The 41-year experiment commenced in 1979 at the Long-Term Fertility Experimental Station of Shenyang Agricultural University (40°48′ N, 123°33′ E), Liaoning, China, with a subhumid climate. The average annual temperature from 1979 to 2019 was 8.5°C (21.2°C from May to September), and annual precipitation was 681 mm, with 538 mm during the crop growing season. The soil is a Haplic Luvisol (FAO, 1998), which contained 48% sand, 29% silt, and 23% clay in the 0–20 cm soil layer in 1979 with 1.18 g cm^–3 bulk density, 15.9 g kg^–1 organic matter, 0.8 g kg^–1 total N, 105.5 mg kg^–1 alkali-hydrolyzable N, 6.5 mg kg^–1 Olsen-P, 97.9 mg kg^–1 available K, and pH 6.5.

Statistical analysis was performed using the SAS software package (Version 9.2). A one-way ANOVA was used to test the effect of different fertilization treatments on total N uptake and that derived from ¹⁵N-urea by soybean and maize, ¹⁵N residual in soil, and unaccounted ¹⁵N losses across the three rotation years. Significant differences between treatments were compared using the least significant difference method at the 0.05 probability level.

At the experimental site, the annual precipitation was 477, 583, and 691 mm in 2017, 2018, and 2019, with 83%, 85%, and 93% (averaged 87%) occurring during the crop growing seasons ( Figure 1 ). The trends in monthly maximum and minimum temperatures were similar across years, with maximum temperatures of 37.5, 38.4, and 36.3°C and minimum temperatures of –22.3, –27.7, and –23.1°C in 2017, 2018, and 2019, respectively.

Soybean and maize grain yields varied between treatments ( Figure 2 ). In the soybean season (2017), the combined chemical fertilizer and manure treatments significantly increased grain yield by 153% on average, relative to chemical fertilizer alone (241% at MN vs N, 103% at MNP vs NP, and 114% at MNPK vs NPK), more so at MNPK (yield being 4544 kg ha^–1). The yield trends in the two succeeding maize seasons were similar to soybean, with the highest yields at MNPK (12336 and 13868 kg ha^–1 in 2018 and 2019, respectively). The MN, MNP, and MNPK treatments had 288%, 19%, and 7% higher maize grain yields than the N, NP, and NPK in 2018, respectively, with corresponding increases of 628%, 26%, and 13% in 2019.

The combined chemical fertilizer and manure treatments significantly enhanced soybean and maize N uptake relative to the sole chemical fertilizer treatments ( Figure 3 ). In 2017, soybean total N uptake ranged from 44.4–398.5 kg N ha^–1, of which 5.3–12.9 kg N ha^–1 came from ¹⁵N-urea, with the maximum amounts at MNPK. The ratio of plant N derived from ¹⁵N-urea to total N uptake decreased from 12.0% at N to 3.2% at MNPK. In the two subsequent maize seasons, plant N uptake ranged from 130.1–405.1 kg N ha^–1 in 2018 and 105.4–524.2 kg N ha^–1 in 2019, of which 3.3–4.3 and 0.8–1.7 kg N ha^–1 came from soil residual ¹⁵N from the soybean season. Maize N uptake from the soil residual ¹⁵N contributed 1.0–2.5% and 0.3–0.8% to total N uptake in 2018 and 2019, respectively. As with the soybean season, the MNPK treatment had the highest values, which significantly differed from the other treatments.

The distribution of ¹⁵N uptake in plant organs among treatments followed similar trends each season ( Figure 4 ). In the soybean season, ¹⁵N uptake was partitioned mainly to grain, with the highest at MNPK (9.3 kg N ha^–1), and ¹⁵NHI ranged from 66.1–75.2% across all treatments ( Figure 4A ). The combined chemical fertilizer and manure treatments significantly increased ¹⁵N uptake in grain by 104.2% on average, with no significant effect on ¹⁵NHI, relative to the chemical fertilizer only treatments. The distribution of ¹⁵N-urea in pods, straw, and roots accounted for 13.2%, 10.5%, and 4.6% of the total ¹⁵N uptake on average, with no significant differences between treatments. In the first subsequent maize season (2018), the NPK and MNPK recovered 1.88 and 1.86 kg ha^–1 soil residual ¹⁵N from the soybean season for distribution to grain, with ¹⁵NHI of 46.3% and 43.0%, respectively. The ratios of ¹⁵N uptake in cobs, bracts, straw, and roots to total uptake were 13.0%, 2.3%, 34.8%, and 9.7%, respectively. In the second subsequent maize season (2019), the recovery of soil residual ¹⁵N declined but followed a similar trend to 2018. The average ¹⁵NHI was 46.6%, with ratios ¹⁵N uptake in cobs, bracts, straw, and roots to total uptake of 3.9%, 4.1%, 37.2%, and 8.3%, respectively.

Of the 30 kg N ha^{–1 15}N-urea applied in the soybean season, 13.4–20.0 kg N ha^–1 remained in the 0–40 cm soil layer at harvest, with 83.2–91.8% in the top 20 cm ( Figure 5A ). Chemical fertilizer alone resulted in more ¹⁵N residual in the top 40 cm soil with no significant difference compared to treatments with manure. The residual ¹⁵N-urea was present mainly as organic N, accounting for 58.6% of the added ¹⁵N-urea in the chemical fertilizer only treatments and 45.4% in the chemical fertilizer combined with manure treatments, compared with 4.9% and 3.0% for mineral N, respectively. At maize harvest in the first subsequent year, the top 40 cm soil contained 8.8–13.6 kg N ha^–1 derived from ¹⁵N-urea, with 69.1–76.9% in the 0–20 cm soil ( Figure 5B ). Similarly, 31.6% and 41.2% of the preceding applied ¹⁵N fertilizer remained in organic form in the 0–40 cm soil profile with and without manure, respectively, compared with 0.5% and 0.6% for mineral N. In the second succeeding maize year, the top 40 cm soil contained 6.7–12.2 kg N ha^–1 derived from ¹⁵N-urea, with a significant decrease with manure application ( Figure 5C ), with 74.3–91.9% distributed in the top 20 cm layer. The ratios of organic N to applied ¹⁵N-urea were 25.2% and 36.7% with and without manure, compared with 0.2% and 0.2% for mineral N.

The chemical fertilizer combined with manure treatments significantly increased soybean N uptake from applied ¹⁵N-urea by 94.6% compared with the chemical fertilizer only treatments ( Table 1 ). In the soybean season, the fertilizer ¹⁵N recovery was 28.8% on average. In the two succeeding maize seasons, 3.3–4.3 kg N ha^–1 and 0.8–1.7 kg N ha^–1 of the soil residual ¹⁵N from the soybean season was recovered, accounting for 10.9–14.4% (12.6% average) and 2.6–5.8% (4.1% average) of theapplied ¹⁵N-urea. Across the soybean–maize–maize rotation system, 31.2–63.1% of the initial applied fertilizer ¹⁵N was recovered, with 21.9–40.5% remaining in the 0–40 cm soil and 20.2–31.3% in the top 20 cm soil. During the 3-year rotation, 14.6–29.9% of the applied ¹⁵N-urea was unaccounted for (including ¹⁵N losses), which decreased by 26.9% with manure addition. The highest recovery rate (63.1%) and lowest residual (22.3%) and unaccounted (14.6%) rates were found at MNPK treatment.

The combined application of chemical and organic fertilizers increased crop yield compared to chemical fertilizer alone ( Figure 2 ), as reported elsewhere (Chivenge et al., 2011; Doan et al., 2015; Zhang et al., 2018). In our study, grain yields significantly increased at treatments with manure in the soybean season, despite not being applied since 1992, which could be due to the increased soil organic matter and nutrient availability ( Table S2 ). Similarly, a ¹⁵N study in Northwest China reported that soils supplied with MNPK had higher wheat grain yields than those supplied with NPK for the same amount of N applied (Liang et al., 2013), as did another study in the Tai Lake region (Pan et al., 2009). In maize seasons, grain yields were more responsive to chemical fertilizer plus manure than chemical fertilizer alone with reduced difference between manure treatments due to the extra nutrient addition. Besides, after long-term fertilization with chemical fertilizer alone, the soil had lower buffering capacity of nutrients and water conservation and supply than that with manure application ( Table S2 ) (Diacono and Montemurro, 2010; Oldfield et al., 2019). However, another study reported no significant effect on maize yield for the same N rate, replacing 50% synthetic N fertilizer with manure in Northeast China (Chen et al., 2014b), and wheat yields even decreased by 14–20% in North China Plain (Sun et al., 2013).

Long-term addition with chemical fertilizer and manure benefited soybean and maize total N uptake and that derived from ¹⁵N-urea, particularly at MNPK ( Figure 3 ), but did not alter its proportional distribution in plant organs. Plant N uptake from ¹⁵N fertilizer decreased in the two successive maize seasons, and totally accounted for 31.2–63.1% of the applied N in 2017, indicating that most of the ¹⁵N-urea remained in the soil or was lost. Previous studies reported comparable results (Wang et al., 2016a; Li et al., 2018; ). In contrast, the MNPK treatment had lower ¹⁵NHI than the NPK treatment due to the dilution effect, as the significant increase in total ¹⁵N uptake with manure decreased the ratio of N distributed in grain.

Co-application with manure significantly increased fertilizer ¹⁵N recovery by 94.6% compared with chemical fertilizer alone in the first soybean season, attributed to the enhancement in soil organic matter (22.4 g kg^–1 with manure and 17.2 g kg^–1 with chemical fertilizer) ( Table S2 ), similar to previous studies (Liang et al., 2019; Quan et al., 2021). However, less than 50% of the ¹⁵N-urea was recovered in plant, with the highest at MNPK (42.9%). Similarly, another study searching numerous papers with ¹⁵N labeled fertilizer reported ¹⁵N recovery rates of 41% for maize, 32% for rice, and 37% for crops with small grains (Yan et al., 2019). Rocha et al. (2019) also reported that the seasonal ¹⁵N recovery for maize was 34%, with 46% residual in soil in the maize–pasture rotation system. It indicated that large amount of fertilizer N was residual in soil, a crucial N source for subsequent crops (Yang et al., 2011; Guo et al., 2021; ). Therefore, it is important to not only consider seasonal N fertilizer efficiency but also the recovery of soil residual N derived from fertilizer in succeeding seasons.

For crop N uptake, 56–64% was derived from soil, with some converted from residual fertilizer N, biotic fixation, and N deposition (Gardner and Drinkwater, 2009; Dourado-Neto et al., 2010; ). Due to the overuse of N fertilizer in China, residual fertilizer N could be the largest contributor to the soil N pool (Zhang et al., 2015). At soybean harvest in 2017, 44.7% (MNP) to 66.7% (N) of the applied ¹⁵N-urea remained in the 0–40 cm soil layer ( Figure 5A ), comparable to irrigated (45–60%) and rainfed (49.8%) maize studies in China (Yang et al., 2011; Wang et al., 2016b; ). While, Guo et al. (2021) reported a lower fertilizer N residual rate (26.1–32.1%) due to the higher N use efficiency of maize (40–50%), and 30% in Shi et al. (2012) at an N rate of 210 kg N ha^–1 with large N losses. In our study, the residual fertilizer ¹⁵N remained mainly in the top 20 cm soil though frequent rainfall occurred during the soybean and maize growing seasons. It might be attributed to the microbial immobilization and some adsorption of mineral N forms (presumably NH 4 + ) in exchangeable sites of the topsoil, in line with other studies (Liu et al., 2015; Karwat et al., 2017; Hu et al., 2019; ). Inversely, higher fertilizer ¹⁵N accumulated in the 20–40 cm soil layer in North Central China, indicating that more N leaching occurred than in our study (Quan et al., 2020). In the second maize season in 2019, the 20–40 cm soil had more mineral N than the top 20 cm, attributed to the high precipitation in August (378.6 mm), which might have increased mineral N leaching. Like Li et al. (2021a), the combined use of manure and chemical fertilizer reduced the soil mineral N amount, potentially decreasing N leaching. Thus, fertilizer type, application method, soil properties, and climate influenced fertilizer N residual in soil. After three years, 21.9–40.5% of the labeled ¹⁵N fertilizer remained in the 0–40 cm soil depth, with the lower proportions at manure treatments, and was primarily in organic form, consistent with earlier studies (Stevens et al., 2005; Li et al., 2015; Van Meter et al., 2016). Subsequent crops can use the organic residual N through remineralization.

Similar to other studies (Smith and Chalk, 2018; Yan et al., 2019), there was an incremental reduction in the recovery of soil residual ¹⁵N by successive crops. In the two maize seasons (2018 and 2019), 10.9–14.4% and 2.6–5.8% of the total applied ¹⁵N-urea was recovered, with the highest at MNPK. The reason is that the high soil fertility and improved physical and biological properties due to the long-term fertilization with chemical and organic fertilizers enhanced the release of soil residual organic ¹⁵N. A 60-year study of ¹⁵N tracer reported a soil residual ¹⁵N recovery of 5.4 ± 4.5% in the first subsequent season and<3% with impairing rate in the second or more successive seasons (Smith and Chalk, 2018). It indicated that the residual fertilizer N in organic form could remain in soil for a long time, which was gradually released and used by subsequent crops, with only 13 kg N ha^–1 mineralized over 28 years (Sebilo et al., 2013). Exceptions to the above results were attributed to high N fertilizer rates and soil N residues, resulting in high N recoveries in the following cropping seasons (Ditsch et al., 1993; Ju et al., 2007; Guo et al., 2021; ). Across the three-year rotation, the MNPK treatment had the highest (63.1%) fertilizer N recovery, which could be because it had the highest ¹⁵N recovery in the first soybean season, and the manure stimulated microbial activity, enhancing ¹⁵N remineralization from the residual organic N (Edmeades, 2003; Ding et al., 2017; ) and thus increasing crop ¹⁵N uptake in the two maize seasons. Therefore, fertilization with manure plays a vital role in promoting fertilizer N recovery efficiency, most likely by improving soil fertility status.

During the three-year rotation, 14.6–29.9% of the applied ¹⁵N-urea was unaccounted for, which might be related to processes of N removal to the deep soil layer and N losses, including gaseous emissions and leaching. Guo et al. (2021) reported that the proportion of unaccounted ¹⁵N fertilizer was 21.8–27.4%, mainly associated with N losses by volatilization, denitrification, and leaching as the soil residual N was monitored to 160 cm depth. Another study in dryland Northern China showed that 23% of total N losses came from ammonia volatilization (Ju, 2014) due to the high pH of the calcareous soil. In our study, ammonium ions were the main form in the soil, not free ammonia, with a pH below 7 ( Table S2 ) (Overrein and Moe, 1967; Hayashi et al., 2011; ), thus ammonia volatilization could not be the dominant unaccounted ¹⁵N loss. Additionally, over 80% of the annual precipitation occurred during crop growing seasons in our study ( Figure 1 ), which would result in more N losses by leaching. A global study on maize cropping systems consistenly reported 28% of unaccounted ¹⁵N under conventional management, with N rates ranging from 100–300 kg N ha^–1 (Quan et al., 2021), whilea higher proportion of 37% was revealed in a meta-analysis with a larger range of N rates (average 114 kg N ha^–1) (Gardner and Drinkwater, 2009). Application of manuredecreased the unaccounted ¹⁵N by 26.9% compared to the chemical fertilizer only treatments, mainly attributed to the long-term fertilization with manure plus inorganic fertilizer increasing soil C content, enhancing N immobilization and retaining organic N (Wang et al., 2019). In addition, the lower unaccounted ¹⁵N with manure was consistent with the higher total N recovery observed across the three years. Likewise, compost and straw application decreased soil mineral N contents, even at 80–100 cm depth, diminishing the risk of N leaching (Zhang et al., 2016). Therefore, raising the awareness of the benefits of manure application is crucial for increasing production and decreasing fertilizer N losses and thus environmental risks.