Introduction

Africa has one of the fastest-growing populations globally, projected to reach approximately 2.5 billion people by 2050, up from 1.3 billion today (United1). Population growth is already surpassing food supply in the Sahel, increasing many countries’ reliance on food imports2. Therefore, sustainable increases in crop production are essential to ensure food security and sovereignty in Africa3. The sustainable intensification of agriculture—through investments in irrigation systems, more efficient use of chemical fertilizers, and pesticides, which are expensive and environmentally damaging—can significantly reduce yield gaps on marginal lands4,5,6.

Rice is one of the world’s most important cereal crops, serving as the staple food for over half of the global population7. Typically, 3000–5000 L of water are needed to produce one kilogram of rice, which is two to three times the amount required for other cereals8. However, water availability is expected to decline in the coming years, especially in the Sahel region9. Reducing water use in rice production could have significant societal and environmental benefits if the saved water can be allocated to areas with high water demand10. Alternate Wetting and Drying (AWD) is one of the most promising methods to reduce water use in rice production systems, particularly in Africa11,12,13. In AWD, irrigation is periodically paused until the soil water table drops to a specific depth, at which point irrigation is resumed14. These cycles of wetting and drying can be applied during either the vegetative or flowering stages or throughout the entire rice growth cycle. This approach can save up to 43% of irrigation water15,16. However, AWD can lead to a yield penalty, which could be mitigated by optimizing agricultural practices, such as fertilization, and plant breeding17.

Due to the high costs of chemical inputs (e.g., nitrogen, phosphorus, and potassium fertilizers, as well as pesticides) and the environmental issues related to their overuse, such as water and soil pollution18,19,20, there is an urgent need to develop input-efficient rice production systems that can maintain or even increase yields while reducing reliance on conventional intensive farming practices21.

Arbuscular mycorrhizal fungi (AMF) form symbiotic relationships with plants, including rice22,23. AMFs enhance plant access to nutrients such as phosphate, nitrogen, and potassium in exchange for photosynthesis provided by the host plant. They play crucial roles in increasing resistance to both biotic and abiotic stresses24,25, improving water use efficiency and overall plant performance in both natural and agricultural systems23,26. AMF inoculation has been shown to improve crop productivity under stress conditions by enhancing water relations, gas exchange, osmotic adjustment, and antioxidant defense mechanisms27. It was shown that AMF inoculation coupled with P fertilization improved the grain yield of rice under limited irrigation conditions in greenhouse conditions28. However, the impact of AMF inoculation on rice production under low-water and low-input conditions in realistic field conditions remains poorly documented.

We hypothesized that AMF could mitigate yield losses under low doses of chemical fertilizers and in water-saving cultivation systems. The objective of this study was to assess the impact of AMF inoculation on rice growth and yield under field conditions, across different levels of inorganic fertilization and irrigation regimes (continuous flooding and AWD). We report two years of field trials demonstrating that AMF inoculation improves the agronomic performance of both irrigated and upland rice varieties under low-input conditions.

Results

Grain yield of irrigated and upland rice as a function of irrigation regime, fertilizer level and inoculation with AMF

To assess the impact of AMF inoculation on rice growth and yield at varying levels of inorganic fertilization and under different irrigation regimes, field trials were conducted over two consecutive years in Fanaye (Senegal). An LME analysis revealed that the effects of irrigation regime, fertilizer level, and AMF inoculation on grain yield were dependent on the ecotype (Fig. 1; Table S2). Alternate wetting and drying (AWD) irrigation (at 60 kPa SMP (soil matric potential)) led to a significant reduction in grain yield compared to continuous flooding (CF), with a more pronounced effect on the irrigated ecotype (29% yield reduction for irrigated vs. 10% for upland ecotype). For both irrigated and upland ecotypes, the effect of AMF inoculation on grain yield was significantly influenced by fertilizer level (Fig. 1; Table S3). A significant AMF inoculation × irrigation regime interaction was detected only in irrigated rice, while both main effects were significant on grain yield of both irrigated and upland ecotype (Fig. 1, Figure S1, Table S3). Additionally, the beneficial effect of AMF inoculation was more prominent at half the recommended dose of NK without P application (F2; 75:00:30 kg NPK ha−1) compared to other fertilizer levels, irrespective of the irrigation regime (Fig. 1, Figure S2).

Fig. 1
figure 1

Grain yield of irrigated and upland ecotype inoculated (InoAMF) or non-inoculated with AMF (NinAMF) under continuous flooding (CF) and alternate wetting and drying (AWD) irrigation at different fertilizer levels (F1 = 00:00:00 kg NPK/ha; F2 = 75:00:30 kg NPK/ha; F3 = 75:30:30 kg NPK/ha; F4 = 150:00:60 kg NPK/ha; F5 = 150:60:60 kg NPK/ha). Different letters indicate significant differences as determined by a Tukey HSD test at the 5% level of significance.

Full size image

In-depth analyses of each rice ecotype revealed that the fertilization x irrigation regime as well as fertilization x AMF inoculation interactions were significantly influenced by rice variety (Fig. 2; Table S3). For irrigated varieties, Sahel 108, Sahel 202, and IR 64 exhibited the lowest yields (1452, 1488, and 1218 kg/ha, respectively) under non-inoculated, unfertilized, and AWD conditions (Fig. 2A). Sahel 108 and Sahel 202 achieved the highest yields (9,558 and 8,698 kg/ha, respectively) under inoculated conditions with half the recommended dose of NK without P (F2; 75:00:30 kg NPK/ha) and CF conditions, while IR 64 reached its highest yield (7,295 kg/ha) under inoculated conditions with the full recommended dose of NPK (F5; 150:60:30 kg NPK/ha) and CF. Under CF, grain yield decreased at half the recommended dose of NPK (F3; 75:30:30 kg NPK/ha) compared to the full recommended dose (F5; 150:60:60 kg NPK/ha), while this trend was not observed under AWD irrigation for all irrigated varieties (Fig. 2A). Furthermore, AWD irrigation caused significant yield losses in non-inoculated plants at half the recommended doses of NPK (F3; 75:30:30 kg NPK/ha) and NK (F2; 75:00:30 kg NPK/ha). However, AMF inoculation almost completely offset these AWD-related yield losses, particularly for Sahel 108 and Sahel 202 (Fig. 2A).

Fig. 2
figure 2

Yield of irrigated (A) and upland (B) rice varieties inoculated (InoAMF) and non-inoculated with AMF (NinAMF) under continuous flooding (CF) and alternate wetting and drying (AWD) irrigation at different fertilizer levels (F1 = 00:00:00 kg NPK/ha; F2 = 75:00:30 kg NPK/ha; F3 = 75:30:30 kg NPK/ha; F4 = 150:00:60 kg NPK/ha; F5 = 150:60:60 kg NPK/ha). Different letters indicate significant differences as determined by a Tukey HSD test at the 5% level of significance.

Full size image

For upland varieties, NERICA 4, WAB56_104, and CG14 exhibited the lowest yields (484, 683, and 1,492 kg/ha, respectively) in non-inoculated and unfertilized plants under AWD conditions (Fig. 2B). CG14 achieved the highest yield (4,913 kg/ha) with inoculated plants under CF at the recommended dose of NK (F4; 150:00:60 kg NPK/ha), while WAB56_104 and NERICA 4 displayed the highest yields (4,652 and 4,130 kg/ha, respectively) with inoculated plants under CF at the recommended dose of NPK (F5; 150:60:60 kg NPK/ha). NERICA 4 showed a substantial reduction in grain yield at half the recommended dose of NPK (F3; 75:30:30 kg NPK/ha) compared to the recommended dose (F5; 150:60:60 kg NPK/ha) under both CF and AWD irrigation. The application of half the recommended dose of NPK (F3; 75:30:30 kg NPK/ha) also resulted in yield losses under CF for WAB56_104, whereas it did not significantly affect the yield of CG14 under either CF or AWD irrigation (Fig. 2B). CG14 exhibited yield losses in non-inoculated plants under AWD irrigation compared to non-inoculated plants under CF; however, AMF inoculation compensated for yield losses caused by AWD irrigation, regardless of fertilizer level. This pattern was also observed for WAB56_104 at the recommended dose of NPK (F5; 150:60:60 kg NPK/ha) and half the recommended dose of NK (F2; 75:00:30 kg NPK/ha). In contrast, NERICA 4 did not show significant yield losses under AWD irrigation, regardless of fertilizer level or inoculation status (Fig. 2B).

Inoculation with AMF reduces yield losses due to AWD irrigation and low fertilization

We determined the relative yield losses (RYL) in irrigated and upland rice varieties, both inoculated and non-inoculated, caused by reduced water and fertilization. The yield of non-inoculated plants under continuous flooding (CF) at the recommended NPK dose (F5; 150-60-60) was used as the yield reference.

At the ecotype level, significant RYL were observed in irrigated rice varieties under alternate wetting and drying (AWD) irrigation at all fertilizer levels, regardless of AMF inoculation. In upland rice, AWD irrigation caused RYL at all fertilizer levels in non-inoculated plants. However, inoculation with arbuscular mycorrhizal fungi (AMF) fully offset AWD-related RYL at half the recommended NK dose (F2; 75-00-60), as well as at the recommended doses of NK and NPK (F4 and F5; see Fig. 3).

Fig. 3
figure 3

Relative grain yield losses of irrigated and upland ecotype inoculated (InoAMF) and non-inoculated with AMF (NinAMF) under continuous flooding (CF) and alternate wetting and drying (AWD) irrigation at different fertilizer levels (F1 = 00:00:00 kg NPK/ha; F2 = 75:00:30 kg NPK/ha; F3 = 75:30:30 kg NPK/ha; F4 = 150:00:60 kg NPK/ha; F5 = 150:60:60 kg NPK/ha). The red horizontal line (0%) represents yield reference which corresponds to the grain yield of non-inoculated plants under continuous flooding at recommended dose of NPK (150:60:60 kg NPK/ha).

Full size image

For both ecotypes, reducing fertilizer levels under CF led to significant yield declines in non-inoculated plants, except at the recommended NK dose (F4). However, AMF inoculation compensated for these yield declines, particularly at half the recommended NK dose (F2). Under CF, the yield of inoculated irrigated and upland rice surpassed the yield reference at the recommended NPK (F5) and NK (F4) doses. However, AMF inoculation was insufficient to offset yield declines at zero NPK application (F1) across all irrigation regimes (see Fig. 3).

For irrigated varieties, AWD-related RYL in non-inoculated plants ranged from 31 to 83% for IR 64, from 32 to 84% for Sahel 108, and from 24 to 81% for Sahel 202, with the greatest losses occurring at zero NPK application (F1; see Fig. 4). In contrast, AMF inoculation reduced AWD-related RYL across all fertilizer levels and varieties (see Fig. 4, Table S4). Yield loss reductions due to AMF inoculation (RYLRInoAMF) under AWD irrigation ranged from 7% at zero NPK application (F1) to 11% at the recommended NK dose (F4) for IR 64, from 9% at the recommended NPK dose (F5) to 22% at half the recommended NK dose (F2) for Sahel 108, and from 10% at the recommended NK dose (F4) to 36% at half the recommended NK dose (F2) for Sahel 202 (see Fig. 4, Table S4).

Fig. 4
figure 4

Relative grain yield losses of irrigated rice varieties inoculated (InoAMF) and non-inoculated with AMF (NinAMF) under continuous flooding (CF) and alternate wetting and drying (AWD) irrigation at different fertilizer levels (F1 = 00:00:00 kg NPK/ha; F2 = 75:00:30 kg NPK/ha; F3 = 75:30:30 kg NPK/ha; F4 = 150:00:60 kg NPK/ha; F5 = 150:60:60 kg NPK/ha). The red horizontal line (level 0%) represents yield reference which corresponds to the grain yield of non-inoculated plants under continuous flooding at recommended dose of NPK (150:60:60 kg NPK/ha).

Full size image

Under CF, the yield of inoculated Sahel 202 plants exceeded the yield reference by 17%, 15%, and 12% at half the recommended NK dose (F2), the recommended NK dose (F4), and the recommended NPK dose (F5), respectively. Similarly, the yield of inoculated Sahel 108 plants surpassed the yield reference by 8%, 4%, and 5% at half the recommended NK dose (F2), the recommended NK dose (F4), and the recommended NPK dose (F5), respectively (see Fig. 4).

For upland varieties, AWD-related RYL in non-inoculated plants ranged from 5 to 65% for CG 14, from 30 to 87% for NERICA 4, and from 20 to 83% for WAB56_104 (see Fig. 5). The greatest RYL reduction due to AMF inoculation (RYLRInoAMF) under AWD irrigation was observed at half the recommended NK dose (F2; 39%) for NERICA 4, at the recommended NPK dose (F5; 31%) for WAB56_104, and at half the recommended NPK dose (F3; 24%) for CG 14 (see Fig. 5, Table S4).

Fig. 5
figure 5

Relative grain yield losses of upland rice varieties inoculated (InoAMF) and non-inoculated with AMF (NinAMF) under continuous flooding (CF) and alternate wetting and drying (AWD) irrigation at different fertilizer levels (F1 = 00:00:00 kg NPK/ha; F2 = 75:00:30 kg NPK/ha; F3 = 75:30:30 kg NPK/ha; F4 = 150:00:60 kg NPK/ha; F5 = 150:60:60 kg NPK/ha). The red horizontal line (0%) represents yield reference which corresponds to the grain yield of non-inoculated plants under continuous flooding at recommended dose of NPK (150:60:60 kg NPK/ha).

Full size image

Interestingly, the yield of inoculated plants under AWD surpassed the yield reference by 1 to 8% for CG 14, and by 7 to 14% for NERICA 4, depending on the NPK fertilizer dose used. Under CF, the yield of inoculated plants exceeded the yield reference by 2 to 15% for CG 14, by 7 to 14% for NERICA 4, and by 2 to 12% for WAB56_104 (see Fig. 5).

Mycorrhizal inoculation effect on yield-related traits of irrigated and upland rice as a function of irrigation regime and fertilizer level

We then assessed the mycorrhizal inoculation effect (MIE) on yield-related traits in irrigated and upland rice and identified significant sources of variation in MIE. Irrigation regime, fertilizer level, and ecotype were significant sources of variation in MIE for total biomass (Table 1), while their interactions were not. Both irrigated and upland rice responded positively to AMF inoculation, with higher MIE in rice plants subjected to AWD irrigation compared to those under CF (Figure S3). Additionally, irrigated ecotypes exhibited a greater total biomass response to AMF inoculation. Regarding fertilizer levels, the highest MIE (0.18) was recorded at zero NPK application (F1), while the lowest MIE (0.04) was observed at the recommended NPK dose (F5; Table 1, Figure S3).

Table 1 Results from ANOVA testing the effects of fertilizer level, irrigation regime and rice ecotype on mycorrhizal inoculation effect (MIE) for yield-related traits (total biomass (Biomass), tiller number (Tiller), panicle number (Pan_Num), panicle weight (Pan_Weight), spikelet fertility (Fertility), thousand grain weight (TWG), harvest index (HI), plant height (Height), and maturity dates (Mat_Days)). Different letters in the pairwise comparisons indicate significant differences at p < 0.05.
Full size table

For tiller numbers, ANOVA revealed that only the irrigation regime and ecotype significantly influenced MIE (Table 1). AMF inoculation led to a substantial increase in tiller numbers in rice plants under AWD irrigation compared to those under CF, as shown by their respective MIE values (0.124 vs. 0.039). At the ecotype level, irrigated rice exhibited a higher MIE than upland rice. On the other hand, fertilizer level was the sole source of variation in MIE for several yield-related traits, including panicle number, spikelet fertility, and maturity dates. For the latter, a negative MIE was observed at zero NPK application (F1), while positive MIE was seen at other fertilizer levels, indicating that the combination of AMF inoculation and fertilizer application delayed maturity compared to AMF inoculation alone (Figure S3). For thousand grain weight (TGW), none of the factors or their interactions were significant sources of variation in MIE (Table 1).

Discussion

In this study, we investigated the response of six rice varieties from two ecotypes (three irrigated and three rainfed upland varieties) to AMF inoculation at five fertilizer levels under continuous flooding (CF) and alternate wetting and drying (AWD) irrigation (at 60 kPa SMP) in field conditions over two consecutive years. AMF inoculation significantly enhanced grain yield across all experimental conditions, thus showing the potential of AMF to boost rice productivity.

Our results further demonstrated that the effects of irrigation regime, fertilizer level, and AMF inoculation on grain yield were dependent on the rice ecotype. Previous studies22,29 found that upland varieties showed the best responses to inoculation, particularly in terms of grain yield, compared to irrigated and lowland ecotypes. These results align with the negative impact of continuous flooding on AMF colonization rates and extraradical hyphal density30.

AWD irrigation resulted in a significant reduction in grain yield compared to CF. This finding is consistent with the results obtained from the same study site by Djaman et al.12, which showed that rice yield declined significantly under AWD (irrigation at 60 kPa SMP). At the ecotype level, AWD induced substantial yield losses at all fertilizer levels for irrigated rice, regardless of inoculation status. In upland rice, AWD-related yield losses were observed at all fertilizer levels in non-inoculated plants, while AMF inoculation almost offset these losses at half the recommended dose of NK and at recommended doses of NK and NPK. Under AWD, AMF inoculation improved performance for all plant parameters beyond yield. Previous studies have shown that mycorrhizal plants absorb more water under drought conditions than non-mycorrhizal plants31,32,33,34,35. AMF fungi can directly transport water to plants, maintaining hydraulic continuity between the soil and the plant during water deficit36. Under AWD conditions, this contribution may be crucial for maintaining water absorption during dry cycles. Our findings suggest that AMF inoculation could be used to improve rice yield in AWD conditions.

Furthermore, in our field trials, the strongest effect of AMF inoculation was observed at low fertilizer doses. AMF inoculation reduced yield losses under low fertilization regimes and, in some cases, resulted in higher yields with lower fertilization. This positive effect was most pronounced at half the recommended dose of NK without P (F2; 75:00:30 kg NPK/ha), regardless of the irrigation regime. These results are consistent with previous reports of the positive impact of AMF inoculation on rice performance under low fertilization conditions29,37. On the other hand, it was shown that AMF inoculation coupled with P fertilization improved rice grain yield under limited irrigation conditions28. However, the latter was performed in pots in greenhouse conditions where the limited soil volume could lead to rapid exhaustion of the soil P reservoir. AMF inoculation can enhance phosphorus (P) uptake by increasing the volume of soil explored for P capture38,39,40,41. In line with this, we observed a nearly 50% increase in plant biomass in P treatments due to AMF. In addition to phosphorus, AMF can efficiently transfer nitrogen (N) to rice plants, enhancing their N nutrition42,43,44. AMF can use both inorganic forms of N, such as NH4 + or nitrate, and organic forms like amino acids45,46. Altogether, our results indicate that AMF inoculation could be used to maintain yield with reduced fertilizer application in rice agrosystems.

Interestingly, we found that the effect of inoculation on yield was variety-dependent, suggesting a genetic control that could be a new target for breeding programs. AMF inoculation significantly improved the agronomic and productive performance of the Sahel 108, Sahel 202, NERICA 4, and GC14 varieties under AWD conditions.

In conclusion, our study showed that both the irrigation system and fertilization level modulated the effects of AMF, regardless of rice ecotype. Our findings suggest that AMF inoculation could significantly reduce irrigation and fertilization requirements without compromising productivity. This strategy could be leveraged to create a more profitable and sustainable cropping system. It would be beneficial to identify genetic traits that promote AM symbiotic efficiency and integrate them into the genetic backgrounds of rice varieties. The effect of AMF inoculation in these low-input conditions was variety-dependent, suggesting that it is at least partially under genetic control and could be targeted by breeders to further optimize this agronomic system.

Methods

Site description

Field trials were conducted at the AfricaRice Saint-Louis station in Fanaye (Senegal; 16° 32′ 05″ N, 15° 11′ 01″ W) for two consecutive dry seasons (February to July). Fanaye is in the middle of the Senegal River valley, approximately 240 km inland. The climate is characterized by an irregular wet season (about 200 mm rainfall) from the end of July to the end of October, a cold dry season from November to February and a hot dry season from March to June. In the hot dry season, there are many cycles of warm and dry dust-laden winds. The temperature is very high and often exceeds 40 °C with a higher temperature amplitude of 21 °C.

AMF strains and inoculum production

The AM fungi used in this study were Glomus aggregatum Schenck & Smith (DAOM 227128, National Mycological Herbarium, Ottawa, Canada) and Rhizophagus irregularis Walker & Schüßler (formerly called Glomus intraradices DAOM 197198, Krüger et al.47). The AMF inoculum was prepared as described in Diedhiou et al.22. Briefly, the AMF strains were propagated separately in a greenhouse using maize plants (Zea mays) and sterilized (2 × 2 h at 180 °C) soil from Sangalkam (Senegal). This soil is composed of 88.8% sand, 5.8% silt, 5.4% clay, 0.6% organic matter, 0.3% total C, 0.02% total N, 333.5 ppm total K and 41.4 ppm total P. The maize roots and substrate were collected after three months of propagation to evaluate spore density in the soil and root length colonized by the AMF as described in Diedhiou et al.22. Briefly, roots were harvested, thoroughly washed with tap water and large lateral roots were collected, bleached in KOH (10%) at 80 °C during 30 min, and stained with trypan blue (0.05%) at 80 °C during 35 min (adapted from48). AMF spores were extracted from culture substrate (soil) by wet sieving and sucrose density gradient centrifugation49 to determine their density. For each strain, spores were counted under a dissecting microscope using up to 40-fold magnification50. For each AMF strain the colonized maize roots were cut into fragments of about 1 cm and perfectly homogenized to the culture substrate containing spores to form the AMF inoculum.

Field experimental design and plant material

Six rice varieties, among which 4 O. sativa (Sahel 108, Sahel 202, IR 64 and WAB 56-104), one O. glaberrima (CG 14) and one interspecific variety (NERICA 4; Table S1), were tested at five NPK fertilizer levels and two irrigation regimes (continuous flooding and AWD). In the continuous flooding (CF) regime, the water level was maintained between 5 and 10 cm from transplanting to harvesting. For AWD, the water level was maintained between 3 and 5 cm for the first ten days after plantation. From 11 days after planting to the beginning of the reproductive phase (with O. sativa Sahel 202 as the reference variety), water was applied to the field when a pressure monitor (tensiometer) probe located 20 cm deep in the soil indicated 60 kPa soil matric potential (SMP) in the center of the plot. During the reproductive phase, the water table was then continuously maintained at 10 cm.

A split-split-plot design was used in each irrigation regime with three replications. The main factor was the inorganic fertilization (NH3-P2O3-K2O) with five levels: F1—no fertilization (00:00:00 kg NPK/ha); F2—half the recommended dose of NK (75:00:30 kg NPK/ha); F3—half the recommended dose of NPK (75:30:30 kg NPK/ha); F4—recommended dose of NK (150:00:60 kg NPK/ha); F5—recommended dose of NPK (150:60:60 kg NPK/ha) in Senegal. The fertilizers were applied as follows: 40% of total amount of N, 100% P and 100% K were applied at vegetative stage (21 days after transplanting), 40% N at panicle initiation and 20% at booting stage. At each fertilizer application, the water level was reduced to 3 cm. The second factor was AMF inoculation with two treatments: inoculated (InoAMF) and non-inoculated (NinAMF). Inoculation was performed with a mixed AMF inoculant (60% R. irregularis (Ri) + 40% G. aggregatum (Ga)). Before germination, rice seeds were maintained at 48 °C for 12 h to break dormancy. Seeds were then surface sterilized with bleach 8% (4 min) and ethanol 96° (1 min) and rinsed three times with sterile demineralized water. Inoculation was performed in the nursery as follows: 2 pre-germinated seeds were planted in each hole of multi-pot plates filled with a mixture of 15 g of Ga + Ri inoculum and 27 g of sterilized Fanaye soil. For the treatments without AMF, 42 g of sterilized Fanaye soil were used per hole. 18 day-olds inoculated or non-inoculated seedlings were transplanted in the field. Prior to transplantation, some inoculated and non-inoculated seedlings were sampled to determine the root colonization by AMF. AMF structures were observed in roots of all inoculated seedlings, while the non-inoculated plants were not mycorrhized. The third factor was the 6 varieties belonging to two ecotypes (irrigated and upland). 180 plots (6 varieties × 2 inoculations × 5 fertilizer levels × 3 replicates) of 0.64 m2 each were set up in each irrigation regime (CF and AWD). In each plot, 16 seedlings were transplanted with 20 × 20 cm spacing and one plant per hill.

Assessment of yield and yield-related traits

Six plants were selected in the center of each subplot and the following traits were assessed: (1) 50% flowering time (when 50% of panicles had flowered); (2) maturity time (when 80% of grains had lost green color); (3) plant height at maturity (between the soil level and the top of panicle); (4) number of mature tillers; (5) yield components (number of panicles per plant, number of filled and empty spikelet per panicle to determine spikelet fertility, weight of 1000 grain, total shoot biomass, harvest index defined as the ratio of grain yield to total shoot biomass); (6) grain filling duration (GFD, defined as the period between flowering and physiological maturity); (7) grain yield corrected to 14% moisture:

$${\text{Grain}}\;{\text{yield}}\;\left( {{\text{kg}}/{\text{ha}}} \right) = \left( {\frac{{wgh*\frac{100 - H\% }{{86}}}}{NHH*0.04}} \right)*10000$$
(1)

where wgh is the weight of grain from harvested hill, NHH is the number harvested hills, H% if the moisture content of the rice grain, (100-H%)/86 represents the 14% humidity correction factor and 0.04 m2 (0.2 m * 0.2 m) is the area of one hill (spacing between hill = 0.20 m).

Data analysis

A linear mixed-effects (LME) model was used to test for main effect of four factors (irrigation regime, fertilizer level, inoculation with AMF, and ecotype) and their interactions (all two-way interactions) on grain yield, using the lmer function from the lme4 package51. The LME model was conducted with rice varieties nested within ecotype and year included as a random factor. The Anova function from the car package52 was used to assess the significance of fixed effects. As we found a highly significant interaction effect between ecotype and irrigation regime and ecotype and fertilizer level for grain yield (Table S2), we performed subsequent analyses separately by ecotype (irrigated and upland). Hence, LME models were run for each ecotype by using year as a random factor to test the effects of irrigation regime, fertilizer level, inoculation with AMF, rice variety and their interactions on grain yield.

Relative yield loss (RYL) in irrigated and upland rice varieties due to the reduction in fertilizer doses and the AWD irrigation was evaluated. To calculate RYL in each rice variety, the yield of non-inoculated plants (NinAMF) in optimal conditions (at the recommended dose of NPK (150:60:60 kg NPK/ha) under continuous flooding (CF)) was taken as yield reference. For each treatment, the following formula was used to estimate the RYL53:

$$RYL\left( \% \right) = \frac{Yield measured - Yield reference}{{Yield reference}}*100$$
(2)

We further assessed yield loss reduction due to rice inoculation with AMF (RYLRInoAMF) using the following formula:

$${\text{RYLR}}_{{{\text{InoAMF}}}} \left( \% \right) = {\text{RYL}}_{{{\text{NinAMF}}}} {-}{\text{RYL}}_{{{\text{InoAMF}}}}$$
(3)

where RYLInoAMF represents the RYL of inoculated plants with AMF and RYLNinAMF represents the RYL of non-inoculated plants.

For each rice variety, we further determined the mycorrhizal inoculation effect (MIE, indicating the effect of introduced AMF inoculum compared with the inherent field inoculum), for each assessed yield-related trait under each irrigation regime at each fertilizer level as follows:

$$\begin{aligned} & {\text{MIE}} = \left( {{\text{mean}}\;{\text{value}}\;{\text{of}}\;{\text{inoculated}}\;{\text{plants}} - {\text{mean}}\;{\text{value}}\;{\text{of}}\;{\text{non-inoculated}}\;{\text{plants}}} \right) \\ & \quad /{\text{mean}}\;{\text{value}}\;{\text{of}}\;{\text{inoculated}}\;{\text{plants}} \\ \end{aligned}$$
(4)

MIE varies between − 1 and 1 denoting negative and positive effect, respectively.

Unless mentioned otherwise, the mean comparisons among treatments were done by using analysis of variance (ANOVA) and post hoc Tukey’s test at p < 0.05. All statistical analyses were conducted in R v4.3.1.