Introduction

Climate change, biodiversity loss, and food security have been identified as three major crises currently facing the global human population (McCouch et al. 2013, 2020; Nelson et al. 2009). These challenges are interlinked: modern intensive farming is known to be a major source of greenhouse gasses (GHGs; IPCC 2020), as well as a major driver of ongoing declines in biodiversity (Bicknell et al. 2023; Jaureguiberry et al. 2022) through habitat degradation, mechanization, and widespread pesticide use (Critchley et al. 2006; Robinson and Sutherland 2002; Storkey et al. 2012).

Ideal solutions to these crises would be win-win-wins, i.e., simultaneously tackling all three challenges. Such solutions would enable a move away from debates based on a zero-sum game, where the route to tackling farmland biodiversity loss is seen as being either taking land out of cultivation (i.e., land sparing) or accepting assumed reductions in the level of productivity from more “regenerative” agricultural approaches. Win-win-wins would allow increased biodiversity within cropped land, reduce GHG emissions, and maintain yields (notwithstanding the important challenge of tackling food waste). Such win-win-wins might also be more acceptable to all stakeholders and would not face the difficulty encountered by some approaches of being seen as directly inimical to the basic purpose of farming, i.e., to produce crops. Such perceptions can be a substantial constraint on scaling up, and if potential solutions cannot get past such hurdles and achieve large scale uptake then they may not be genuinely viable.

Crop mixtures could be a win-win-win option with the potential for application at scale. The definition of crop mixtures, and the mechanisms underpinning them, have been reviewed in detail elsewhere (see for example Brooker et al. 2015, 2021; George et al. 2022; Pakeman et al. 2020). In brief, crop mixtures are two or more crop species growing in the same location and sufficiently closely integrated in time and space such that there is substantial interaction between the component crops. Through mechanisms such as within-crop N fixation and pest, disease, and weed suppression, crop mixtures can be grown with reduced inputs including fertilisers, herbicides, and pesticides (see Brooker et al. 2015, 2016; Martin-Guay et al. 2018). Reducing inputs and increasing crop above- and below-ground structural complexity in turn has benefits for farmland biodiversity. Similar - though smaller - benefits can also be achieved through cultivar mixtures (see, for example, Newton et al. (1997) and Newton and Guy (2009)). Such positive effects of both crop and cultivar mixtures concur with ecological studies which demonstrate that across a wide range of ecosystems increasing biodiversity enhances system functions including productivity, resilience, and C storage (Brooker 2016; Brooker et al. 2021; Cardinale et al. 2012).

Importantly, many crop mixture studies are indicating that, as well as potential benefits for biodiversity and GHG emissions, crop mixtures also maintain yields compared to conventional cropping, and in many cases enhance yields. For example, a recent pan-European study focussing on arable crops found a crop performance ratio (CPR) of 1.28, indicating a roughly 30% average yield gain across sites (Brooker et al. 2024). This scale of benefit concurs with the average Net Effect Ratio value of 1.28 observed by Li et al. (2023), and is similar to the average LER value found by Martin-Guay et al. (2018) (although, as discussed by Li et al. (2023), these metrics express different aspects of crop mixture performance). We might expect such levels of productivity to address concerns about possible yield reduction. However, and despite such data concerning yield, widespread uptake remains a key challenge for crop mixtures, particularly in modern, mechanised, and “intensive” or “conventional” agricultural settings.

Remaining barriers to uptake include knowing how best to tailor crop mixtures to local environmental conditions (Brooker et al. 2024). An important aspect of tailoring crop mixtures is understanding the extent to which crop mixture yields are consistent or whether they vary with changing environmental conditions, necessitating changes in the type of crop or agronomic practice depending upon the local environment. Climate is a critical aspect of the abiotic context within which the mixtures grow. The processes operating within crop mixtures include direct and indirect beneficial and competitive interactions between the component crops, and ecological theory has demonstrated that the balance of such interactions can vary according to the environmental conditions (for reviews see, for example, Brooker et al. 2008, 2015). However, the results of previous studies of crop mixtures are equivocal on the influence of climate on yield. The meta-analysis by Martin-Guay et al. (2018) showed no effect on LER of water availability (measured as the occurrence of irrigation and the aridity index in non-irrigated intercrops). In contrast, Brooker et al. (2024) found a significant positive effect of mean daily rainfall on CPR. Differences in scale and scope may contribute to these between-study differences in results. Martin-Guay et al. worked at a global scale, and “spanned the globe” from arid to humid environments, also including a wide range of intercrops in their study. In contrast Brooker et al. (2024) focussed on intercrops within Europe, and so looked at a narrower range of environmental conditions and crop types. It may be that different relationships are again evident if we focus down still further to the national or regional level, often the level at which farm advisory (‘extension’) services operate.

A second aspect which causes concern amongst growers, and which again may limit the uptake of crop mixtures, is ensuring a market for the product (Brooker et al. 2023). Part of the process of ensuring a market is to meet standards in terms of crop quality. For example, in Scotland the brewing and distilling industry is a major buyer of malting barley, and there are clear expectations of quality for grain which is sold into this premium market, and so it is important where possible to add to the growing body of evidence concerning grain chemical composition (one aspect of grain quality) and the impact on this of practices such as intercropping.

Finally, there is very considerable focus currently on agricultural soil carbon (C). This is in part due to concerns about farming’s greenhouse gas (GHG) emissions, to which soil degradation contributes, and in part because there is a hope that appropriate management can enhance C storage in agricultural soils. This provides benefits both to farmers – who might be able to realise the value of this stored C by selling carbon credits or might be able to get farm payments for increasing soil C – and to society in general by helping to mitigate the extent of climate change. If intercropping helps to reduce farm GHG emissions, either by reducing outputs or by enhancing soil C capture and storage, then this too may be an important element in driving the wider uptake of this type of cropping system.

In this study we aimed to address these various knowledge gaps with respect to crop mixtures, with the ultimate aim of supporting wider uptake of crop mixtures as a sustainable farming practice. We focussed in particular on understanding the response of crop mixture yields to climate and management drivers, and in the same crops explored responses in crop seed chemistry, and soil C and nutrients. Overall, and based on previous studies, we expected the CPR of crop mixtures to show crop production levels equal to or better than that delivered by monocultures, and to be positively related to daily rainfall. We expected “less intensive” management regimes, with fewer inputs and less severe soil disturbance, to show the highest CPR. With respect to the responses of soil nutrients, soil carbon, and crop seed chemistry, our null hypothesis was that these factors would not be influenced by crop mixtures.

Our study focusses on on-farm trials conducted in Scotland during 2020, 2021 and 2022. Data from some of these trials had been included in a previous analysis (Brooker et al. 2024) but the study presented here includes data from an additional year – 2022 - which was particularly hot and dry in many parts of Europe (MetOffice 2024). This enables us first to compare at different scales the responses of intercrops to factors such as climate (Scotland vs. Europe), and second to assess whether the benefits of intercrops were affected by a year which represented a climate extreme (for Scotland). This study also moves beyond the previous yield-focussed response of Brooker et al. (2024) to examine impacts on crop seed chemistry, and soil C and nutrients, which as noted above are related to key factors either promoting or limiting the wider uptake of intercrops.

Materials and methods

Field trial design

Between 2020 and 2022, 32 intercropping field trials of varying complexity were grown on commercial farms and research farms in Scotland. These were concentrated around the main arable crop growing region of the east coast of Scotland (see Suppl. Mat. A). Each trial included a minimum of three treatments, comprising the intercrop and each intercrop component in monoculture. Plot sizes ranged from < 1 to > 5 ha. Except for the research farm trials, there was no replication of treatments within a trial, with trials commonly being two areas sown with the monocrops and between these a strip of crop mixture.

The trials varied in their design, management treatments and crop combinations tested (see Suppl. Mat. A). To help with the aim of making crop mixtures relevant to farmers, the crop mixture grown was often determined by the farmer; some farmers had strong preferences and others were advised to grow mixtures that were expected to be relatively straightforward and reliable in the Scottish context, for example pea and barley. In addition, we did not specify a formal planting arrangement, but allowed the farmer to make the mixture to the best of their ability and using standard on-farm equipment. In practice this involved either pre-mixing seeds and sowing the mixture in a single pass of the drill, or sowing one component in one pass, and the second component in another pass of the same area (for example simply overlapping the edges two neighbouring patches of monoculture crop). This meant that in all cases the mixtures were a “fully mixed” intercrop (Brooker et al. 2015), but there was not tight control of the spatial arrangement of plants within the crop.

Yield and seed chemistry assessment

Mixed crops were not split prior to yield assessment: yield assessment at harvest was done at the whole crop level, i.e., as the total yield of the crop containing either one species (monocultures) or multiple species (mixtures). Yield data were provided by the farmers on standardised reporting templates. The metadata and data collected from the trials were collated and included sowing and harvest dates and final yield (seed or whole crop biomass) of each treatment. Metadata also included tillage type (ploughed/inversion or non-ploughed/non-inversion), fertiliser use (inorganic, organic or none), and herbicide use (yes or no), and whether or not the farming approach was organic (yes or no).

In 2021 and 2022 farmers also provided seed samples from the final harvested and bulked crop (see Suppl. Mat. Table A for a list of trials from which seed was provided). These samples were analysed for phosphorous, nitrogen and carbon (P, N and C) content. Sample extractable P content was analysed following 50% nitric acid tissue digest with reflux for 30 min; after cooling, the digest was analysed for elemental concentrations using inductively coupled plasma optical emission spectroscopy (ICP-OES). Percentage N and C content of the samples was determined by an automated Dumas combustion procedure (Pella and Colombo 1973) using a Flash 2000 Elemental Analyser, (Thermo Scientific). P concentrations were expressed in mg kg−1 and N and C content as a percentage of sample dry mass. Hereafter we refer to seed rather than grain chemistry or quality. Grain structure affects quality also, particularly malting performance, but this was not assessed. In addition, many trials included peas, field beans and linseed that are not generally termed grains.

Weed cover and soil chemistry assessment

In all years before harvest, but at a point when crops were approaching maturity (mid-July through to early August; see Suppl. Mat. Table A), measurements were made within the crops, and soil samples were taken for chemical analysis. In those cases where there were small trial plots at each site, one plot was taken as representing one replicate, and a single sampling quadrat was placed within each plot. In the more frequent case where a single large area of the mixture and the component monocultures was sown, for example a field sown with two areas of monocrop with an area in the middle where both component crops overlapped to give the mixture, each large area was sub-sampled generally three times, with a quadrat being placed at three locations spatially dispersed across the cropped area to provide this subsampling.

Sampling quadrats were 0.5 m x 0.5 m for cereal monocultures and 1 m x 1 m for non-cereal constituents or mixtures of cereals and non-cereal crops. To minimise edge effects, in small trial plots these were located roughly in the middle of the plot. In the larger areas of mixture or monocrop, quadrats were located at least 50 cm away from the nearest vehicle passage line. Within this quadrat we visually assessed the percentage cover of non-crop vascular plants, hereafter referred to as weed cover. We did not record weed cover at species level, but common species included Persicaria maculosa, Chenopodium album, Poa annua, Stellaria media, and Matricaria discoidea. Prior to analysis, weed cover values were averaged within treatments across each site; so, for a site with a 2-species mixture and two associated monocrops there would be three averaged weed cover values, one each for the two monocultures and one for the mixture.

In 2020 and 2021 we also took five soil samples within each sampling quadrat using a 3 cm diameter corer and to a depth of approximately 10 cm. Sampling points were positioned at the four corners of the quadrat with one additional sample being taken from the middle of the quadrat. These five samples were then bulked and stored at 4 °C before being subsampled for further chemical analysis.

Soil NH4-N and NO3-N were determined colourimetrically after extraction using a 1 M potassium chloride solution (10:1, solution: soil). Extractable P and total N and C content of the soil samples were determined following the same procedure as used for the seed samples. Soil pH was determined on the supernatant of a 3:1 water to soil mix (pH(H2O), and the supernatant in a 0.01 M CaCl matrix (pH(CaCl2) (McLean 1982; Sumner 1994). Extractable soil P, NH4-N and NO3-N levels are expressed in mg kg−1 and total N and C content as a percentage of sample dry mass.

Climate data

In addition to the direct measurements of crop responses, weather data for the growing season of each trial was sourced from the E-OBS ENSEMBLES daily gridded observational dataset for precipitation and temperature in Europe (Cornes et al. 2018). Data for daily maximum, mean and minimum temperatures, as well as daily precipitation, at a resolution of 0.1° (c. 11.1 km latitude) were used to calculate the following growing season indices for each trial: mean temperature (°C), growing degree days (> 5 °C), mean daily maximum temperature, highest maximum temperature, mean daily minimum temperature, lowest minimum temperature, mean daily rainfall (mm) and total rainfall.

Data analysis

For analysis of yield data, each trial represented a single replicate, with the research farm trials included in the analysis by using the mean values of each treatment. Yield data were also standardised before analysis. Although Land Equivalent Ratio (LER) is widely used to assess the yield benefits of intercropping (Khanal et al. 2021), LER was not seen as a suitable metric for our study because it does not account for varying ratios of the species in the intercrop, and calculation of LER necessitates the intercrop components be separated which was not possible for many of our trials because of a lack of labour or appropriate equipment. With these factors in mind, we selected Crop Performance Ratio (CPR; Harris et al. 1987) as our response metric. CPR allows for the ratio of components to vary and does not need separation of the intercrop. An additional benefit is that it measures the yield gain for a fixed area, making it easily interpretable.

For a two-species intercrop, CPR is expressed as:

$$\:CPR=\:\frac{{Y1}_{c}+\:{Y2}_{c}}{{Z1}_{c}{Y1}_{m}+\:{Z2}_{c}{Y2}_{m}}$$

where Y1c and Y2c are the yields of crops 1 and 2 when grown together, Z1c and Z2c are the proportions of crops 1 and 2 in the intercrop, and Y1m and Y2m are the yields of the two crops when grown separately as monocultures. Following Oyejola and Mead (1982), means of the monocultures and intercrops across each trial were used in calculating CPR.

For the purposes of analysing the weed cover data we created a similar metric to CPR, Weed Performance Ratio, and calculated it as:

$$\:WPR=\text{l}\text{n}\:\left(\frac{{WC}_{mix}}{{Z1}_{c}{WC1}_{m}+\:{Z2}_{c}{WC2}_{m}}\right)$$

Where WCmix is the weed cover in the mixture, WC1m and WC2m are the weed covers when the crops are grown as monocultures. The index was loge transformed to make its behaviour symmetrical for positive and negative effects; negative values indicate greater weed suppression by mixtures.

All analyses were carried out in R version 4.2.2 (R Core Team 2022) using linear mixed models from the lme4 package (Bates et al. 2015) with probability testing from lmerTest (Kuznetsova et al. 2017) and predicted means calculated in emmeans (Lenth 2022).

We did not obtain a full set of data from all sites in all years – for example for some of the sites for which we have grain and soil chemistry data we do not have yield data (see Suppl. Mat. Table A) and, overall, our dataset lacked power to explore the interactive effect of key factors of interest.

Table 1 Mixed effect model output from the analysis of the relationship between CPR and each separate weather variable
Full size table

The mean effect of intercropping was calculated using a model with no fixed effect but with a random term (CPR ~ 1 + (1|Site) as multiple trials were run at some sites. Separate categorical fixed models were tested for the effects of different factors on CPR, namely mixture complexity, cereal and legume crop species and management with Site as the sole random term. Separate quantitative models were run for overall yield, weather and weed cover with the same random term. For tests of seed chemistry, the random model was Site/Block as the research farm trials had multiple experimental blocks, and for tests of soil chemistry the random model was Site/Block/Plot as multiple samples were taken per plot; the block term is only relevant for the replicated experiments.

3. Results

Impacts of crop diversity and composition on yield

Average CPR across all studies was 1.19 (95% CI 1.04 to 1.35). This indicates that crop mixtures yielded roughly 20% more per unit area of land than the expected value based on the component monoculture crops. It is also notable that the lower limit of the 95% confidence interval is greater than 1, indicating that – at worst – crop mixtures performed as well as component monocrops, with the majority doing better than expected based on monocrop performance.

There was a strong positive effect of increasing the number of species in the intercrop from 2 to 3. The predicted mean CPR of two species mixtures was 1.16, whereas the mean CPR of three species mixtures was 1.53 (df = 31.1, t = 3.527, p = 0.001). CPR was also higher where overall yields were lower (df = 30.13, t = -4.21, p < 0.001), indicating that the benefits of crop mixtures were greatest in conditions which monocrops find more marginal.

Crop composition affected CPR. Compared to mixtures with no legumes, a combination of beans and peas in the mixture gave significantly greater CPR (df = 24.92, t = 2.496, p = 0.020), while having either beans or peas alone did not (Fig. 1a; full model results Suppl. Mat. B). With respect to cereals, compared to mixtures with barley, CPR was significantly increased by the presence of oats (df = 26.62, t = 5.114, p < 0.001), with a trend towards increased CPR due to the presence of wheat (df = 12.64, t = 1.849, p = 0.088), with no impact of the presence of rye (Fig. 1b; full model results Suppl. Mat. B). Compared to mixtures without linseed (Fig. 1c), the presence of linseed significantly increased CPR (df = 25.74, t = -3.68, p = 0.001).

Fig. 1
figure 1

Crop Performance Ratio (CPR) values for plots with different types of (a) legume composition (b) cereal composition, and (c) presence/absence of linseed. Values shown are model estimated (+/- 95% confidence intervals) means. Figures on the x-axis indicate the level of replication (number of trials) within each category

Full size image

Impacts of climate drivers on CPR

As well as an influence of composition, we found significant impacts on CPR of climate variables (Table 1), in particular negative effects of growing degree days (Fig. 2a), and negative effects of mean daily and total rainfall (Fig. 2b and c), indicating that CPR was greatest in cooler and drier conditions. However, there was also a trend toward a positive effect of highest temperature recorded. Changes in CPR can be driven by either an increase in the mixture yield, or a decrease in the monoculture yields. To explore this further we assessed the relationship between average monoculture yield and CPR; this showed a significantly negative relationship between CPR and average monoculture yield (df = 30.13, t = -4.213, p < 0.001), indicating that the yield gains from mixtures are greater when monoculture yields are lower.

Fig. 2
figure 2

Fitted relationship from the mixed effect model of CPR and (a) Growing Degree Days (GDD °C), (b) mean daily rainfall (mm) and (c) total growing season rainfall (mm). Dashed lines indicate the 95% confidence intervals of the relationship

Full size image

Impacts of management on CPR and weed cover

CPR did not differ between crops grown with standard inorganic fertiliser and either no fertiliser (df = 27.69, t = -1.61, p = 0.118) or organic N fertiliser (df = 15.41, t = 0.52, p = 0.611). There was no significant effect on CPR of organic vs. non-organic management (df = 32.46, t = -0.85, p = 0.401), but there was a trend towards ploughing having a positive effect on CPR compared to non-inversion tillage (df = 29.89, t = 1.93, p = 0.063). Ploughing (inversion tillage) can be important in suppressing weed growth, and herbicide use also had a significantly positive effect on CPR (df = 26.13, t = -3.32, p = 0.003). For full results of all tests of the impacts of management on CPR see Suppl. Mat. C.

Of all the factors tested, including the presence or absence of crop mixtures, the number of components in the crop mixture, cereal, or legume type, presence of linseed, ploughing, and herbicide or fertiliser addition, only ploughing had a significant impact on weed cover (df = 27.79, t = -2.10, p = 0.045), with mixtures having lower weed cover than expected on ploughed plots (WPR = -0.56) but greater weed cover than expected on non-ploughed plots (WPR = 0.47). For full results of all tests of the impacts of management weed cover see Suppl. Mat. D.

Seed chemistry

Analysis of seed chemistry data focussed on cereals grown either as monocultures or in two-species mixtures with legumes, and on legumes grown either as monocultures or in two-species mixtures with cereals.

For cereals, only wheat showed significant changes in seed chemistry between monocultures and mixtures. Wheat seed phosphorous content was not affected by the mixture treatment, but wheat seed nitrogen content (df = 73.67, t = -2.31, p = 0.023), and seed carbon content (df = 65.67, t = -2.14, p = 0.035), were both enhanced in mixtures with no significant impacts on C: N ratio (Fig. 3). For full results of cereal seed chemistry analyses see Suppl. Mat. E.

Fig. 3
figure 3

For cereals (barley, oats, wheat) grown with legumes, model estimated (+/- 95% confidence intervals) mean values of seed (a) nitrogen concentrations (N as a percentage of sample mass), (b) carbon concentrations (C as a percentage of sample mass), (c) P concentrations (mg kg−1) and, (d) seed C: N ratio. Significant within-species effects of growing with or without legumes shown as * p < 0.05. Figures on the x-axis indicate the level of replication (number of trials) within each category

Full size image

For legumes, only peas showed significant changes in seed chemistry between monocultures and mixtures. There were significant changes in pea seed nitrogen content (df = 55.02, t = 3.87, p < 0.001) and changes in phosphorous content (df = 54.41, t = 2.29, p = 0.026), with N and P content being lower in peas from mixtures. There were no significant effects on legume seed carbon content, resulting in a significantly higher C: N ratio in peas from mixtures (df = 56.10, t = -3.62, p < 0.001) (Fig. 4). For full results of seed chemistry analyses see Suppl. Mat. F.

Fig. 4
figure 4

For legumes (beans, lupins, peas) grown with cereals, model estimated (+/- 95% confidence intervals) mean values of seed (a) nitrogen concentrations (N as a percentage of sample mass), (b) carbon concentrations (C as a percentage of sample mass), (c) P concentrations (mg kg−1) and, (d) seed C: N ratio. Significant within-species effects of growing in mixtures or monocultures shown as * p < 0.05, *** p < 0.001. Figures on the x-axis indicate the level of replication (number of trials) within each category

Full size image

Soil chemistry

There were several significant differences between the soil chemistry of monoculture cereals, and that of other crops and mixture combinations (Fig. 5). NH4-N was enhanced in mixtures with cereals and linseed (df = 206.68, t = 2.016, p = 0.045) and three-component cereal-legume-linseed mixtures (df = 185.44, t = 2.206, p = 0.029). NO3-N was enhanced by the presence of legumes (df = 197.90, t = 4.78, p < 0.001), and enhanced in mixtures of cereals and legumes (df = 197.77, t = 2.04, p = 0.043), cereals and linseed (df = 203.77, t = 2.09, p = 0.038), and three-component cereal-legume-linseed mixtures (df = 200.12, t = 2.08, p = 0.039).

Fig. 5
figure 5

Response of soil chemistry to crop type including cereal (C), legume (L) and linseed (O = oilseed) monocultures, and mixtures of cereal + legume (CL), cereal + linseed (CO), and cereal + legume + linseed (CLO). Panels show estimated mean values (+/- 95% confidence intervals) for (a) extractable NH4-N (mg kg−1), (b) extractable NO3-N (mg kg−1), (c) total C (percentage of sample mass), (d) total N (percentage of sample mass), (e) soil C: N and (f) extractable P (mg kg−1). Letters indicate significant differences (P < 0.05) in soil chemistry between crop types. Figures on the x-axis indicate the level of replication (number of trials) within each category

Full size image

Total soil N (% dry mass) was reduced in mixtures of cereals and linseed (df = 210.64, t = -2.55, p = 0.012), and significantly reduced in three-component cereal-legume-linseed mixtures (df = 205.44, t = -3.89, p < 0.001). Similarly, soil C (% dry mass) was reduced in mixtures of cereals and linseed (df = 202.90, t = -1.99, p = 0.048), and in three-component cereal-legume-linseed mixtures (df = 198.61, t = -4.07, p < 0.001), and showed a trend towards a reduction in the presence of legumes (df = 195.99, t = -1.75, p < 0.082). Soil C: N showed a trend towards an increase in mixtures of cereals and linseed (df = 204.30, t = 1.84, p = 0.067) and P showed a trend towards a decrease in these same mixtures (df = 232.84, t = -1.86, p = 0.064). For full results of soil chemistry analyses see Suppl. Mat. G.

Discussion

Addressing the numerous challenges involved in combatting the climate-nature crisis while delivering food security will be aided by identifying situations where a particular land management approach delivers benefits relevant to all three critical issues (Brooker et al. 2024). Crop mixtures have the potential to deliver such a “win-win-win”, with crop yield and resilience potentially being boosted by the reintroduction of biodiversity into the crop system (Brooker et al. 2015, 2021; Martin-Guay et al. 2018) bringing further benefits for the environment and climate in terms of reduced greenhouse gas emissions and agrochemical inputs (Brooker et al. 2016, 2023).

While showing promise – with studies demonstrating yield gains from crop mixtures (see for example Brooker et al. 2024; Martin-Guay et al. 2018) or increased diversity of crop products (Li et al. 2023) - there remain some key questions that need to be addressed to provide reassurance to farmers and buyers and so help up-scale the growing of crop mixtures (Brooker et al. 2024). In part, we need to provide more information with respect to the optimisation and management of crop mixtures, and in part address knowledge gaps about their impact on crop quality and the wider environment.

Our study addressed these issues by combining data from multiple crop mixture trials conducted in arable systems in Scotland, many of them on working farms and managed using standard farming equipment. Using the substantial database assembled from these trials, we examined some key questions including whether crop mixtures provided yield gains, the response of any yield gains to climate and management, and the impacts of growing crops in mixtures on both crop chemical composition and soil chemistry.

Yield responses to crop diversity and composition

With respect to crop yield, our results mirror previous studies such as those of Martin-Guay et al. (2018), Schöb et al. (2023), and Brooker et al. (2024) which demonstrated yield benefits of growing crop mixtures. In our study focussed on Scottish systems we found approximately 20% higher yield per unit area compared to monoculture crops, with the majority of CPR values being greater than 1. In the Scottish context, and for the type of mixtures grown in these trials (predominantly cereal-legume mixes – see suppl. mat A), even at their worst, crop mixtures do as well as expected based on the yield of the monocultures, and they often achieve this with reduced inputs. Notably crop mixtures also showed their highest relative yield benefit when the productivity of monocultures was lower. This matches results from Brooker et al. (2024) which at the European scale found the yield benefits of intercropping to be greater when yields were lowest. This also matches expectations from ecological theories such as the Stress Gradient Hypothesis (Callaway 2007) that beneficial effects in plant communities - which could be delivering yield gains in crop mixtures (as discussed below) - could become stronger or more frequent in more stressful environmental conditions. However, we did not see an impact on CPR values of some factors which we might expect to regulate monoculture yield, including the addition of fertiliser or the occurrence of organic management, so the specific factor limiting monoculture yield may be important. CPR values were highest when monoculture yields were lowest, and this was associated with cooler and drier (i.e., lower rainfall) climatic conditions; the mechanisms driving mixture benefits may have been more strongly influenced by climate than by soil nutrient conditions or management.

As also expected from ecological theory, CPR increases with the diversity of the mixture. This type of positive response – a positive Biodiversity Ecosystem Function (BEF) relationship – is common in many ecological systems and for many functions such as ecosystem productivity, with substantial meta-analyses showing that on average these positive BEF relationships are the norm (Cardinale et al. 2012). Such responses have also been found in studies of crop diversity (Schöb et al. 2023). There can be a number of mechanisms driving these responses (these so-called overall Net Biodiversity Effects), and these have been broadly categorised as Selection Effects and Complementarity Effects (Loreau and Hector 2001). Selection Effects occur when the most productive species in monoculture dominates the mixture while Complementarity Effects result from niche complementarity or facilitative effects (both direct and indirect). Complementarity Effects have been shown to contribute substantially to yield benefits of intercropping, for example driving 90% of yield gain in a meta-analyses of intercrops in China (Li et al. 2020a). In most cases in our study the harvested crop was not separated such that the productivity of individual components could be assessed, meaning we cannot calculate Selection and Complementarity Effects; there would be clear benefits from doing this in future studies to enable an assessment of the relative strengths of these effects and how they might vary with management or climate.

Importantly, within this overall positive effect of diversity, crop identity also matters. Interestingly, given what we might expect about the potentially beneficial effects of legumes (Brooker et al. 2015, 2024 and references therein), and compared to mixtures without legumes, adding either beans or peas alone did not increase CPR. Martin-Guay et al. (2018) also found no difference in the LER values of intercrops including and not including a legume. This indicates that while particular mechanisms associated with legumes may operate to generate a positive CPR when legumes are present – for example complementary N use in mixtures with legumes (Engbersen et al. 2021; Schöb et al. 2023) - other mechanisms (for example niche complementarity) can generate equally strong CPR responses in the absence of legumes.

Although adding either beans or peas alone had no significant effect on the CPR of intercrops, adding both beans and peas in combination did enhance CPR. There were also differences between mixtures depending on the identity of the cereal, with strong positive effects of oats, and whether the mixture contained linseed. The effect of beans and peas in combination may simply be an extension of the positive BEF effect discussed above, and because there was only one study with a pea-bean combination it is perhaps not sensible to generalise too far from this result. However, the observed differences in CPR between different 2-species mixtures points towards differences in the compatibility of particular trait combinations. Oats, for example, are known to have allelopathic characteristics which can suppress soil-borne pathogens (e.g., van Elsas et al. 2002; Williams-Woodward et al. 1997) and have been shown to promote legume nitrogen fixation (Tsialtas et al. 2018) and micronutrient accumulation (Zuo and Zhang 2008), possibly mediated by root exudates (Homulle et al. 2022). The potentially beneficial characteristics of linseed are less clear, but may be related to rooting characteristics that complement relatively shallow-rooted cereals. Flax roots can grow to 90–120 cm deep (Gill 1987), although only 4–7% of roots grow deeper than 60 cm (Hall et al. 2016). Work is clearly needed to further explore the combinations of plant traits that help to enhance crop mixture yield gains, for example the work summarised by Schöb et al. (2023). However, it is important to remember that such work will help to improve the yield gains that already exist in intercrops, and that - irrespective of their composition - at worst our crop mixtures yielded as much as we would expect based on their component monocultures.

Impacts of climate on yield

Climate as well as mixture composition had clear and strong impacts on CPR. Our study showed enhanced CPR under cooler and drier conditions, in contrast to the results of Brooker et al. (2024) whose trans-European study showed higher CPR associated with higher rainfall. These differences in results may be related to the different scales of these two studies, the prevailing limiting factors across the environmental gradients that they encompass, and the range of crops grown. At the European scale and along the gradient explored by Brooker et al. (2024), low soil moisture might represent droughting conditions, whereas in our Scottish-focussed study higher levels of soil moisture might represent an excess (i.e., waterlogging of the soil). Both these extremes of soil water availability could limit beneficial crop interactions. Schöb et al. (2023) found differences in crop mixture performance between experimental crop mixture mesocosms located in Spain and Switzerland; in their case LER values were higher in Spain, but there were variations in these responses depending on crop identity and fertiliser applications. Ramirez and Wright (2023), studying the ‘Three Sisters’ polyculture system in an urban garden setting found polyculture yield advantages in a high light environment when soil moisture was also limiting, and also that reduced microclimate temperatures were associated with increased polyculture yield advantage. These studies demonstrate the ability of crop mixtures to mediate the impacts of large-scale environmental drivers and gradients. While understanding the causes of differences in polyculture or crop mixture responses between studies - and linking these to impacts on local micro-climatic conditions within the crop - are important future challenges, irrespective of the causes these results emphasize the need for regional-scale studies to help tailor crop mixtures to local conditions.

In our study of Scottish systems there is also an interesting pointer towards crop mixtures providing resilience under climate extremes. Specifically, we observed a trend (P = 0.085) toward a positive relationship between CPR and the highest temperature recorded during the growing season (Table 1). Our study period included the extremely hot and dry (for Scotland) summer of 2022, within which Scotland achieved a record summer temperature of 34.8 °C (MetOffice 2024). As noted above, the impacts of mixed crops (or polycultures) on microclimatic conditions might buffer growth-limiting environmental conditions (Ramirez and Wright 2023; Schöb et al. 2023). The potential of more biodiverse vegetation to better ameliorate more severe – particularly hot and dry - climatic conditions has been demonstrated by other studies as well, for example that of Wright et al. (2021). While this capacity will vary depending in part of the functional groups (and hence identity) as well as richness of the mixture, overall increasing the biodiversity of plant communities could act as a nature-based solution to the impacts of climatic warming and drying (Wright and Francia 2024). However, while we cannot be certain of the precise mechanism operating in our study, our field observations in 2022 indicated particularly strong negative effects of the hot and dry weather on some of the monocrops, especially peas and beans. It may be therefore that we are seeing two types of benefits from crop mixtures: under less extreme growing conditions we see greater CPR when biological interactions between mixtures components are occurring or are enhanced (i.e., in the absence of waterlogging), whereas in extreme years the mixtures contain at least one component capable of growing under the extreme conditions. To put it another way, under normal growing conditions mixture benefits result from active beneficial interactions (Complementarity Effects) between neighbours, whereas under extreme climatic conditions mixture benefits result from an “insurance effect”, an important component of the benefits expected from enhancing biodiversity in general (Cardinale et al. 2012) including in food production systems (Brooker 2016).

Impacts of management on yield and weed cover

Interestingly we found no clear impacts of management on CPR except a significant positive effect of herbicide use and a trend towards a positive effect of ploughing. The absence of a significant effect of fertiliser addition is in contrast to the findings of Li et al. (2020b) who found stronger yield gains from intercrops under lower levels of nutrient input. This may result in part from the increased competitiveness of cereals and decreased competitiveness of legumes under higher N-input conditions (Yu et al. 2016), and matches results showing that yield gain from cereal-legume intercrops in China did not increase with increased N input (Li et al. 2020a). It is possible though that even the fertilised trials in our database correspond to the low input ‘alternative production syndrome’ of intercrop production (Li et al. 2020b).

Returning to the positive effects of herbicide use and ploughing, these may be linked; ploughing helps regulate weed cover, and this is supported by our observed reductions in weed cover in systems which use ploughing. Weed management is a clear challenge for crop production as weeds compete with crops for resources (George et al. 2022). However, it is notable that while ploughing reduced weed cover, herbicide use did not, so is it reasonable to argue that the positive effect of ploughing and herbicide use on CPR was due to reduced crop-weed competition? A factor here may be the different timings and scale of impact of treatments and measurements, and the variability in measured responses. First, herbicide additions occur early in the growing season to “knock back” germinating weeds prior to crop germination, enabling enhanced establishment and initial growth of the crop. But crop yield and weed cover are assessed later in the season by which time, while the signal of early enhanced establishment may still be present in the treated crop (resulting in a detectable positive effect of herbicide treatment on CPR), weed community cover may have recovered towards that found in the untreated plot (resulting in no effect of herbicide treatment on weed cover). In contrast the effect of ploughing may be more substantial and long lasting such that it is still detectable at the time of weed cover assessment. Second, there might be much greater variability between plots in the composition of the weed community (and weed cover) than there is in the composition of the crop (and crop yield), making a herbicide signal more difficult to detect in the weed community.

An alternative mechanism is that ploughing and fertiliser addition have different effects on the composition of the weed community which in turn can influence the interactions between the weed and the crop (Stefan et al. 2021). More diverse weed communities can limit the negative impact on the crop of particularly dominant and competitive weeds (Adeux et al. 2019). If, for example, herbicide selectively reduced the cover of dominant weeds, this could explain why there may have been a beneficial effect on CPR without an observable reduction in weed cover. While the precise mechanism is unclear, it is important to emphasise that our comparison of monocultures and mixtures was done within-site, and on areas of land with similar seedbanks and cropping histories (i.e., generally within a single field). Substantial and systematic differences in weed composition between the monoculture and mixture plots are therefore unlikely – certainly at the start of the growing season.

We also need to consider why CPR – the benefit from the crop mixture – is enhanced when weeds are controlled. As for the responses to climate drivers, this may be linked to the processes which deliver a positive CPR, in this case niche complementarity. Weeds compete with crops for resources (George et al. 2022). We would expect crop monocultures to utilise less of the available resource pool than two crops growing together in a mixture, especially if those two crops have contrasting traits such as rooting depth. Therefore, for a given level of weed cover, monoculture crops may experience in total less competition from the weeds than do mixtures, which are utilising more of the resource pool that would otherwise be free for the weeds to use. This would then mean that control of the weeds would have a proportionately bigger benefit for the performance of the mixtures compared to the monocultures, and hence a positive effect on CPR. Again, this is an interesting possibility and would reward further work which drills down into the mechanism underlying the net responses of overall crop yield as measured by CPR.

Responses of seed chemistry and soil C

A particular concern for farmers is that growing crops in mixtures might alter the quality of the crop and reduce its commercial value. We found few changes in seed chemical composition as a result of growing crops in mixtures. In the cereals, we only detected significant changes in wheat grain chemical composition, with enhanced N and C content in wheat from mixtures (Fig. 3) perhaps as a consequence of increased seed oil rather than carbohydrate content. Alternatively reduced cereal yield in the mixtures may have resulted in higher protein and hence N content (for example see Fossati et al. 1993), although this couldn’t be assessed because in most cases we did not have partial yields from components in the mixture. In the legumes we detected reductions in pea N and P content, perhaps because some of the soil available P and N pool was being utilised by the legume’s crop mixture companion, and an enhanced C: N ratio (Fig. 4). Although statistically significant these changes are comparatively small. Notably barley – around which there is considerable interest because of its sale into the high-return malting market – showed no significant changes in the grain chemistry parameters assessed, although we recognise this is only one aspect of overall grain quality.

As noted above, changes in seed chemistry might result from changes in soil chemistry, for example reduced pea N and P content resulting from competition with cereals in crop mixtures. Some of the changes in soil chemistry that we detected are not surprising, for example increased (extractable) soil NO3-N in the presence of legumes, a response which might result from legumes utilising atmospherically derived rather than soil N (Brooker et al. 2015 and references therein). However, mixtures with legumes were not the only combinations where we detected changes in soil chemistry. Cereal-linseed mixtures also showed enhanced NO3-N, and NH4-N, which would indicate that in these mixtures at least soil N availability is being governed by processes other than N fixation. One possibility is that increased plant (crop) species diversity – through mechanisms such as increased soil exudate and structural complexity – enhances soil processes including decomposition, leading to greater N release (Chen et al. 2021; Zhu et al. 2023). This mechanism might also drive the reductions in soil C content observed in cereal-linseed and 3-species mixtures. Mixtures with legumes did not show such as strong reduction in soil C, again indicating that the processes driving enhanced soil N availability differ between mixtures with and without legumes. Notably 3-component mixtures seem to have some of the largest responses (Fig. 5) which may indicate a combined impact of N fixation and other processes such as soil priming and mixture-driven manipulation of the soil microbiome (Liu et al. 2020; Murphy et al. 2017; Paterson et al. 2007; Schöb et al. 2023) on soil nutrients and C. Finally with respect to soil chemistry responses, it is worth noting that our soil samples came from the top 10 cm of the soil. Although we see reductions in soil C in this soil layer, this does not mean that crop mixtures reduce total soil C. It is possible that soil C is instead being sequestered in deeper soil layers, particularly if deep rooted and legume species are included in the mixture (Peixoto et al. 2022).

Conclusion

Crop mixtures may provide an important part of the toolkit needed to deliver sustainable cropping systems which maintain crop production while addressing the climate and nature crises. Our study aimed to address a number of uncertainties which may at present be slowing the wider roll-out and uptake of crop mixtures, specifically understanding the response of crop mixture yields to climate and management drivers, and the impacts of growing crop mixtures on aspects of crop quality (crop seed chemistry), and soil C and nutrients. Overall, we found on average that crop mixtures delivered yield gains, which were enhanced by increasing the diversity of the mixture, and had limited impacts on crop seed chemistry.

A number of factors were associated with variation in the benefits observed from crop mixtures. We found influences of management which appeared to be linked to weed control (ploughing and the use of herbicides), as well as influences of climate, in particular enhanced crop mixture performance under cooler and drier conditions. We also found variation in mixture performance depending on the composition of the mixture. But irrespective of such variability between trials, in all cases crop mixtures performed at least as well as expectations based on monocultures. This indicates that, while further study can help tailor management and crop composition to maximise the benefits from crop mixtures, even without such refinement crop mixtures provide reliable crops without yield losses or substantial changes in seed chemistry.

However, a key area for future research must be the impact of crop mixtures on soil C, wider soil chemistry and processes and soil biology that mediate many key soil processes. We have suggested that crop mixtures, while appearing to reduce soil organic matter in the upper soil layers, may be associated with C sequestration in deeper layers. It is essential to assess first whether the detected response of soil organic matter content is found in other crop mixture systems, and second whether we do indeed see compensatory soil C changes in deeper soil layers. If this is the case it will help to provide further reassurance that crop mixtures can indeed deliver the “win-win-wins” needed to tackle the triple challenge of climate change, biodiversity loss, and sustainable food production.