Introduction

Yield and quality of grassland may be altered in an alley-cropping agroforestry system compared to a treeless cropping system. Microclimate can be modified in a zone next to the tree strip that is neither static nor fixed but depends on the agroforestry design (e.g., orientation, planting density, tree age), time of the year and the climatic variable. For example, wind speed can be reduced in a zone of more than 20 m, whereas air temperature is only changed a few meters from the tree line (Rivest et al. 2022; Swieter et al 2022). This may result in changed yields, nutritional quality, or species composition of grassland in this zone (Grass et al. 2020; Sutterlütti et al. 2023a). Whether the effect of tree strips on grassland yield and quality is positive, negative, or neutral depends on various factors such as agroforestry design (e.g. orientation, planting density, tree age) or weather conditions (rainfall, temperature, extreme weather events) (e.g., Sutterlütti et al. 2023b).

We compared yields and crude protein contents of a grassland in an alley-cropping agroforestry system in northern Germany with those of a treeless control system for five consecutive years. We hypothesized reduced yield and enhanced crude protein contents of grassland in the immediate vicinity to the tree strips.

Material and methods

The study was conducted on a short-rotation alley cropping system, established in 2008 in northern Germany near Mariensee (52°33′51″N 9°27′50″E). The alley cropping system consisted of three 135 × 11 m rows of the willow clone ‘Tora’ (Salix schwerinii x S. viminalis) in a precise north–south orientation. Tree density was 10,500 trees per hectare. Willows were harvested in March 2016 and March 2021. Tree strips were separated by 48 m wide alleys of grassland. A 2 ha treeless reference grassland site bordered directly on the agroforestry site. Both control and agroforestry grassland were dominated by Poa pratensis, Agropyron repens, Alopecerus pratensis and Holcus lanatus. Both grasslands were managed site-specific with one yearly cut for fodder production in late spring and a second low-yield mulch cut in October. Between March and April, both control and agroforestry grassland were hauled to level mole hills. Watery pig and cattle slurries were applied each February (20 m3/ha; total N 1.5–2 kg/m3). Tree strips were not fertilized. The experimental site is characterized by two distinct Histosol soil types. The soil type in the northern part is a Dystric Folic Histosol (DFH) with an organic layer of about 40–50 cm. In the southern part a transition soil is found which can be described as a Dystric Folic Histosol-Gley (DFHG). For each of these soil types, grassland yield and quality were determined with three replicates, both in the alleys and in the treeless reference site. Yield and quality were determined at short distance from the tree strip edge (1 and 4 m) where effects were expected based on literature and own research, at intermediate (7 m) distance as well as in the alley center (24 m). In the control, yield and quality were measured at the corresponding distances from the field edge. The sampled field edge with DFHG soil was directly adjacent to a seldom-travelled grass pathway, a country road with a ditch and some roadside trees (height of about 9 m). The 1 m point of the grassland control transect was located at approximately 15 m from the roadside trees so that no influence of the trees on the grassland sample points was to be expected. The sampled field edge of the control field with DFH soil was also adjacent to a seldom-travelled grass pathway and a drainage channel. There were no trees or shrubs here. Annually yield surveys were carried out between 28 May and 15 June from 2017 to 2021 by cutting manually an area of 0.25 m2 at 3 cm stubble height. Dry matter yield of each sample was determined after drying at 105 °C for 24 h. As a measure of the forage value of the grassland, crude protein (CP) contents were analyzed. Therefore, the plant materials were oven-dried, ground (< 1 mm) and scanned, utilizing a near-infrared reflectance spectrophotometer (NIRS) (Foss 5000, Foss GmbH, Hamburg, Germany). Estimations were carried out on the basis of a robust, plant-specific NIRS calibration. The climate at the study site is temperate with a long term (1981–2010) mean annual air temperature of 9.7 °C and a mean annual precipitation of 657 mm. Annual precipitation totals were approximately equal to the long-term average in 2017, 2019, and 2021, but drier in 2018 and 2020 at 418 and 530 mm, respectively.

Data analyses were performed with RStudio (R Core Team 2022). To analyse grassland yield and quality data, generalised linear mixed effect models were fitted with distance from the tree strip/field edge (1, 4, 7 and 24 m), the experimental year (2017–2021), the cropping system (alley-cropping, open grassland) as well as the soil type (DFH, DFHG) as fixed effects and the plot ID as random effect using the package glmmTMB (Brooks et al. 2017). After automated model selection (Bartoń 2023), final models were selected based on the lowest Akaike information criterion (AIC). Subsequent analyses of variance (ANOVA type II) were followed by post-hoc comparisons of means using the emmeans package (Lenth et al. 2023). All data were square root transformed to achieve homogeneity of variance. Significance level for analysis was set at p < 0.05.

Results

Grassland dry matter yield was significantly determined by the factors distance into the field, year and soil type, but not by the cropping system. Moreover, yield was explained by the significant interaction of the variables year*cropping system, distance*cropping system, year*soil type, and cropping system*year*soil type. In both cropping systems, dry matter yields at 4, 7 and 24 m distance did not differ significantly amongst each other and were significantly higher than those at 1 m distance (Fig. 1A). Not considering involvement in interactions, dry matter yields were significantly lowest in 2018 and 2020 as well as significantly lower on the soil type DFHG. Overall, yields were lowest, though not always significantly, in the reference in 2018 and 2020 on both soil types as well as in the agroforestry in 2020 on both soil types.

Fig. 1
figure 1

A Estimated mean dry matter yield (Mg ha−1) and B estimated mean crude protein content (% dry matter) of the grassland over all harvest years (2017–2021) and both soil types (DFH, DFHG). Error bars are the confidence intervals of the candidate model with filled squares as predicted mean values. Lowercase letters in Fig. 1 show significant differences (p < 0.05) between dry matter yield values

Full size image

CP content was significantly affected by the factors distance, year and soil type, but not by the cropping system. Moreover, CP content was explained by the significant interaction of the variables year*cropping system, distance*cropping system as well as distance*year. Not considering interactions, the CP content was significantly highest in 2017–2019, significantly higher on DFH soil as well as significantly higher at 1 m than at 4 m. The interaction distance * cropping systems showed highest, though not significantly, CP contents in the agroforestry at 1 m (Fig. 1B).

Discussion

Yields of grassland and agricultural crops at field edges are usually reduced ('t Mannetje 2000), which is why field edges are generally excluded from sampling or excluded from yield maps (Vega et al. 2019). The reasons for lower yields at field edges are manifold and can relate to deviating management measures, such as fertilizer or crop protection applications, soil compaction from field traffic, or field margin structure (Liu et al. 2023). In studying the effect of different field margin structures on winter wheat yield, Raatz et al. (2019) found 95% of field centre yields at a distance of about 17 m from hedgerows or forest edges, at 11 m for a pothole, and about 7 m for an agricultural road or field-to-field border. In grassland, a different species composition is possible in the field edge (Grass et al. 2020, Nuñez et al. 2022). Thus, to study the effect of tree rows on grassland or crop yield at different distances from these rows, identical distances should be sampled in a treeless control. Many agroforestry studies have lacked this comparison and have compared yields adjacent to tree rows with those of field centres (e.g., Grass et al. 2020; Sutterlütti et al. 2023a). In our 5-year study, yields at 1 m from the field edge were reduced in both cropping systems (agroforestry vs. treeless control) and did not differ from each other. This was unexpected; we assumed a larger effect of tree rows due to altered microclimate as well as competition for resources such as light, water, nutrients, and space (Sutterlütti et al. 2023b; Swieter et al. 2022). In their work, Sutterlütti et al. (2023b) explained reduced grassland yields adjacent to tree rows in an alley cropping system by reduced light levels, whereas air temperature and relative humidity had a minor effect in their study. Grassland yields of a Canadian agroforestry system were similar to those of the control at 5 m from the tree strips (Rivest et al. 2022), which is in line with our findings at 4 m. The typical field edge structures at our control grassland site, i.e. grass pathway, road, ditch and roadside trees, may have influenced yield and quality of the grassland, although the type of influence is different than in the case of agroforestry tree rows. By tendency, the highest CP levels were found at the 1 m distance of the agroforestry variant. This is in line with the findings of Sutterlütti et al. (2023b), who also found elevated CP levels of grassland adjacent to a row of willow trees. CP levels of grassland are mostly enhanced under shaded conditions; for example Pang et al. (2019) and Lin et al. (2001) provide details on this topic.

Data availablity

The datasets generated during the current study are available from the corresponding author on reasonable request.