Agroforestry in northeastern Tanzania is practiced on Mount Kilimanjaro, in South Pare, and in the West and East Usambara Mountains, each occupying an agricultural area of approximately 8,000 km² (Figure 1; Burgess et al., 2007; Heckmann, 2011; Zech et al., 2014) with elevations ranging from 800 to 2,000 m asl. The climate is humid and monsoonal. Annual rainfall has a bimodal distribution with the main rainy season occurring between March and June (locally called Masika) and a shorter rainy season from October to December (Vuli). Each mountain range has its own unique indigenous agroforestry system (Akinnifesi et al., 2008; Reetsch et al., 2020a,b). These mountain ranges are referred to as ‘Kihamba’ or ‘Chagga homegardens’ on the southern slopes of Mount Kilimanjaro (Hemp and Hemp, 2008; Banzi and Kalisa, 2021), ‘Ginger agroforestry’ in South Pare (Ndaki, 2014; Mmbando, 2015), ‘Miraba’ in West Usambara (Msita, 2013), and ‘Mixed spices agroforestry’ in East Usambara (Hall et al., 2011; Patel et al., 2022).

The contribution of native and non-native tree species to ES differed among the studied areas (as shown in Tables 3, 5 and Figure 5). Native tree species are perceived as important for food and fodder very commonly in Kihamba (80% of native tree species) and Ginger agroforestry (75%). Shade also was an important service of native trees in those systems (70%, as compared to ≤20% for non-native species). Non-native trees are also used for food or fodder but much less for shade. In the Usambara, no native trees were encountered. Non-native trees were mostly valued as important for fuel and timber in Miraba and food (clove and cinnamon; data not shown) in Mixed Spices agroforestry (Figure 5).

When split according to species (Tables 3, 5; Figure 6; Supplementary Figure S3), it becomes evident that farmers have different requirements in different systems and use different trees to meet them. Moreover, a tree can have a different function in different agroforestry systems. In Kihamba, the largest share of trees was reported to be planted for food and fodder, and they belonged to a wide range of species (with Margaritaria and Rauvolfia being the main native species, and avocado as an important non-native). The second largest group was planted for shade, most of them being natives (Albizia, Cordia, Croton, and a variety of other species). Albizia is also important for fuel in Kihamba, and Grevillea is found to be an important non-native used for fuel and timber in the system. In Ginger agroforestry, trees were mainly planted for food and fodder (with fruit trees having the largest share) and shade (Albizia and a range of other native species). Few trees were encountered in Miraba, and fuel and timber were the most sought-after ES, with large shares for Grevillea, Acacia, and pine. Fruit trees are relatively rare (loquat, apple, and mango). Grevillea and pine are used for shade although farmers in the West Usambara use the term ‘shade’ also to denote soil and water conservation.

A PERMANOVA (Table 6) and NMDS ordination plot of Bray–Curtis community dissimilarities (Figure 7) confirmed that there is a significant difference between the identified tree species in the studied systems and the smallholder farmers reported most important ES (p < 0.001) across the study areas. This implies that the identified tree species have a most significant influence on the smallholder farmers who reported multiple ES (food/fodder, fuelwood, timber, and shade) at p of <0.05 across the studied agroforestry systems.

Moreover, our results generated by multipatt command from indicator species analysis (Supplementary Table S6) show statistically significant native tree species abundance (p < 0.04) for Maragaritaria discoidea and (p < 0.02) for Albizia schimperiana associated with Kihamba. These tree species have high relative abundance in the provision of food/fodder and shade. Locally native tree species Mguthuru (p < 0.03) and Newtonia buchananii (p < 0.05) were found statistically significant associated with Ginger agroforestry (Supplementary Table S7).

The agroforestry systems studied, i.e., Kihamba, Ginger agroforestry, and Miraba and Mixed spices agroforestry, are unique to Tanzania and East Africa (O'kting'ati and Mongi, 1986; Rugalema et al., 1994; Hemp and Hemp, 2008; Namwata et al., 2012; Kinyili et al., 2019). Although all studied systems are multi-layered with three or four vertical layers, our study shows that they have notable differences in their salient features mainly because of the unique climate, landscape setting, soils, historical background, habitat, and species adaptation that exists in this region (Table 1; Figures 3, 4; Namwata et al., 2012). Therefore, understanding the salient features of these systems including arrangements and patterns in space and the composition of their components will be of paramount importance in conserving these important agricultural heritage systems (cf. Charles, 2015; Reetsch et al., 2020a,b).

Kihamba homegardens have existed for over 800 years and most closely mimic a tropical montane forest, often containing mature tree species with a canopy layer height of more than 40 m and a large variety of native and non-native species (Figure 3; Tables 2, 4). This layout offers optimal growing conditions for coffee and banana on the volcanic soils of Kilimanjaro, while an important integration of cattle in the homegardens keeps soil fertility up to par for those demanding crops. Nevertheless, a crash in coffee prices has induced a shift in tree species toward other cash crops, notably avocado. The Ginger agroforestry in the South Pare mountains has a cultural link to the Kihamba on Kilimanjaro (Kitalyi et al., 2013; Ndaki, 2014), but as coffee and banana income declined even faster on the poorer, Precambrian soils, a boom of pests and coffee diseases motivated farmers to switch to growing ginger (70% of production in Tanzania) and sugarcane (Ndaki, 2014). As ginger is a root crop requiring a dappled shade, farmers kept the shade trees that are also common in Kihamba, but with a fewer dense canopy lowering light and root competition (Table 4; Figure 3). The introduction of ginger, hence, escalated the deforestation of the native tree species in the Pare mountains (Ndaki, 2014; Mmasa and Mhagama, 2017), while the lower nutrient requirements also contributed to a reduction in heads of cattle and fodder trees. This concurs with the increasing importance of non-native trees, mainly fruits, in the overstory layer (Figure 5; Table 3), consistent with earlier findings of Nath et al. (2016).

In Miraba, the culture of maize and vegetables requires ample sunlight, so farmers only plant scattered trees among the Miraba hedge lines (Figure 4). There was no culture of traditional homegardens as in Kilimanjaro and Pare, yet the use of Miraba (hedges) around fields was traditionally practiced mainly by women (Msita et al., 2010, 2012). The use of trees native to the area in farming was not part of the tradition. Historically, the Usambara mountains were covered by native forest tree species, such as Albizia gummifera, Prunus africana, Catha edulis, Ocotea usambarensis, Podocarpus usambarensis, Parinari excelsa, and Milicia excelsa (Msuya et al., 2008). However, much of the forest tree species were logged for timber before the logging ban of 1984 (MNRT, 2001; FAO, 2005). Increasing population densities expanded community activities, i.e., cultivation of maize, beans, and Irish potatoes on slopes and vegetables on valley bottoms. The intensification in land use led to severe soil erosion on the mountain slopes and flash floods in the valley bottoms (Haruyama and Toko, 2005; Msuya et al., 2008). Due to these challenges, Miraba was promoted as a soil conservation measure. Other interventions in the landscape, for example, the Gesellschaft für Technische Zusammenarbeit (GTZ) project on Soil Erosion Control and Agroforestry (SECAP) in collaboration with other institutions including the Tanzania Forestry Research Institute (TAFORI), introduced non-native tree species, such as Grevillea, pine, and eucalyptus for curbing soil erosion and to reduce logging (Johansson, 2001; Msuya et al., 2008). Hence, these species remain important in the landscape (Table 3; Figure 6).

East Usambara receives very high amounts of rain from the Indian Ocean. Combined with the Precambrian, easily leachable soils, it makes it difficult to get good yields of arable crops, banana, or coffee. As the region has cultural ties to Zanzibar, Madagascar, and India, a system of spice crops that thrive in high humidity and on well-draining soils has been practiced here for over 50 years (Figure 3; Hall et al., 2011). This type of agroforestry starts with the thinning of canopy trees to create 50% shade and the complete clearance of the lower strata of a once natural forest (Reyes et al., 2005; Hall et al., 2011). Those authors noticed an absence of young native tree species in two-thirds of the active agroforest sites, questioning the ability of the Mixed spices agroforestry to contribute to the conservation goals of the East Usambara Mountains (Reyes et al., 2010; Hall et al., 2011). During our field campaign in 2021, native trees were absent from the canopy. Mature clove trees are most productive in full sunlight, and their conical–cylindrical shape allows ample sunlight for the cinnamon trees below. Land scarcity may push to a further clearing of the original canopy to prevent competition for light and space with clove and cinnamon trees. Black pepper and cardamom require partial to full shade underneath the trees, and pepper uses the clove trees for support.

Consistent with the farming strategies described above, farmers have different ES requirements for trees in the different systems (Tables 3, 5), providing a more diversified image as compared to earlier studies stressing the importance of on-farm tree resources for the provision of food, fodder, shade, timber, and fuel (Munishi et al., 2008; Charles, 2015; Wagner et al., 2019).

Moreover, our analysis shows that tree species are used in different ways in the different agroforestry systems (Figures 6, 7).

The identification of dominant native tree species, such as Albizia schimperiana, Maragaritaria discoidea, Rauvolfia caffra, and Cordia africana, in the studied agroforestry systems holds significant implications for ecosystem services provisioning. These trees play a pivotal role in the sustainability and multifunctionality of the agroecosystems in northeastern Tanzania. Notably, they serve as crucial shade providers for coffee cultivation, contribute to fodder production, serve as a source of fuelwood, and in some instances, are employed for medicinal purposes. As such, native species are mainly valued in systems requiring shading of coffee, banana, or ginger, and in systems with an important cattle component, notably in Kihamba (Banzi and Kalisa, 2021). In Ginger agroforestry, Albizia schimperiana remains as a shade tree but fodder trees are being replaced by fruits (Figure 6).

The findings of this study align with prior research in the same study area, reinforcing the importance of Albizia schimperiana as a primary choice for shading coffee in both smallholder farms and large-scale commercial coffee plantations (Hundera, 2016). Findings in our study revealed that native species in the Kihamba remain important for communities in accessing ES, such as food, fodder, fuel, and timber, and in providing shade for the production of coffee, banana, firewood, roots, and tuber crops as well as vegetables (Figure 5; Table 5). In this system, farmers have accumulated wide indigenous knowledge and use a wide range of trees and shrubs (Figure 6; Akinnifesi et al., 2008; Hemp and Hemp, 2008; Reetsch et al., 2020a,b).

However, we demonstrated that the proportions of non-native tree species are becoming competitive with native tree species in the studied areas, and native species are not or no longer used in farming in the Usambara Mountains. For example, Grevillea robusta, Persea americana, and Eucalyptus camadulensis have been introduced for the timber market and have replaced part of the native trees used for fuel and timber in Kihamba and Ginger agroforestry (Table 3; Figures 5, 6). Fruit trees are replacing native food and fodder trees most notably not only in Ginger agroforestry but also in Kihamba. In systems with no shade requirements (Miraba) or where native tree species would compete with tree crops (Mixed spices agroforestry), native trees are now absent from fields and homegardens.

The tree component of agroforestry systems is important not only for provisioning services but also for supporting, regulating, and cultural ES important for conservation and ecosystem resilience (Soini, 2005; Graham et al., 2022). Mature, native trees in Kihamba have been reported as important for biodiversity conservation and carbon sequestration (Fernandes et al., 1984; Gupta et al., 2009). The taller canopy provides a diverse range of habitats and niches, supporting a greater variety of flora and fauna, contributing to overall biodiversity in the agroforestry system (Hemp, 2006). The Kihamba native tree layer has, moreover, been shown efficient in controlling landslides, in reducing soil erosion, in improving soil fertility, and in protecting sources of water for local and downstream users (Kitalyi and Soini, 2004; Hemp and Hemp, 2008; Mbeyale, 2010; Santoro et al., 2020; Reetsch et al., 2020a,b; Banzi and Kalisa, 2021; Mbeyale and Mcharo, 2022). In the North Pare Mountains, agroforestry tree species help to improve the resilience of smallholder farmers against environmental extremes by modifying temperatures (Charles, 2015). The absence of native tree species has, moreover, changed the outlook of the landscapes in terms of their pristineness, cultural history, and land use/cover arrangements. Restoration efforts and re-introduction of native species have, thus, been proposed to improve the resilience of the studied systems and are advocated as an avenue to minimize conflicts and encroachment into the protected areas (Johansson, 2001; Kueffer et al., 2013; López et al., 2017).

Over the past 100 years, farming systems in the northeastern Mountains of Tanzania have undergone several transformations due to colonial and post-colonial policies, land scarcity, migration of younger generations to urban areas, crop pests and diseases, and collapse in coffee prices (Chuhila, 2016; von Hellermann, 2016). The results of our study corroborate the importance of livelihood strategies on the tree component of agroforestry systems (Figures 3–7), corroborating the statement that these challenges have led the smallholder farmers in the area to diversify their sources of income to accommodate external changes and market dynamics (Namwata et al., 2012). The majority of smallholder farmers have adopted the introduced non-native tree species, sometimes for conservation value but more so for their economic benefits (von Hellermann, 2016; Figures 5, 6). Hence, differences in the context of smallholder farming conditions and ES requirements, as evidenced in our study, should be taken into consideration for restoration efforts to be successful.

von Hellermann (2016) stressed the importance of an increased sale of coffee for agroforestry during the 1940s. Our study corroborates that shade ES required for coffee farming promotes the use native tree species (Figure 5) and supports the hypothesis that a collapse in coffee prices since has led to a gradual abandonment of the coffee crop and diversification of crop production in Kilimanjaro and Pare (Ndaki, 2014), leading to a deforestation of the native tree species (Ndaki, 2014; Mmasa and Mhagama, 2017; Table 4). If native trees are to be restored in this region, additional research and supporting measures are needed to help farmers build alternative value chains for products that can benefit from the ES from native species, such as the sale of milk or honey from (stingless) bees (Eersels, 2022; Tersago, 2022).

In the East Usambara mountains, protecting habitat for endemic species is one of the most important conservation objectives (Burgess et al., 2007; Hall et al., 2011). In Mixed spices agroforestry, the strata of a once natural forest (Reyes et al., 2005; Hall et al., 2011) have now completely disappeared (Table 4). Several authors, therefore, question the contribution of Mixed spices agroforestry to conservation goals (Reyes et al., 2010; Hall et al., 2011). Although such a tree-covered agricultural system may provide additional ecological services compared to sun-grown agriculture, a lower compositional and structural diversity will affect the ES not related to food production as compared to natural forests. Furthermore, a more profitable cardamom market could be beneficial to local farmers, which may encourage agroforestry establishment in currently deforested areas but could also lead to the expansion of cultivation into protected areas (Reyes et al., 2010). Some previous studies suggest that sustainable cultivation of spice is possible (Kumar and Nair, 2004; Reyes et al., 2006; Swallow et al., 2006) and that some farmers are already adopting ecologically sound intensification practices in homegardens (Reyes, 2008; Reyes et al., 2010). Therefore, any efforts to encourage integrated Mixed spices agroforestry with other native agroforestry tree species should be explored. Nevertheless, as all farms in our study do not have productive ES requirements for trees other than clove and cinnamon (Tables 3, 5), these efforts will not be straightforward to realize for farmers from a livelihood perspective without flanking measures. The protection of native vegetation in forest reserves, therefore, also remains an urgent priority.

The role of policy and knowledge bias in agroforestry tree composition has been highlighted by several authors. Worboys (1979) and Sheridan (2001) mentioned the role of policy and mass promotion by government regimes with a motive to produce timber for export and also restore previously cleared forests. Interventions to control erosion and reduce logging introduced non-native species, such as Grevillea, pine, and eucalyptus, as these are well studied in the international literature on soil and water conservation, as compared to species native to the Usambara (Johansson, 2001; Msuya et al., 2008). Policies to restore the native tree cover can, therefore, only be successful if underpinned by a better knowledge of local species and their potential to be aligned with the diverse ES needs of local communities (Figures 5–7). Kihamba agroforestry can serve as an inspiration as it shows a kind of resilience in terms of available native tree species that are the remnants of the forest tree species (Table 5; Figure 6) and has been shown very efficient in the provisioning of ES for conservation purposes (Hemp and Hemp, 2008; Reetsch et al., 2020a,b). The fact that Kihamba farmers still use native tree species for ES that are also required in systems without native species (Figures 6, 7) indicates potential for the exchange of indigenous knowledge between distant communities as well as for driving scientific research toward the potential of these trees.