The experiment was conducted during the 2020 and 2021 cropping seasons at the Kouaré Research Station, located in the Sudano-Sahelian zone of Burkina Faso, particularly in the province of Gourma ( Figure 1 ).

The study area is characterized by a distinct rainy season from May to October and a dry season from November to April. Climatic data (2020-2021) indicate that the site receives average annual precipitation of approximately 850 mm, with peak rainfall occurring in July, August, and September ( Figure 2 ). During these months, monthly rainfall often exceeds 150 mm. Average monthly temperatures remain relatively stable throughout the year, ranging from 23˚C to 32˚C ( Figure 2 ). According to the WRB (World Reference Base for Soil Resources), the main soil types in the study site include Regosols, Lixisols, Cambisols ( Figure 1 ), which are typical of Burkina Faso and are moderately leached, with low fertility [35] [36] . These pedoclimatic conditions, characterized by seasonal rainfall, high temperatures, and sandy, nutrient-poor soils pose significant constraints to agricultural production and influence weed community dynamics, making the site ideal for evaluating the effects of agroecological management practices.

The plant material used in this study included a cereal species, sorghum and a legume species, represented by two different varieties of cowpea: Nerwaya and Moussa local. The sorghum variety used was ICSV 1049, which has a growth cycle of 110 to 115 days. The cowpea variety Nerwaya is characterized by a semi-erect growth habit and a short maturity period of about 70 days. Moussa local, another cowpea variety, has a cycle of 75 to 80 days and a creeping habit.

A split-split plot design was used with complete randomization and three replications ( Figure 3 and Figure 4 ). The main plots were assigned to three tillage treatments (T):

T1: no-tillage before sowing.

T2: conventional tillage (mechanized plowing at a depth of 20 cm with tractor before sowing);

T3: ridge tillage.

The subplots were used for four cropping systems (C), involving different spatial arrangements of sorghum and cowpea:

C1: two rows of sorghum (128 plants × 4 = 512 plants) alternated with two rows of semi-erect cowpea (64 plants × 4 = 256 plants);

C2: two rows of sorghum (128 plants × 4 = 512 plants) alternated with two rows of creeping cowpea (64 plants × 4 = 256 plants);

C3: one row of sorghum (96 plants × 4 = 384 plants) alternated with one row of semi-erect cowpea (96 plants × 4 = 384 plants);

C4: one row of sorghum (96 plants × 4 = 384 plants) alternated with one row of creeping cowpea (96 plants × 4 = 384 plants).

The sub-subplots received four fertilization treatments (F):

F1: no fertilizer;

F2: compost at 2.5 t.ha⁻¹;

F3: mineral fertilizer at 100 kg.ha⁻¹ NPK and 50 kg.ha⁻¹ urea;

F4: compost (2.5 t.ha⁻¹) + mineral fertilizer (100 kg.ha⁻¹ NPK and 50 kg.ha⁻¹ urea).

The compost used is based on cow dung. It had a neutral pH of 7.02 and contained 1.75% total nitrogen and 33.83% total carbon. The total phosphorus and total potassium contents were 2.83 g.kg⁻¹ and 4.48 g.kg⁻¹, respectively. Each elementary plot measured 6 m × 4 m (24 m²) and the soil is characterized by carbon content of 0.56%, a clay content of 11.80%, a silt content of 12.08% and a sand content of 76.12%. Sorghum and cowpea were sown in hills spaced 80 cm apart between rows and 40 cm between hills. Sowing was done in July for both experimental years. Thinning was performed 14 days after sowing (DAS) to retain only two plants per hill. NPK fertilizer (14-23-14) was applied using microdosing techniques 15 DAS. Urea (46% nitrogen) was applied at 45 DAS, also using microdosing.

Weed surveys were conducted in both 2020 and 2021, with two sampling periods per cropping season. The first inventory was performed 35 days after sowing (DAS) and the second at 65 DAS to ensure an exhaustive inventory of all weed species during crops cycles. The perimeter walk method [37] was employed to identify all emerged weed species. Species identification was based on direct morphological observation and confirmed using standard botanical references [38] [39] . Weed abundance was estimated using the density scale of Barralis (1976), while ground cover was assessed according to Marnotte’s visual scale [40] [41] . Weed biomass was determined following each survey by randomly placing a 1 m² quadrat in each plot. All weeds within the quadrat were uprooted, bagged ( Figure 5 ), oven-dried at 50˚C for 72 hours, and weighed to determine dry biomass.

Data collected over the two cropping seasons were analysed using R program version 4.4 [42] . Canonical Correspondence Analysis (CCA) and Permutational Multivariate Analysis of Variance (PERMANOVA) were used to assess the influence of agricultural practices (tillage, intercropping systems, and fertilization) on weed community ( Table 1 ). CCA was performed using weed community data matrix and agricultural practices as constraining matrix. Bray-Curtis distance was used for PERMANOVA with 999 permutations. Following CCA and PERMANOVA, the IndVal (Indicator Value) method described by Dufrêne and Legendre [43] was employed to identify characteristic weed species (Species almost exclusively associated with a given agricultural practice occur regularly and abundantly in that environment but are absent elsewhere. IndVal is an index used to identify indicator species in ecological analyses. It measures the association between a species and a specific group of sites. The IndVal statistic (shown in Table 2 ) ranges from 0 to 1, where 1 represents a perfect indicator species that occurs only within a particular group and is present at all sites in that group). The effects of agricultural practices on weed development, assessed through ground cover and biomass, were analysed using generalized linear models (GLMs) with a Gamma error distribution. This choice is justified by the fact that weed cover and biomass data are positive and continuous response variables. These are quantitative variables that can take an infinite number of values within a given interval, including fractional and decimal values. In the models, weed development (ground cover and biomass) is response variable and agricultural practices are explanatory variables. Tukey’s HSD test was further used to separate the means when GLMs revealed significant differences. The CCA and PERMANOVA were performed with the vegan package [44] and IndVal value was calculated with the indicspecies package [43] . All figures were performed with the ggplot2 package [45] , and the significance threshold for all statistical tests was set at 5%.

The tillage method is the main agricultural practice significantly influencing weed diversity ( Figure 6 and Table 1 ). No-tillage (T1) is characterized by the presence of five species (p-value < 0.05), among which Cochlospermum tinctorium Perr. Ex A. Rich. is the sole characteristic species. In conventional tillage (T2), Mollugo cerviana (L.) Ser. ex DC is the characteristic species. In contrast, ridge tillage (T3) supported eight weed species (p-value < 0.05) characterized by Ceratotheca sesamoides Endl., Sida urens L. and Echinochloa colona (L.) Link ( Table 2 ).

R², F-value and P-value provided by PERMANOVA analysis.

Index, stat and p-value provided by IndVal analysis. T1: no tillage before sowing; T2: conventional tillage; T3: ridge tillage. Abbreviations written in capital letters are taken from the EPPO Global Database. All abbreviations, and the corresponding list of species, are provided in Appendix.

Gray scatter plot and green color represent weed species community and red arrow represents agricultural practices. The length of arrow shows the influence of the variable on weed community.

The longer the arrow, the greater the influence, and the shorter the arrow, the less influence the variable has on the species. T1: no tillage before sowing; T2: conventional tillage; T3: ridges tillage ; C1: two rows of sorghum alternated with two rows of semi-erect cowpea; C2: two rows of sorghum alternated with two rows of creeping cowpea; C3: one row of sorghum alternated with one row of semi-erect cowpea; C4: one row of sorghum alternated with one row of creeping cowpea; F1: no fertilizer; F2: compost at 2.5 t.ha⁻¹; F3: mineral fertilizer at 100 kg.ha⁻¹ NPK and 50 kg.ha⁻¹ urea; F4: compost (2.5 t.ha⁻¹) + mineral fertilizer (100 kg.ha⁻¹NPK and 50 kg.ha⁻¹ urea). Abbreviations written in capital letters are taken from the EPPO Global Database. Those written in lowercase do not currently have an EPPO code. All abbreviations, and the corresponding list of species, are provided in Appendix.

Weed ground cover was significantly influenced by fertilization and tillage practices, while total biomass was significantly affected only by tillage modes ( Table 3 ). Mineral fertilization alone (F3) resulted in the highest weed cover (19.39 ± 15.57%; p-value < 0.05), representing increases of +35.49% compared to the non-fertilized treatment (F1) and +18.14% compared to compost application alone (F2) ( Figure 7 ). Additionally, ridge tillage (T3) and no-tillage (T1) presented significantly lower weed cover compared to conventional tillage (T2), which showed the highest value (20.93 ± 15.05%). Compared to conventional tillage (T2), no-tillage (T1) and ridge tillage (T3) reduced weed cover by −17.73% and −37.17%, respectively ( Figure 7 ). Finally, the lowest biomass (17.59 ± 13.70 g.m⁻²) was observed with ridge tillage (T3), corresponding to reductions of −33.97% relative to no-tillage (T1; 26.64 ± 14.48 g.m⁻²) and -32.29% compared to conventional tillage (T2; 25.98 ± 14.41 g.m⁻²) ( Figure 8 ).

F-value and P-value provided by GLM analysis.

This study shows that tillage intensity is a major determinant of weed community composition in sorghum-cowpea intercropping systems within the Sudano-Sahelian zone. Each tillage system created distinct weed communities with characteristic species. No-tillage systems were characterized by Cochlospermum tinctorium as the sole indicator species, reflecting this West African savanna plant’s adaptation to undisturbed soil conditions [46] - [48] . Conventional tillage favored Mollugo cerviana, typical of opportunistic annual species that exploit regularly disturbed habitats. Ridge tillage supported the most diverse community, characterized by Ceratotheca sesamoides, Sida urens, and Echinochloa colona, representing both broadleaf and grass functional groups [49] [50] .

The observation of higher weed species richness under ridge tillage presents both opportunities and challenges for crop management [51] . Indeed, the diverse nature of the weed flora can, paradoxically, mitigate the pressure exerted by certain highly competitive species [52] . Additionally, some weeds can secrete allelopathic compounds that inhibit the germination or growth of more harmful competing species, contributing to a less destructive community balance for the crop [53] .

However, higher species richness can complicate certain aspects of weed control such as resistance risk. The solution would be to regularly monitor weed composition and adapt strategies based on the evolving dominant species and their responses to interventions [54] .

The differential responses reflect underlying seed bank dynamics and microhabitat heterogeneity. Indeed, no-tillage concentrates weed seeds near the soil surface, favoring established perennial species like C. tinctorium [55] [56] . Moreover, reduced soil disturbance in no-tillage systems limits vertical soil movement, thereby restricting the exposure of buried weed seeds to light and oxygen, key factors for germination [57] [58] . Conventional tillage distributes seeds throughout the tillage layer, creating optimal conditions for fast-germinating annuals. Ridge tillage creates the most heterogeneous environment with varied moisture, light, and soil conditions across ridge-furrow topography, supporting diverse ecological niches within the same field [51] [59] .

These findings show that tillage system selection can strategically influence weed community composition. While ridge tillage supports higher characteristic species, diverse weed communities are often more stable and less prone to dominance by highly competitive single species [60] [61] . In West African semi-arid conditions, these tillage effects are mediated by environmental constraints including variable rainfall and nutrient-poor soils [62] [63] . The observed patterns reflect both direct tillage disturbance effects and indirect effects through altered soil water retention and organic matter dynamics, which influence weed-crop competition relationships.

Interestingly, variations in intercropping configurations did not significantly influence weed species diversity. This lack of effect may be attributed to similar canopy architecture and ground coverage among the intercropping treatments, which likely did not generate sufficient differences in microclimatic conditions to affect weed assemblages [64] - [66] . Indeed, previous studies have shown that more pronounced differences in crop spatial arrangement, such as strip intercropping versus intimately mixed configurations, are often needed to significantly alter weed suppression dynamics [57] [67] .

Likewise, while fertilization increased crop growth, it had no significant effect on weed diversity, suggesting that nutrient enrichment mainly stimulates the growth of dominant weed species without substantially altering species composition [68] [69] .

These observations align with the theory of disturbance-mediated community assembly, whereby the nature and intensity of disturbance act as ecological filters that select species with specific functional traits [70] - [72] . Conventional tillage, by enhancing disturbance, fosters trait divergence among weed species, whereas no-tillage favour species with conservative traits such as high seed production, dormancy, and shallow germination [73] [74] . While our study did not include direct measurement of functional traits or formal classification into functional groups, the observed shifts in species composition suggest changes in the functional profile of weed communities across tillage treatments. Managing tillage regimes may therefore represent a valuable tool for shaping the functional composition of weed communities and enhancing ecosystem services such as erosion control and pollinator support [12] [33] [72] .

Regarding weed development, our results reveal that both tillage and fertilization practices significantly influenced weed ground cover, while weed biomass was mainly affected by tillage intensity. Mineral fertilization led to the highest weed cover, likely due to rapid nutrient availability promoting nitrophilous species with high growth rates and competitive ability [32] [75] . These results support prior research showing that fertilization enhances weed emergence and dominance, especially of fast-growing species [10] [76] [77] .

Conversely, compost-based treatments, whether used alone or in combination with mineral fertilizer, reduced weed ground cover. This may be explained by the slower nutrient release of organic amendments and their positive effects on soil structure and microbial activity, which create less favorable conditions for the proliferation of opportunistic weeds [78] . Compost applications have also been linked to improved soil moisture retention and biological resilience, thereby suppressing weed growth and stabilizing weed community dynamics [11] [79] [80] .

Tillage practices also influenced weed biomass, with conventional tillage supporting higher biomass than no-tillage and ridge tillage. Ridge tillage achieved the lowest biomass values, representing a reduction of over 30% compared to the other tillage methods. This may result from the creation of micro-topographical heterogeneity, which affects water distribution and soil temperature dynamics, thereby limiting favorable conditions for weed establishment [81] [82] [83] .

These findings underscore the value of ridge tillage as a component of integrated weed management strategies. Ridge tillage not only reduces weed pressure but also supports agroecological goals by minimizing external inputs and potentially contributing to system sustainability [4] [83] [84] . When combined with organic amendments, ridge tillage offers a synergistic pathway for sustainable intensification by suppressing weeds, improving soil health, and potentially supporting a more balanced weed community structure. However, this assertion requires further investigation through dedicated trait-based analyses and functional diversity measurements [85] [86] .