The study was conducted on the slopes of Mount Kilimanjaro along an elevation gradient in smallholder farming communities in Hai District (latitude −3.232 to −3.384 S and longitude 37.238 to 37.281 E; Figure 1). Permission to carry out the research was granted by the district council and community-based officials, with all farmers involved providing their consent. The study was done under the Farmer Research Network (FRN) project with an official research permit from the Tanzania Commission for Science and Technology (2021-181-NA-2021-061). The annual temperature ranges from 15 to 30°C, with a mean annual rainfall ranging from 500 mm in the lowlands and 2,000 mm in the highlands. Rainfall is bimodal with a long rainfall season from March to June and a short rainfall season from November to December. Altitude ranges from 700 to 1,500 meters above sea level (m.a.s.l) (Lema et al., 2014). Soils in the higher elevations are Acrisols (https://soilgrids.org/) (extracted study area map in Supplementary Figure 1) which have low base cation status, low activity clay on topsoil, and more organic matter (FAO, 2006). Soils in the middle elevations are ferralsols with good physical properties and poor chemical fertility having relatively weak cation retention capacity (Massawe and Mrema, 2017). Soils in the lower elevation are chromic luvisols having high base status and low activity clay on topsoil (FAO, 2006). Farming practices in the area consist of smallholder agriculture where most farmers practice intercropping of maize and beans. Other crops grown include bananas, coffee and leafy vegetables. Common beans are often grown as a mono-crop during the short rain season and as an intercrop with maize in the long rain season and have been grown in the region for several decades. Farmers in higher and middle elevations grow common bean on rented land or in small pieces of land around their homestead while most farmers in the lower elevations own larger farm fields.

Farmer's knowledge, attitudes, and practices with regard to bean cropping and particularly SPD were collected in the Narumu Ward of Hai District. Focus group discussions (FGD) with 64 farmers within eight different groups were facilitated with the assistance of village extension officers and research assistants involving farmer representatives nominated by farmers in different areas. Eight participants were selected for each focus group in a purposive way to cover a wide possible range of experience and knowledge in common bean production. Preliminary information on the level of farmer knowledge and their local taxonomy of SPD were collected. Farmers' observations and knowledge of field disease symptoms and insect damage were also recorded. The discussions allowed farmers to share their knowledge and experiences regarding SPD and their management practices including synthetic and pesticidal plants use. General guiding questions were used, and provoking questions were asked when necessary (questions available in online Supplementary material 1). Topics discussed included knowledge on common bean production, SPD, use of pesticidal plants and other practices for controlling pests and diseases, constraints to their control as well as methods observed to be successful by farmers in managing SPD.

A farm-level, individual questionnaire survey was also carried out to capture farmer demographic information as well as their knowledge of bean fly, foliage beetle and root rot disease damage, and management practices including pesticidal plant use practices. A total of 54 common bean farmers were interviewed individually. Some participants in the farm-level questionnaire were also involved in the FGDs but new participants were included who did not have experience in using pesticidal plants to control insect pests and diseases. To provide clear references for the responses and observations of pest incidence in the field, live insects were shown to farmers to see the color and size of insect pests. Also, damaged plants showing feeding or oviposition behavior of bean fly and bean foliage beetle were displayed before starting the interview. Questions for the farm-level individual questionnaire were prepared using Kobo Toolbox (https://kf.kobotoolbox.org/) and data were collected using smartphones with Open Data Kit (ODK) open-source mobile data collection software (https://opendatakit.org/). The questionnaires were pre-tested in a pilot study before being used by the targeted respondents.

Farmer practices with respect to soil fertility management were collected in individual discussions in farmer fields with the same 54 farmers involved in questionnaires about pest and disease management. Information on livestock keeping, herd sizes, land ownership and soil management was collected as part of a nutrient balance survey. A partial nutrient balance reflecting nutrient inputs, harvests, and estimated nitrogen (N) fixation was adapted from the NUTMON framework and previous nutrient balance assessments in smallholder systems (Smaling and Fresco, 1993; Ampofo et al., 1998; Vanek and Drinkwater, 2013; Nyamasoka-Magonziwa et al., 2020). The nutrient balance information on field size, crop type, harvests, and types and amounts of manure, fertilizer, and other inputs were gathered using a second survey implemented in ODK (see above). Nutrient balance information was collected on 45 farm fields belonging to farmers included in the larger field-level individual questionnaire survey. Field areas were assessed, and partial nutrient balances were obtained as nitrogen (N), phosphorus (P), and potassium (K) in inflows (mineral fertilizer input, compost, manure, and estimated N fixation of beans as equal to harvested N) (Ojiem et al., 2007); minus export flows (harvested grain and nutrients within crop residues when these were exported). Nutrient contents of crops and manures were based on data from similar smallholder systems in the literature and unpublished data from analyses. The frequency of use of organic nutrient sources and the retention of crop residues on fields was also assessed in the survey to characterize the tendency of farmers' management to sustain soil organic carbon.

Composite soil samples were collected from the 45 farm fields in the field-level nutrient balance survey. The samples were collected to evaluate key soil properties that are known to influence plant resilience and the presence of soil-borne pathogens using a tool kit of accessible assessment methods (Nyamasoka-Magonziwa et al., 2020). These measurements included particulate organic matter (POM), permanganate-oxidizable soil carbon (POXC), aggregate stability, soil pH and phosphorus. From each field, five sub-samples were sampled to a depth of 20 cm, collected, mixed, and a sample of 1 kg was taken. Care was taken to not break aggregates during mixing for subsequent aggregate stability analysis. The soil samples were air-dried and sieved through a 2 mm sieve prior to analysis, except for the aggregate stability analysis for which a portion of the sample was carefully broken along natural planes of weakness by hand to pass a 10 mm sieve prior to drying and stability analysis (see below).

Soil pH was measured using a portable pH meter (Test Equipment Depot, Woburn, MA, USA) and particulate organic matter (POM) was determined by gentle wet-sieving of particles between 250 microns and 2 mm, followed by density flotation and decanting of organic matter in clean tap water. Permanganate oxidizable carbon (POXC; adapted from Weil et al., 2003) was measured based on the oxidation of labile soil C by potassium permanganate in a calcium chloride solution (0.015 M KMnO₄ and 0.1 M CaCl₂). Soil available phosphorus was determined using a modified Olsen method (Olsen et al., 1954) in which Olsen solution (0.5 M NaHCO₃, adjusted to pH 8.5 using NaOH) was used to extract reactive P which was then acidified using sodium hydrogen sulfate (NaHSO₄) prior to analysis for dissolved phosphate using a reagent pack for the molybdate blue colorimetric method and a low-cost colorimeter (Hanna Instruments, Providence, RI, USA). Aggregate stability was determined using wet sieving methods adapted from Nyamasoka-Magonziwa et al. (2020) and the mean weighted diameter (MWD) of three size classes (2–10, 0.25–2, and 0–0.25 mm diameters) calculated as a single parameter summarizing aggregate stability, with larger MWD representing higher stability. Soil texture was assessed by the USDA feel method (Thien, 1979; Rictchey et al., 2015).

Discussions with farmers in FGDs and during surveys were recorded and transcribed together with notes taken during the sessions. Transcripts were then coded and analyzed to identify recurrent themes and patterns on farmers' awareness of soil-borne insect pests and diseases, the use of pesticidal plants and other soil management issues. Data about soil nutrient and residue management were analyzed using a principal component analysis (PCA) to understand associations between site and management variables. Regression analysis was used to assess which site, soil, and nutrient management parameters were associated with farmers' perceptions of damage by soil-borne pests and diseases, subject to the insights provided by the PCA on potential confounding of soil, site, and management effects. To examine the relationship between soil factors and SPD damage, a single stunting and wilting disease index was created that was rated 1 when any symptoms of stunting and wilting were observed by farmers (e.g., wilting of seedlings, death of taproot, adventitious root formation, stunting), and 0 when none of these symptoms were observed. Multiple correspondence analysis assessed potential relationships between respondent demographics from the farmer surveys (age, gender and education level) and their responses on soil-borne insect pests and diseases and management practices. All analyses were carried out in XLSTAT statistical package version 2022 (Addinsoft, New York, USA).

Most farmers interviewed were male (>60%) and over 80% were over 30 years old with most farmers (>50%) aged over 50 years (Table 1). Most farmers (>85%) had only completed primary education, and over 60% of respondents reported that farming was their main livelihood activity. From the sample of interviewed farmers, synthetic pesticide was reported to be a major (>72%) control method used by most farmers for whatever insect pest or disease appeared in common bean including SPDs (Figure 2). Less than half (37%) were using pesticidal plants for the management of insects in their field. On cultural management practices, >50% of farmers reported practicing intercropping of common bean and maize, >50% reported practicing timely planting, and < 30% practiced crop rotation. Demographic parameters influenced some responses provided where multiple correspondence analysis showed some relationships (Figure 3). For example, the use of synthetic pesticides, intercropping and early planting correlated significantly with farmers of age >60 years whereas crop rotation correlated with farmers with age between 40 and 60 years having primary education and female farmers. Among farmers who reported using pesticidal plants, >35% used Tephrosia vogelii, 7% used Tithonia diversifolia and >3% used Lantana camara. All farmers reported that the parts of pesticidal plants used were leaves, which were extracted in soapy water and sprayed onto crops. These pesticidal plants were reported to work well for a variety of pests and diseases on foliar parts of plants but were not perceived as effective on SPDs.

Most farmers (60%) owned land with a majority of respondents (44%) reported farming on a land size of 0.2 ha. Farmers with larger sizes of land accounted for 40% having a land size between 0.2 and 0.4 ha. Most farmers renting land had land sizes of 0.2 ha (>55%) while almost all (90%) farmers farming on owned family farms had land sizes between 0.2 and 0.4 ha (Table 2). In the study area, fertilizer use bore no relationship to whether farmers permanently own or borrow/rent land for farming and there was no significant relationship between land size and fertilizer use.

Most farmers owned chickens (>60%) and cattle (>50%) while those who owned goats, ducks, pigs and sheep constituted a small proportion of the respondents. More than half of farmers with cattle had a small number of cattle ranging from 1 to 6 with most farmers owning 1–4 cattle (50%) kept at their homestead surroundings. These zero-grazing practices promoted farmers to collect bean and maize residues from their fields to feed the cattle. Although it was expected that only farmers with livestock would collect crop residue for feeding their animals, all farmers with or without livestock harvested crop residues, whereas farmers with no livestock would sell the residue to farmers with livestock. Cattle manure was the only manure used for managing soil fertility in farms; however, only 3 of 45 maize and bean fields in the survey received manure, and at very low rates, suggesting that recycling of organic nutrient sources to bean and maize fields is very uncommon in these systems. Rather, all but one of the surveyed farmers reported the use of manure on homestead farms or plots near their houses ranging from small vegetable plots to large mixed plots with coffee, banana and beans (Table 3), which were often adjacent to the homestead.

Although highly variable, the manuring rates on these near plots were much higher than in the maize/bean plots, averaging 14.3 Mg ha⁻¹ (standard deviation = 19.4 Mg ha⁻¹) across farms. Over 80% of manure was from farmers' own livestock with few obtaining manure from family or neighbors. Most farmers (>70%) reported that the distance to the farm was the major constraint for using manure on bean and maize fields, particularly where farmers were farming at a distance greater than five kilometers from the homestead. Another constraint reported by farmers was a lack of improved practices for composting manure as an alternative to passive methods in which the manure decomposes for 6–12 months.

Focus group discussions indicated that farmers in the study area were aware that insect pests damaged common bean from within the soil (Table 4). The symptoms of crop damage by soil-borne insect pests are observed mainly by uprooting the plants; however, only farmers from one group discussion reported uprooting plants to see if the damage was caused by soil-borne insects. Farmers reported root rot diseases in the field where three groups of farmers knew of such disease as soil-borne, and they reported having seen it in their farms. They perceived the diseases as soil-borne as they could observe field sections of healthy bean seedlings without any symptoms then turning yellow a few weeks after planting. Five groups reported wilting plants, which they thought were caused by soil pests or pathogens. Participants provided with pictures of plants with Pythium or Fusarium root rot were unable to differentiate between the two pathogens. Many farmers did perceive the link between soil-borne diseases and plants turning yellow, wilting, and dying a few weeks after planting. However, participants' knowledge was limited with some remarking that root nodules on beans that provide N fixation via the symbiosis with Rhizobium bacteria were a symptom of the disease.

Field surveys showed that farmers reported bean fly and bean foliage beetle to be present in their fields for more than 3 years. Almost all farmers (90%) reported bean foliage beetle adults to be present in their fields with holes on the bean seedling leaves as the main observed symptom of insect damage. Only 6% of farmers reported seeing bean fly adults in their field but >35% of farmers reported wilting and dying of seedlings as the widely observed symptom of damage by the insect. No participant reported the presence of larvae or pupa in the soil nor attributed plant damage to them. Some farmers misdiagnosed bean fly damage, indicated by stunted plants with yellowed leaves, as anthracnose (Colletotrichum lindemuthianum), a fungal disease normally triggered by cold weather and high humidity. About 37% of respondents reported having root rot disease in their fields were from higher altitudes with higher humidity. All participants were aware of one key symptom of root rot, which was leaves turning yellow and dropping, followed by plants wilting and dying.

Soil assessments showed that ~70% of the fields had active carbon (POXC) in the low and very low ranges (< 400 mg kg⁻¹) with the remaining group having a medium amount of POXC with the highest recorded value of < 700 mg kg⁻¹ (Supplementary Table 1). Almost three-quarters of the soils were either low or very low in available Olsen P (< 10 mg P kg⁻¹; Supplementary Table 1) with a median value for Olsen P of 5.3 mg P kg⁻¹. Nevertheless, an analysis of soil nutrient balances on the survey fields showed that most farmers had field P balances not different from zero, regardless of whether fields were farmer-owned or not (Figure 4), due to the widespread use of synthetic fertilizers (over 95% applied fertilizer) that replaced crop exports. Meanwhile, fields owned by farmers, mostly at lower elevations, had negative N balances while those not owned had balances not different from zero on average. Negative N balances on owned fields likely relate to more frequent growing of maize in intercrops with beans on these fields, resulting in high N exports from the cereal crops, whereas on rented fields beans were more often the sole crop with lower N exports due to N fixation and the lack of a non-fixing cereal (Figure 4). On the other hand, the infrequent use of manure and the removal of residue across all the bean and maize fields, described above, is notable because it implies that little organic carbon is being returned to any of these fields (Table 3).

A principal component analysis (PCA) of site and management variables for fields (Figure 5) demonstrated a strong first component encompassing just over 30% of variability, combining pH, elevation, total inorganic N inputs, organic matter (POXC) as well as soil available P, and aggregate stability to a lesser extent. Aggregate stability (MWD of aggregates) was also positively correlated to POXC (r = +0.55, p < 0.0001, n = 45). Meanwhile, soil nutrient balances, total P fertilizer additions, and soil clay content were segregated into a second component (Figure 5, PC2). The third axis of variability in PC3 with 81% of the variability was loaded highly only on MWD related to aggregate stability (Supplementary Table 2). The negative association of fertilizer N inputs with elevation likely is another result of the higher frequency of maize-bean intercrops in lower fields with N fertilizer inputs during both surveyed seasons, while the association of total inorganic P inputs with P balances indicates how these inputs may be dominating the P balances across all the surveyed fields.

Logistic regressions testing the association of each of the site and management principal components, with a binary variable testing whether wilting, stunting, or root damage (S/W) was observed at each site, showed that the three principal components were all significantly associated with these farmer observations related to soil pests and diseases (Table 5). Within principal component 1 (PC1), Soil pH, soil active Carbon, and Olsen P were negatively correlated to S/W observations so that for example, fields with higher active C had lower rates of S/W reported by farmers (in fields with S/W, POXC levels of 300 ± 23 mg C kg⁻¹ compared to 467 ± 41 mg C kg⁻¹ in fields without S/W). By contrast, only total inorganic N inputs correlate positively with S/W, while elevation in itself was not significantly related to the S/W synthetic variable. Principal component 2 (PC2) parameters were not significantly related to S/W when tested separately (Table 5), while in PC3, fields with higher levels of aggregation (higher MWD) had lower levels of S/W.

The surveys and focus group discussions with farmers suggested that most do not uproot seedlings or investigate roots in other ways when SPD are suspected. This is partly because farmers are not fully aware of how to assess and confirm their presence but also because it is considered a destructive method that could reduce crop yield. Farmer knowledge about bean fly and foliage beetle life cycles was largely absent with little understanding of how to observe plant damage caused by their larval stages in the soil. Damage symptoms of these pests can be similar to those arising from root rot diseases (Abate and Ampofo, 1996), where farmers made general descriptions of the yellowing of plants, stunting, wilting and death. Our results are consistent with other research that reported farmers are often confused by damage caused by bean fly and root rot disease (Buruchara et al., 2010). Holes on bean leaves were commonly reported symptoms of adult bean foliage beetle damage; however, farmers generally did not know the larval stage was a soil pest, as has been observed in other studies (Abate et al., 2000). Further confusion was evidenced by farmers who wrongly attributed bean fly larval damage to the fungal disease anthracnose (Colletotrichum lindemuthianum). This may indicate the inability of farmers to recognize the minute adult bean fly as a pest and then rather describe wilting and dying caused by bean fly infestation as a disease (Letourneau, 1994). Our surveys with farmers highlight that plant damage symptoms from different soil-borne pests and pathogens, such as wilting and stunting, are difficult to distinguish by farmers.

This lack of awareness about SPD has clear implications for farmers' abilities to take appropriate management actions. The first line of defense for most farmers is to use cultural practices such as early synchronous planting, using a varietal mixture of seed and intercropping (Abate and Ampofo, 1996). Such practices are considered affordable and are part of local customs, as indicated by the large number of practices in use by the oldest farmers in the MCA (Figure 3). Farmers in the area grow local bean varieties with seeds obtained from their previous harvest. The choice of varieties is based on productivity and adaptability to environmental conditions such as temperature and rainfall. This finding is consistent with research in Uganda that indicated varietal adaptability to farm conditions was one of the keys to farmer preferences for seed varieties (Bruno et al., 2018). Intercropping is commonly used in a variety of cropping systems to reduce pest incidence such as the bean foliage beetle (Farrow et al., 2011; Srinivasan, 2014). However, intercropping of beans with maize was reported to be ineffective in reducing bean fly incidence (Abate and Ampofo, 1996). Some surveys suggest that farmers do not perceive intercropping to be a pest management strategy as they report a prevalence of insect damage despite the use of the method (Laizer et al., 2019), and the observation may be quite correct in that intercropping over many adjacent farm fields supplies pests and pathogens with a constant supply of host bean plants. By contrast, crop rotation is a common cultural practice known to be effective in managing root rot and soil-borne insect pest in common bean (Mohamed and Teri, 1989). Unfortunately, in the study area, very few farmers practiced crop rotation, and among those who did, they were not doing it appropriately as 2–3 years rotation with non-host crops as recommended (Buruchara et al., 2010).

A commonly reported approach used by farmers in the study area for preventing losses from SPDs is the use of synthetic pesticides. Prior studies have noted this to be a common practice in different regions (Mwanauta et al., 2015; Stevenson et al., 2017; Andersson and Isgren, 2021). Insecticide seed treatments and foliar sprays have been recommended for bean fly and foliage beetle (Mwanauta et al., 2015), whilst fungicides and fumigants have been recommended for the control of root rots (Mahmood et al., 2017). Despite the often high cost of synthetics, many farmers use them as they fear the loss of their livelihoods and ability to feed their families. The economics of using relatively safer synthetic products is often restricted by a farmer's perceived limited capacity and autonomy to reduce pesticide use as well as what is considered “normal” practice (Bakker et al., 2021; Deguine et al., 2021). Use of synthetics is exacerbated at the smallholder level as many smallholders are unaware of alternative agro-ecological approaches (Anjarwalla et al., 2016), nor are they often aware of the hazards of synthetic use to their health (Nicolopoulou-Stamati et al., 2016; Andersson and Isgren, 2021) and the environment (Suganda et al., 2020). Farmer use of pesticidal plants has been undermined through commercial advertising and systematic promotion of synthetics over several decades by businesses and governments (Isman, 2008; Lykogianni et al., 2021). Desneux et al. (2007) advocate studying the sub-lethal effects of synthetic pesticides on natural enemies to be sure of their safe use in order to optimize IPM programs involving the use of both natural enemies and pesticides against pests. Further, Belmain et al. (2022) suggest the elimination of synthetic pesticides in order to support these natural processes and food sovereignty. Mkindi et al. (2021) recommend pesticidal plants as one alternative to synthetic pesticides. One-third of farmers in the study area reported using pesticidal plants such as T. vogelii and Tithonia diversifolia. Extracts of T. vogelii have been reported to have broad pesticidal properties which make them effective against most foliar and soil pests (Mkenda et al., 2014; Stevenson et al., 2017). Its use can also increase plant resilience and growth by acting as a foliar fertilizer (Mkindi et al., 2020). Other pesticidal plant species such as neem (Azadirachta indica) have been shown to be effective against soil-borne insects (Abate and Ampofo, 1996; Buruchara et al., 2010) where Karel and Rweyemamu (1984) reported A. indica to be effective in controlling bean fly in common bean.

While the cultural practices and crop protection approaches discussed above have relatively direct and short-term impacts on arthropod pests, soil fertility and organic matter management that we explored in this study are important agroecological aspects of smallholder intensification, with effects on pests and diseases at a variety of different timescales and through a complex set of mechanisms (Altieri et al., 2012; Krey et al., 2020; Belmain et al., 2022; Han et al., 2022). In this light, it is interesting that farmers observed less stunting and wilting in bean crops in fields that differed in terms of organic carbon, soil aggregation, soil available P, and recent applications of inorganic N fertilizer. These results for smallholder management contexts are consistent with more controlled experimental and observational studies suggesting for example that organic management can have positive impacts on plant defense (Krey et al., 2020), or via more diverse soil biological communities that can suppress pests such as pythium root rot (Larkin, 2015). It has been observed that soils with higher amounts of organic matter content have more water-stable aggregates (Gachene, 2018), and this was also the case in our study. Stable aggregates provide a habitat for a larger and more diverse microbial population which can create competitive and/or antagonistic environments between microorganisms resulting in disease suppression (Leon et al., 2006; Kevan and Shipp, 2011; Larkin, 2015). Meanwhile, the contrast between inorganic N application and soil available P in their opposite associations with stunting and wilting observations is also consistent with previous research showing that soil fertility influences the outcomes of SPDs. Soil-borne insect pests such as bean fly (Ophiomyia spp.) and soil-borne diseases such as root rot from Pythium spp. and Fusarium spp. are reduced in intensity when crops are grown in fertile soils and crops are less vulnerable, with phosphorus sufficiency being a key aspect (Hillocks et al., 2006; Buruchara et al., 2010; Huber et al., 2011; Samago et al., 2018). Letourneau (1994) also reported phosphorus deficiency resulted in an increase in population densities of all species of bean fly; however, increased total N increased the population of O. spencerella. The observation that high levels of available N predispose crops to pest damage is thought to be more generally true (Han et al., 2022).

In linking soil properties and management to stunting/wilting, our study is exploratory rather than definitive since this is an observational study that does not completely control for location, e.g., with paired plots of differing fertility and soil organic matter (SOM) status at a number of different locations. However, given the potential, and plausible links between organic matter, soil fertility, and reduced incidence of soil-borne pests and disease, the existing management of farmers raise concerns regarding sustainable intensification of bean cropping in these systems. For example, although farmers clearly used fertilizers to maintain the productivity of their maize/bean intercrops, few farmers used manure or compost in bean production fields as they lacked manure or materials for composting, and tended to use such materials in fields closer to home, often at high rates (>10 Mg ha⁻¹), which suggests the possibility of allocating some of these organic inputs to more remote maize and bean fields. The preference for enriching home fields vs. far fields in smallholder systems, and not applying fertilizer or other inputs onto beans in years where these are grown as a sole crop, has been documented in other systems (Masvaya et al., 2010; Saimon et al., 2017). In addition, although fields were in general balanced for P inputs vs. exports, available P was low across all sites and may be limiting to bean production. We note here that even though a few soil sites had strongly acidic soils which has been thought to lower the efficiency of the Olsen P extraction we used, a number of studies have shown that Olsen P performs reasonably well in acidic soils and was thus well-suited to our study which covered a wide range of soil pH from neutral to acidic (Farina and Channon, 1979; Fixen and Grove, 1990). Growing common bean with insufficient P can be detrimental to grain yield and degrade soil fertility (Araújo et al., 2000; Samago et al., 2018; Meya et al., 2020) as P contributes to effective N fixation (Hernández et al., 2009) and resistance to SPDs as discussed above (Huber et al., 2011). One important observation from our study is that regardless of whether farmers owned or did not own fields, they prioritized soil health and fertility improvement less than they did crop productivity. Other studies have argued that neglecting soils is especially prevalent for farmers who do not have long-term farm tenure and are less committed to investment in soil improvement (Williams, 1999). Conversely, land ownership can potentially lead to management practices that increase SOM affecting soil health (Mganga et al., 2016). However, in our study, few farmers were using manure irrespective of owning the field or not, likely due to transport costs, lack of skills in manure processing and composting, and relatively few cattle in the area. They were however applying more fertilizer to owned plots in order to plant maize every season, even if the balances in these plots were negative on average, due to high levels of N export by maize.

Crop residue removal can also contribute to the loss of SOM (Mganga et al., 2016) which could have beneficial effects on the management of some pests and diseases (Leon et al., 2006; Janvier et al., 2007). Farmers in the area generally use crop residue as livestock feed, and according to our survey, even farmers without livestock sold residue to those with animals. Beyond distance to fields and knowledge of manure management, our results likely do not capture all the factors that determine farmers' use of manure and crop residues, such as labor constraints and the monetary gain from selling residues; part of increasing the recycling of organic materials to soils would be to better understand and address these factors with evidence regarding soil health benefits of recycling residues (Adimassu et al., 2016; Mponela et al., 2016). Crop harvests, residue removal and lack of any organic inputs through animal manure or green mulches suggest that soil structure and organic matter content have likely been degraded in the majority of farmer fields in the area over decades of using such farming practices (Rurangwa et al., 2018), resulting in the low levels of active C observed in this study in many fields. Reversing these trends will likely require investment in better understanding residue use decisions and raising awareness of farmers and training that aims at restoring soil health such as maintaining agricultural residues, application of organic fertilizers, crop rotation, nutrient and carbon recycling (Bunning and Jiménez, 2003; Martinez-Salgado et al., 2010). Though challenging for many farmers, these practices can augment the soil's organic carbon pool, a key indicator of soil quality linked to soil aggregation, available water holding capacity and reduced erodibility of soil (Lal, 2006). Increased SOM will support an increased abundance of microorganisms that are necessary to sustain many soil functions such as the decomposition of organic matter (Janvier et al., 2007), maintenance of soil structure (Puget et al., 2000) and suppression of above and below-ground pests, parasites and diseases (Leon et al., 2006; Janvier et al., 2007; Altieri et al., 2012).