expense of natural soil fertility (10). Reports show that over the last three decades, nutrient mining from arable lands in SSA has been estimated to be 660, 75, and 450 kg ha^-1 yr^-1 of N, P, and K, respectively (11). Consequently, these practices contribute to land degradation through nutrient mining, threatening the sustainability of existing farming systems (12). Understanding the interaction between farmers’ socioeconomic attributes of soil fertility management and soil nutrient depletion is critical in developing appropriate approaches to address soil fertility problems (13–15). Designing biophysical nutrient management without consideration of socioeconomic factors is likely to yield low adoption, despite being technically sound (14). In the past, soil management interventions were derived from less participatory, top-down policies and thus did not work because they appeared to interfere with farmers’ decision-making (16). Adoption of nutrient management is determined by accessibility and affordability of the technology and the respective requirements for input, materials (17) and labor. To that end, smallholder farmers in SSA often respond to these challenges in various alternative ways by using easily available resources within their environment, such as animal manure, mulching, or intercropping (18). However, due to the insufficient quantity and often low quality of animal manure produced within farmsteads, their contribution to improving soil fertility is negligible (19, 20).

The majority of smallholder farmers are still struggling to increase crop production in order to feed more people, a result of the high population growth rate (21, 22). However, reaching potential yields has remained a challenge due to smallholder farmers’ reliance on rain-fed agriculture, insufficient soil nutrient supply, use of low-yielding varieties, and a lack of mechanization (3, 14). According to Aschonitis et al. (23), the introduction of the “green revolution” in the 1960s greatly improved crop yields ranging from 3 to 5 MT ha^-1 in Asia and China, respectively, to 10 MT in North America, Europe, and Japan through the use of improved crop varieties, fertilizers, pesticides, and advanced farm machinery. The green revolution was not realized in developing countries, leaving Africa with the lowest yield at around 1.5 MT ha^-1 (24), due to the inaccessibility and high cost of the agricultural technologies. Nonetheless, previous research findings, such as those by Omuto and Vargas (25) and Takele et al. (14), revealed that agricultural technological change, such as the use of advanced machinery, high yielding varieties, fertilizers, and pesticides, has been linked to land degradation in many arable lands, including erosion, salinity, and soil nutrient depletion.

Global population growth is expected to reach 9.7 billion people by 2050, with African population growth expected to reach 2.5 billion in 2050, up from 1.3 billion in 2020, and the SSA population expected to reach 3.1 billion by 2100, up from around 1.24 billion today (3). With these population projections, the governments, including those in the SSA region, have to take critical actions to be able to feed the growing population by addressing the rapid decline in soil fertility and increased food constraints (3). It is, therefore, important that global and/or regional food production be increased through holistic strategies to meet the demand of the growing population. While research centers have developed many promising systems of soil amendment techniques for nutrient enrichment, the majority of them rely on mono-disciplinary approaches with a focus on the biophysical aspects (26–29), with little consideration of the socio-economic aspects. Integrating biophysical and socio-economic disciplines could address the problem of soil nutrient depletion more holistically (29). It all starts, nevertheless, with estimating nutrient budgets, and this has been gaining popularity among the researchers (30). A nutrient budget can be viewed as a reliable indicator for nutrient mining and related land degradation, allowing for improved soil nutrient management. The nutrient budget was defined by Bindraban et al. (31) as the difference between the system’s nutrient inputs and outputs within predefined spatial-temporal boundaries. The difference is calculated based on the nutrient stocks present within the top 30 cm of the soil profile (32) and the depth where most of the crop roots are active (33).

There are several modes available for better evaluation of nutrient flow and budgets and the limitations of soil nutrient content in SSA. A system for quantitative evaluation of the fertility in tropical soil (QUEFS), for example, was designed to assess the efficacy of N, P, and K ratios during fertilizer application (34). Other models, like NuMass, were developed to diagnose soil fertility in terms of N, P, and soil acidity (35). However, the NUTMON methodology, introduced in 1990 by Stoorvogel and Smallings, mainly focused on nutrient flow and balance by indicating the nutrient inputs and outputs of a certain land use and farming system (36–38). Because it combines biophysical and socioeconomic approaches to soil fertility management, the NUTMON concept has been opened to a wide range of studies related to nutrient budget and flow (30). NUTMON is essentially a decision support model that has been modified from the Nutrient Balance Model (NUTBAL), which was previously developed to generate quantitative nutrient balances for the major macronutrients (N, P and K) for African land use systems (33). NUTMON, goes beyond NUTBAL by including, in addition to nutrient balance, changes in land use, farm activities, and economic analysis to generate qualitative and quantitative nutrient stock and flows data within and outside the farm (29, 39–41). The economic analysis tool was included in order to estimate the farm’s economic performance (33, 42). As a result of the integration of biophysical and economic performance, farmers and researchers can make recommendations for alternative methods of implementing Integrated Nutrient Management (INM) while keeping the underlying constraints in mind.

The NUTMON methodology can be used by researchers and farmers to assess the environmental and financial sustainability of tropical farming systems (11, 29). Other research has shown that NUTMON can be used to assess the degree of nutrient mining in an agro-ecosystem and the effects of the various nutrient management strategies on soil nutrient stocks (32, 33). NUTMON categorizes inputs into five groups (N1 to N5): incoming nutrients from fertilizers (mineral and organic), wet and dry deposition, nitrogen fixation, and sedimentation. Harvested products (grain, tubers, or animal products), crop residues, leaching beyond the rooting zone, gaseous N and S losses (denitrification, valorization, and burning), and erosion are the five output categories (OUT1–OUT5) (16, 40, 43). As shown in Figure 1 , a farm-level NUTMON consists of a structured questionnaire, a data base, and two statistical models: one for calculating nutrient flows (NUTCAL-model) and the other for calculating economic performance (ECCAL-model) (33).

Because there have been numerous studies on NUTMON in SSA and elsewhere, this review intends to investigate the sustainability of smallholder farming systems in terms of nutrient flow and balance by utilizing previous studies on nutrient monitoring (NUTMON) in smallholder farming system soils. However, due to inconclusive results from diverse settings of smallholder farmers worldwide, this review will only focus on Sub-Saharan Africa (SSA) as a representative of other areas with extensive nutrient mining and low-resource farming systems. As a result, this paper intends to; (i) examine biophysical attributes that influence soil fertility management in SSA smallholder farming systems, (ii) provide narrations on how socio-economic categories affect farming systems and nutrient flow and balance., (iii) Identifying existing smallholder farming systems and their soil nutrient balance status (iv) Identifying the nutrient monitoring (NUTMON) research gap in SSA.

The emphasis of this review is on smallholders’ characteristics and soil nutrient management in SSA. Soil fertility, being the most important indicator that determines the capacity of the soil to produce crops, is controlled by many factors ranging from biophysical (110) to socio-economic (14, 106). In the past, smallholder farmers in SSA used to cultivate land by moving from one place to another (the practice is known as shifting cultivation); thus, farms were left fallow to rejuvenate their fertility (9, 111). Clearly, the aforementioned practice is no longer appropriate due to increased population, which has resulted in pressure on agricultural lands (112, 113). The current smallholder farming systems in SSA are comprised of homestead farms (home gardens), which are typically small pieces of land (1 acre), and distant farms, which are relatively large (> 1 acre but less than 5 acres) (3, 114). We discussed soil fertility management by smallholder farmers in terms of biophysical and socioeconomic attributes in this section.

According to the findings of this review, the majority of studies on nutrient flow and balances have been conducted on mixed crop-livestock systems. The most researched farming systems in Kenya are highland perennials (dominated by coffee, bananas, and tea, with annual crops such as maize and legumes) and lowland cereals such as maize, sorghum, millet, and root crops like cassava. All of these systems are linked to livestock like cattle, goats, sheep, and poultry ( Table 2 ). Similarly, the dominant farming systems in Ethiopia are crop-livestock farming systems, mostly enset-based with banana and cereals like barley, wheat, and teff and vegetables in the highlands, and teff-based with barley, wheat, maize, sorghum, and legumes in the lowlands ( Table 2 ). Other farming systems investigated include agro pastoral in Nigeria, Mali, Tanzania, Uganda, and Burkina Faso (46, 47, 49, 72, 73, 81, 82, 137, 138), and maize-mixed with crops such as legumes, cassava (57, 62, 71, 75). Based on the current review, Table 2 shows the average nutrient balance in the most studied countries over the last three decades. Regardless of the farming systems, the overall mean nutrient balances (particularly for N and K) reported in all studies were negative, with P being relatively small positive. Nitrogen, unlike other soil nutrients that are most likely derived from parent materials, must be supplemented externally. Nonetheless, N is the most required nutrient by plants and the most mobile nutrient, making it easily lost from the soil through harvested products, leaching, volatilization, erosion, and denitrification if not managed properly (88, 139). Potassium, on the other hand, is the third most important nutrient for crops after phosphorus (140). Despite the fact that potassium is abundant in soils, the readily available pool is so small, and the fate of K in soil is almost similar to that of N (141). This has been attributed to the high negative nitrogen and potassium balances found in most studies, particularly when the same field is cultivated for an extended period of time with little or no nutrient replenishment. The degree of nutrient depletion between N and K varied from study to study, depending on the cropping system in a given area. For example, Wortmann and Kaizzi (54) found a high negative balance of K in banana, bean, and sweet potato cropping systems, while maize and soybean cropping systems had a high negative balance of N in the same agro-ecological zone. Similarly, studies by Amann et al. (21); Diarisso et al. (103); and Abegaz (18), to name a few, found a high negative balance in K compared to N, whereas De Jager et al. (27); Bekunda and Manzi (67); and Tunya (98) found a high negative balance in N compared to K. The positive P balance observed in some studies could be attributed to the residue effect of applied P fertilizers or manure in those farming systems (98, 142). It is estimated that 70-90% of applied P fertilizer in soil becomes sorbed to soil particles and transforms into less available forms very quickly (143).

Crop harvesting and soil erosion are reported to be the major nutrient outputs in smallholder farming systems. According to Smaling and Fresco (40), harvested crop products and crop residues exported up to 61, 11, and 46 kg ha^-1 year^-1 of N, P, and K, respectively, accounting for 50–70% of total losses, while soil erosion contributed roughly one-third of total losses. Other nutrient outputs, such as leaching and gaseous losses, varied depending on soil type, climatic conditions, and management practices, but their contributions to nutrient loss were significantly low (40, 68).

The current review discovered significant differences in nutrient balances among farmers with similar agro-ecological conditions (i.e., soils, climate, and infrastructure). It demonstrates the importance of management practices in nutrient flow and balance. Given the diverse settings of smallholder farming systems across all studies, we found it difficult to make comparisons on nutrient balance in various farming systems for the current review due to variation between studies in terms of type of balance (partial or full), scale of study (farm, plot, village, district), and wealthy categories (rich, medium, poor). However, studies comparing wealthy farmers found that farm nutrient flow and balances were more positive in richly endowed resource farmers (i.e., those with high resource access) than in poor resource farmers [Shepherd & Soule (5, 50, 56, 62, 81, 87, 114, 144)]. It follows that richer farmers use large amounts of inorganic fertilizers and organic inputs on their fields (5). Moreover, richer farmers own more livestock which accounts for large amounts of manure produced on their farms and are likely to outsource additional manure/and or crop residues as fodder (145). This contrasts with poor farmers, whereby most of them are faced with multiple constraints including shortage of family labour and low purchasing power to inorganic fertilizers and/or additional source of manure (28, 62, 114). However, this is not always the case; for example, Elias (74) found high nutrient depletion, particularly N (-85 and -102 kg ha^-1 yr^-1) in some rich farmer fields, whereas poor farmers had a low negative N balance (-50 and -56 kg ha^-1 yr^-1). This explains that not how much input is put into the soil but how to manage those inputs matters.

Similar observations are reported in studies that compared homestead farms with distant fields ( Figure 3 ) (62, 84). As smallholder farmers have limited access to inputs and focus on surviving, a considerable soil fertility gradient usually develops from the homestead to the so-called “far” fields. (68). The limited attention to soil fertility management on fields that are located away from home derives directly from the low farmers’ income (62, 114). Hence, the cost of transporting inputs such as fertilizers (both inorganic and organic fertilizers) and the labour requirement preclude optimal nutrient management in these fields. Gender differences also contributes to differences in adoption of soil fertility management, hence nutrient balance. A study by Gebresamuel et al. (106), found that female-headed households had more positive N, P, and K balances in their fields than male-headed households. Among other things, it appears that female-headed households have fewer animals or no livestock, resulting in low crop residue removal from their fields (106). Although women seem to be good soil fertility managers, due to multiple tasks obliged by women such as handling children their effort in soil fertility management is negligible thus crop production remains below potential yields. There are conflicting findings regarding the relationship between farming systems and nutrient balance; however, most studies concluded that farms with high levels of inputs (both organic and inorganic) performed well in terms of productivity and nutrient balance and stocks.

The decline in soil fertility in SSA soils is threatening soil productivity in most arable lands. Most nutrient balance studies have found more negative nutrient balances, particularly for macronutrients (N, P, and K), and the trend continued to rise year after year. In central Kenya, for example, the N, P and K balances were -55, 9 and -15 kgha^-1 yr^-1, respectively, in 1998, but had more than doubled to -116.2, -22.1, and -31.7 kgha^-1 yr^-1 five years later (56). Studies indicate that understanding the dynamics of farm nutrient flow is fundamental for proper implementation of appropriate soil nutrient management techniques (38, 146–148). Over the last three decades, numerous researchers have developed nutrient balance and budget models, including the system for quantitative evaluation of fertility in tropical soils (QUEFS), universal soil loss equation (USLE), nutrient monitoring (NUTMON), and material flow analysis (MFA), to name a few (21). NUTMON, however, is the most popular model for evaluating nutrient flow and budget in several SSA countries, particularly in East Africa, due to its ability to integrate both biophysical and socioeconomic approaches (29, 33, 38). When used properly, NUTMON provides an insight indicator for soil nutrient depletion and/or surplus, aiding in the planning of proper soil management practices (37). A better understanding of the nutrient budget may also raise awareness among agricultural stakeholders and policymakers. Nutrient budgets at the farm level can provide a comprehensive picture of nutrient flow from the village to the national level, informing stakeholders’ interventions (149).

In this review, we discovered that almost all NUTMON studies focused on the major three nutrients, namely N, P, and/or K, rather than other essential nutrients such as magnesium, calcium, sulfur, and micronutrients. Reports have shown that micronutrients such as zinc, iron, boron, and copper have been gradually decreasing in SSA soils, leading to malnutrition, particularly in children under the age of five (148, 150, 151). Iron and zinc deficiencies, for example, have been reported in many African countries, particularly among children under the age of five (151, 152). Despite this, little to no effort has been made to monitor these plant nutrients. Thus, researchers should investigate the balances of other essential plant nutrients. Furthermore, in most SSA countries, a lack of research capacity, particularly for long-term trials, has made it difficult to draw valid conclusions from NUTMON research (120, 153). Despite the success stories of NUTMON reported in SSA and elsewhere, the validity of the remains in doubt (30, 154). Transfer functions, which rely heavily on regression models, are too general and may not be applicable everywhere, contradicting the actual losses (36). Simply put, extensive use of NUTMON tools may result in an overestimation or underestimation of actual nutrient losses (30, 64). This suggests that there is still much work to be done on NUTMON, particularly in smallholder farming systems.