Introduction

The overuse of nitrogen (N) fertilizers in croplands has been a global concern for decades1,2,3 with losses still exceeding recovery in harvested materials (global N use efficiency of 55%)4. A large amount of non-harvested N is thought to be redistributed and resides in earth critical zones (CZs)5,6,7,8, a vital thin skin of the planet open to the complex interactions among air, plants and other organisms, water in the saturated zone, and solid minerals in soils to bedrock9,10. These excess N fertilizers (approximately 90% applied in their reduced form such as ammonium N, i.e., NH4+-N) applied to croplands alter the global N cycle11,12,13, leading to a series of environmental problems, including soil acidification14, water quality degradation15, formation of harmful algal blooms16,17 and hypoxic water zones18, and increased emissions of greenhouse gases to the atmosphere19,20,21. The CZs are vital in processing and attenuating N overloads in global croplands yet, to date, out of a total of 257 global critical zone observatories (CZOs), only ~30% of them were explicitly stated to be established in or near croplands (Supplementary Fig. 1 and Supplementary Table 1), with even fewer examining how the structure of cropland CZs regulates the productive use of N fertilizers8. Particularly, vertical nitrification patterns resulting from the magnitude and vertical location of N fertilizer transformation zones may be valuable indicators of the capacity of the croplands CZs regulating these overloads at both local and global scales22,23.

Nitrification has heterotrophic and autotrophic pathways13,24, with the primary substrates being NH4+-N and organic N, and the primary product being nitrate N (NO3--N)23. In non-cropland ecosystems, heterotrophic nitrification predominates and is tightly coupled to the soil total N (TN, primarily as organic N)22,24. We demonstrate that soil TN exhibits an exponential decline with depth across global non-cropland ecosystems (Fig. 1a, b). This gradient dictates a parallel decrease in nitrification rates, a pattern well-established by site-level studies25,26,27,28. In croplands, however, excessive N fertilizer inputs shift the process to autotrophic dominance13,22. Our synthesis of global cropland profiles reveals a stable deviation: a persistent TN bulge at approximately 0.6 m soil depth across time (p < 0.001) and strongly related to historical N input (p < 0.01) (Fig. 1a, c, d, e). Given the classical exponential decay of soil organic carbon (Supplementary Fig. 2), this TN bulge signals a profound N-transformation driven by inorganic fertilizers rather than by the mineralization of soil organic components. The shallow bulge’s position is well above the peak global water table depth (WTD, 2–7 m), within an oxygenated zone that is optimal for nitrification29,30. We therefore propose that this feature constitutes a previously overlooked Enhanced Nitrification Layer (ENL). This ENL could function as a biogeochemical engine, converting subsurface NH4+-N into mobile NO3--N, thereby not only sustaining the observed TN bulge but also directly driving the soil acidification that reduces crop yield and the NO3--N leaching that pollutes groundwater, both processes that undermine agroecosystem sustainability. The identification of this ENL thereby serves as a precise and manageable subsurface target, expanding the paradigm for sustainable N management to integrate the critical depth of peak transformation alongside surface measures.

Fig. 1: Vertical patterns of soil total nitrogen (TN) in global croplands and non-croplands.
Fig. 1: Vertical patterns of soil total nitrogen (TN) in global croplands and non-croplands.The alternative text for this image may have been generated using AI.
Full size image

a Site distribution of soil TN in both global croplands and non-croplands. Global soil TN data are obtained from the existing database of the World Soil Information Service (WoSIS) and extensively compiled from the site-level investigations of croplands in the literature (Supplementary Tables 6 and 7). Sites from the WoSIS are classified into cropland and non-cropland types based on the global distribution of land uses at different periods (see Methods). b, c The exponential decline and peak patterns of soil TN in global non-croplands and croplands, respectively. The data are shown with a mean ± 95% confidence interval. d, e Soil TN change at 0.6 m depth compared to the mean of the adjacent five soil layers (the main subsoil depths within 1 m where nitrification occurs, based on refs. 48,49,50,51) across time, as well as its relationship with global mean nitrogen input, respectively.

Results and discussion

Potential location of cropland ENL

Given the scarcity of direct vertical nitrification rate measurements in previous studies, which predominantly focus on surface soils22,23, we thus inferred the potential depth of the ENL from the vertical distributions of nitrification primary substrate (NH4+-N, n = 1590) and product (NO3--N, n = 2945) in global croplands (Fig. 2). Our analysis revealed a distinct subsurface peak in soil NH4+-N at 0.30–0.75 m depth, whereas the peak soil NO3--N occurred deeper at 0.45–0.90 m. The results are highly consistent across different climate zones (Supplementary Fig. 3). Low inorganic N concentrations in the upper layers reflect N transformations and crop uptake in the root zone12, which together deplete upper soil NH4+-N and NO3--N pools and amplify the relative prominence of the subsurface peaks. This spatial disconnect, with the NH4+-N peak consistently above the NO3--N peak, suggests that while some downward movement of upper‑formed NO3--N inevitably contributes to subsurface NO3--N pools, the observed depth offset requires an additional source of nitrate at depth, most plausibly in‑situ nitrification converting NH4+-N to NO3--N along the flow path29,31. Critically, the averaged bulges of nitrification products NO3--N were found to be twice as high as the averaged nitrification substrate NH4+-N, although fertilization in croplands predominantly takes the form of NH4+-N12. This disproportionate accumulation of product relative to substrate cannot be explained by leaching alone, which would redistribute rather than amplify NO3--N. Instead, it points to active N transformation at a specific depth. The inference is further supported by the consistency of these patterns with established correlations between nitrification rates and substrate (R2 = 0.15, p < 0.001) and product (R2 = 0.32, p < 0.001) pools in global croplands22. We, therefore, infer that the ENL represents a subsurface depth interval (0.3–0.9 m) between these inorganic N peaks, where nitrification activity is enhanced relative to adjacent soil layers. Within this interval, approximately 0.6 m serves as a global reference point that coincides with the persistent TN bulge (Fig. 1c), providing a consistent basis for linking subsurface nitrification to long-term N accumulation across croplands.

Fig. 2: Vertical patterns of soil inorganic nitrogen in global croplands.
Fig. 2: Vertical patterns of soil inorganic nitrogen in global croplands.The alternative text for this image may have been generated using AI.
Full size image

a, b Vertical profiles of soil ammonia nitrogen (NH4+-N, nitrification substrate) and nitrate nitrogen (NO3--N, nitrification product) in global croplands, respectively. The coloured dot line represents the mean values (±95% confidence interval) at each soil depth, with the adjacent numbers indicating the sample sizes. The gray dots represent the distribution of the sample values at varying depths. The site locations of soil NH4+-N and NO3--N in global croplands are presented in the world map at the lower right corner. Soil NH4+-N and NO3--N of varying depths are compiled from literature and are detailed in Supplementary Tables 6 and 7. ENL denotes the enhanced nitrification layer.

Subsurface drives of ENL formation

Having delineated the potential depth of the cropland ENL, we further evaluated whether subsurface conditions could support active nitrification at this horizon. Nitrification process is usually regulated by substrate availability, oxygen content, and the abundance of nitrifiers under favorable hydrological conditions22,23,29. We thus examined these key attributes in the cropland subsurface, namely root extension and decomposition (indicating substrate conveyor and supply), soil texture (influencing oxygen supply), and nitrifying microbe abundance (constituting the nitrification engine), which can modulate the key conditions of nitrification.

First, root extension and subsequent decomposition create voids and channels above the WTD, and facilitate the transfer of nutrients downward from the topsoil32. Thus, we used the maximum rooting depth (MRD), a proxy of root extension for crops, as the attribute to determine the NH4+-N supply to the ENL. We analyzed the MRD of global plants (n = 2603) and found that it predominantly falls within the subsoils of 0.6–0.8 m depth (Supplementary Fig. 4), closely aligning with the depth of the ENL and TN bulge. Further supporting this connection, comparative analysis of soil NH4+-N profiles reveals that deep-rooted crops (primarily fruit and tea trees) exhibit significantly deeper NH4+-N accumulation peaks than shallow-rooted species (primarily cereals and vegetables) (Fig. 3). These findings indicate that root channels facilitate the vertical transport of applied fertilizers into the ENL32, while simultaneous decomposition of root debris and exudates contributes additional NH4+-N33. Together, these root-mediated processes sustain the substrate supply required to maintain enhanced nitrification within the ENL.

Fig. 3: Formation mechanism of the Enhanced Nitrification Layer (ENL) in global croplands.
Fig. 3: Formation mechanism of the Enhanced Nitrification Layer (ENL) in global croplands.The alternative text for this image may have been generated using AI.
Full size image

a, b Conceptual and data-based views of the ENL formation mechanism, respectively. In a and b, subfigures 1 to 4 represent the four attributes in the formation of ENL, and subfigure 5 shows how ENL leads to elevated nitrate nitrogen (NO3--N). The data are shown with a mean ± 95% confidence interval. The data on soil ammonium nitrogen (NH4+-N), nitrifying microbes, and soil NO3--N in global croplands are extensively compiled from the literature and can be found in Supplementary Tables 6 and 7. The data on soil sand content in croplands are extracted from the World Soil Information Service (WoSIS) database based on the global distribution of land uses at different periods (see Methods). c Site locations of data used in subfigures 1-3 and 5 of b. Subfigure 4 regarding rainfall-activated substrate transport and ENL formation can be found in Fig. 4 and Supplementary Fig. 13. The amoA-AOA and amoA-AOB represent the ammonia-oxidizing archaea and ammonia-oxidizing bacteria, respectively.

Second, soil texture creates an oxygenated microenvironment34 that supports nitrification within the ENL. We analyzed sand content as a proxy for soil porosity and observed a relatively high sand layer that showed a consistent increase in sand content from 0.45 m depth downward until the 1.2 m depth (Fig. 3). This phenomenon is consistent across different climate zones, except in arid areas (Supplementary Fig. 5). Further supporting this, decadal soil moisture data reveal a distinct relatively dry layer within the ENL depth range (Supplementary Fig. 6). This soil moisture pattern is likely driven by the relatively high sand content in this layer (low water-holding capacity), coupled with root uptake, water table fluctuations, agricultural drainage practices, and anthropogenic climate change30,35,36,37,38. Especially under the future scenario, the subsoil drought (28–100 cm) will continue to increase36. The co-occurrence of elevated sand content and reduced soil moisture within the ENL implies enhanced oxygen availability, thereby fostering favorable conditions for nitrification.

Third, nitrifying microorganisms actively colonize the ENL, providing the biological capacity for enhanced ammonium oxidation29. Our analysis of global nitrifying functional gene profiles (n = 259) shows that ammonia-oxidizing archaea (amoA-AOA) are present throughout the ENL depth interval (0.3–0.9 m), with highest abundances in the root zone (approximately 0.4 m) and a distinct secondary enrichment at approximately 0.6 m depth (Fig. 3). Sensitivity analysis excluding paddy systems confirms this pattern, showing that the mean amoA-AOA abundance at approximately 0.6 m remains elevated, despite wider confidence intervals reflecting greater subsurface heterogeneity (Supplementary Fig. 7). This consistency confirms that the ENL represents a robust subsurface feature independent of potential confounding effects from flooded systems. These amoA-AOA can acclimate to low pH and soil NH4+-N levels, and drive nitrification in a wide range of acidic cultivated soils39. The elevated abundance of these specialized nitrifiers provides a biological basis for the enhanced nitrification activity observed in this subsurface layer.

Case studies for ENL formation

The intrinsic drivers of ENL formation were explored through the comprehensive analysis of subsurface environments in global croplands. However, verifying these mechanisms required field investigations at the CZO scale. To this end, we performed a 3-year measurement on N-components in different CZ sections (regolith and saturated sections) in the Fengpu Stream Watershed (FSW) CZO, an agricultural system characterized by high fertilizer inputs (Supplementary Figs. 8–12 and Supplementary Tables 2–5).

Integrated analysis of the FSW-CZO identified two key subsurface characteristics: (1) an MRD of 0.6–0.8 m (Fig. 4a and Supplementary Fig. 10), and (2) a relatively high sand layer (0.6–0.8 m) positioned above the WTD fluctuation, as evidenced by high-resolution (daily) and monthly WTD monitoring, together with soil texture investigation (Fig. 4a and Supplementary Table 2). These structural characteristics closely align with the depth of the global ENL. In particular, elevated NH4+-N and NO3--N in both soil and soil water within the ENL during wet seasons (Fig. 4b, c) suggest that soil N is effectively transported to the relatively high sand layer (ENL depth) through root-mediated pathways under rainfall infiltration, where it undergoes nitrification. Conversely, exponential declines of NH4+-N and NO3--N above WTD during dry seasons indicate that soil NH4+-N derived from fertilizer and organic matter mineralization is not effectively loaded into the ENL to fuel nitrification. These patterns are consistent with the first ENL formation attribute (root-mediated substrate transport) and the second attribute (oxygenated microenvironments maintained by soil texture), while highlighting the importance of hydrological controls.

Fig. 4: Comprehensive observations in the Fengpu stream watershed critical zone observatory (FSW-CZO), southeastern China.
Fig. 4: Comprehensive observations in the Fengpu stream watershed critical zone observatory (FSW-CZO), southeastern China.The alternative text for this image may have been generated using AI.
Full size image

a The position of the relatively high sand layer, the maximum rooting depth (MRD), and the dynamic changes of regional rainfall, water table depth (WTD), river flow, and baseflow at a daily scale in the FSW-CZO. A detailed FSW-CZO overview can be found in Supplementary Fig. 8. Field images for the distribution of crop roots and WTD in the FSW-CZO can be found in Supplementary Fig. 10. b, c Vertical changes of soil and soil water nitrogen at a seasonal scale, respectively. d, e Vertical changes of soil nitrogen, nitrification rate, and gene abundance after two rainfall events in the FSW-CZO. Rainfall indicates the total rainfall during the 20 days before sampling. WTD was recorded on the day of sampling. The amoA-AOA and amoA-AOB represent the gene abundance of ammonia-oxidising archaea and ammonia-oxidising bacteria, respectively. n.d. denotes not detected. ENL denotes the enhanced nitrification layer.

We further examined how rainfall intensity regulates ENL activation. Following light rainfall, soil N components, nitrification rates, and amoA-AOA abundance all exhibited exponential declines with depth (Fig. 4d, e). In contrast, heavy rainfall events triggered substantial increases in substrate NH4+-N, nitrification rates, amoA-AOA abundance, and product NO3--N within the ENL. This coupled response demonstrates that the ENL is an actively responding biogeochemical reactor, rather than a passive recipient of leached NO3--N. These results from the FSW-CZO demonstrate consistency with the first, second, and third ENL formation attributes (root-mediated transport, oxygenated microenvironments, and nitrifier communities), while emphasizing the fourth attribute (hydrological activation). Similar patterns were observed in the cropland-dominated Jianghan Plain CZO40, where rainfall drives subsoil NH4+-N accumulation and ENL formation in the relatively high sand layer (Supplementary Fig. 13).

Together, the results from our case studies are consistent with the ENL formation pathways identified in the global analysis, corroborating the inferred nitrification mechanisms. While strong nitrification also occurs in the root zone (e.g., elevated amoA-AOA at ~0.4 m depth; Fig. 3b), the exact depth of the ENL varies with local environmental conditions, as illustrated by the ~0.7 m ENL in the FSW-CZO, and the ~0.5 m ENL in the Jianghan Plain CZO, both of which lie within the defined ENL depth interval (0.3–0.9 m). Our findings are derived primarily from intensively managed croplands in humid and sub-humid regions with neutral to acidic soils, as represented by two case study sites in southeastern China and central China, respectively. To test the generality of the ENL concept and develop region-specific management strategies, further experimentation is needed across a broader range of agricultural systems, including semi-arid to arid regions with alkaline soils, etc.

Implications for agroecosystem sustainability

The ENL constitutes a critical yet overlooked threat to agroecosystem sustainability by driving interconnected environmental degradation and N loss via two synergistic pathways: Firstly, the ENL accelerates the subsoil acidification trend (Fig. 5). Intensive N fertilization increases nitrification, a process that generates hydrogen ions29,39, thereby driving soil acidification and ultimately reducing crop yields41,42. Soil pH data (1961–2015, n = 104,759) establish that a pervasive acidification trend in global cropland soils is significantly more acute within the ENL (ranging from 0.3 to 0.9 m), particularly at the depth of the TN peak (approximately 0.6 m) after the 1980s. This sharp and localized pH decline is strongly associated with soil TN content (p < 0.01). This finding highlights the urgent need for targeted management to alleviate acidification in this layer.

Fig. 5: Cropland soil pH declines at the depth of the enhanced nitrification layer.
Fig. 5: Cropland soil pH declines at the depth of the enhanced nitrification layer.The alternative text for this image may have been generated using AI.
Full size image

a Vertical patterns of soil pH in global croplands at different periods. The data are shown with a mean ± 95% confidence interval. The data on soil pH in croplands are extracted from the World Soil Information Service (WoSIS) database based on the global distribution of croplands at different periods (see Methods). The site locations of soil pH (1961–2015) in global croplands are presented in the world map at the top right corner. b Soil pH changes at 0.6 m depth compared to the mean of the adjacent five soil layers (the main subsoil depths within 1 m where nitrification occurs, based on refs. 48,49,50,51) over time. c The relationship between soil pH and soil TN at 0.6 m depth.

Secondly, the ENL acts as a biogeochemical engine for soil N leaching and groundwater NO3--N pollution. We observed high NO3--N (nitrification product) at the ENL depth, regardless of climate zone or crop root system length (Fig. 3b and Supplementary Fig. 10). However, below the ENL, both the sand content and the root distribution decline with soil depth (Fig. 3b and Supplementary Figs. 4 and 5), along with a decline in saturated soil hydraulic conductivity43. These conditions will favor the production of NO3--N from NH4+-N after fertilizer applications in the ENL, and cause a consistent release of NO3--N to groundwater. Our 3-year field observations at the FSW-CZO are consistent with this mechanism, showing a low NH4+-N concentration in groundwater with no clear response to fertilization, but significantly high NO3--N levels that exhibit a marked delayed response to fertilization (Supplementary Fig. 11). This demonstrates that the ENL intercepts and transforms most NH4+-N into mobile NO3--N, promoting ongoing NO3--N leaching and groundwater contamination5,6,7,31,44.

In the past decades, various cropland management practices have been employed to control nitrification, enhance N use efficiency, and minimize environmental pollution45. Optimizing fertilizer rate, timing, and placement at the soil surface remains the cornerstone of sustainable N management45,46, as it directly reduces the overloaded N entering the subsoil system. However, a substantial fraction of residual N bypasses surface retention via root channels, preferential flow, or after sorption capacity is exceeded, and is transported to the subsurface, where it fuels the ENL. The identification of this biogeochemical engine thus addresses a long‑standing ignored pathway, calling for an integrated two‑pronged strategy: continued improvement of surface-based practices to minimize loading, combined with depth-specific interventions (e.g., targeted placement of nitrification inhibitors, irrigation scheduling to limit hydrological activation) that intercept N at the zone of peak transformation46,47. Integrating both surface management and ENL-targeted interventions thus offers a more complete pathway for improving N use efficiency and safeguarding agroecosystem sustainability.

Methods

Global database

We integrated a comprehensive database of vertical soil N profiles from global croplands to explore the nitrification pattern. Our literature search on the Web of Science, using the keywords “cropland”, “agriculture”, “nitrogen”, “soil depth”, “soil profile”, and “vertical soil”, identified more than 2000 published papers (as of the end of 2022). Finally, a total of 107 papers were chosen according to three criteria: (i) the data on soil N originated from croplands; (ii) at least one variable related to soil N (TN, NH4+-N, NO3--N) was reported in the paper; and (iii) the soil profiles were reported with at least two layers. In summary, we extracted 14 key data categories, including the first author, published year, country, longitude, latitude, soil depth, TN, NH4+-N, NO3--N, amoA-AOA, amoA-AOB, mean annual precipitation, mean annual temperature, and crop type. Given that soil nitrification rates are typically low (<0.1 mg kg⁻¹ d⁻¹) below 1 m depth48,49,50,51 and the global WTD distribution peak (2–7 m) limits oxygen supply29,37, we collected soil N data from the 0–2 m depth. However, the hydrological and redox conditions in flooded paddy fields may limit nitrification29. During our data collection, most studies only reported the dominant crop type, including paddy, making it difficult to accurately distinguish whether sampling was conducted during the fallow period or under flooded conditions. Therefore, our current data analysis includes both dryland and a small number of paddy systems. An analysis of the vertical patterns of soil nutrients excluding paddy systems is presented in Supplementary Fig. 7. The key patterns (including the subsurface peaks in TN, NH4+-N, NO3--N, and amoA-AOA) remained consistent with those obtained using the full dataset, confirming the robustness of our findings. Detailed data and references are provided in Supplementary Tables 6 and 7, respectively.

Global distribution of plant MRD was derived from the Global Root Trait (GRooT) Database52. The vertical distribution of global soil TN, TOC, sand content, and pH was derived from the World Soil Information Service (WoSIS, December 2023) database (https://www.isric.org/explore/wosis). Global monthly soil moisture and temperature at depths of 0–0.1, 0.1–0.4, 0.4–1, and 1–2 m from 2010 to 2019 were collected from the Famine Early Warning Systems Network (FEWS NET) Land Data Assimilation System (FLDAS) products (FLDAS_NOAH01_C_GL_M; https://disc.gsfc.nasa.gov/datasets). Soil physicochemical properties data are classified into cropland and non-cropland (including forest, grassland, and shrubland) types based on the global distribution of land uses53 at different periods (Supplementary Table 8). Soil inorganic N and sand content data were categorized according to the Köppen-Geiger climate classification (1980–2016)54. Data regarding global agricultural N input were obtained from the FAOSTAT.

Fengpu stream watershed critical zone observatory (FSW-CZO)

Our FSW-CZO (117.2812°E, 24.3856°N) is located in the upper reaches of the West River of the Jiulong River, Fujian Province, southeastern China (Supplementary Fig. 8). The FSW is an upper head watershed of the West River, which covers approximately 85 km2, with regional elevation ranging from 49 to 999 m. The slope ranges from 0 to 71 degrees, and areas with a slope greater than 15 degrees account for 79%. This study area falls within a subtropical oceanic monsoon climate, with a mean annual temperature of 22.8 °C. The mean annual precipitation is 1,411 mm, mainly from March to September (wet season) with a high flow. The FSW is typically an agricultural watershed, primarily planted with pomelo, and covers 64% of the total area, with the rest as forest land (32%) and built-up land (4%).

Observation in the FSW-CZO

We conducted a three-year (2019–2021) monitoring of field management, hydro-meteorological conditions, and the dynamics of the soil and water environment in the FSW-CZO. Information on fertilization was collected based on detailed records of four farms in 2019. Since pomelo was the dominant industry in this region, there was no significant difference in field management between the years. The main stages of orchard management were summarized in Supplementary Fig. 9. Chemical fertilizer (diammonium phosphate, urea, and mixed fertilizer) and manure were the main nutrients to promote pomelo growth. The amount of fertilizer N was then converted based on the N content in each type of chemical fertilizer and manure used. Generally, approximately 1100 ± 200 kg N ha-1 yr-1 was applied to the soils in FSW-CZO, which is significantly higher than the optimal N input (150–250 kg ha-1 yr-1) in citrus orchards55.

Soil samples were collected by depth at the entire watershed scale to explore the vertical pattern of soil texture in the FSW (n = 2, June 2019; n = 14, July 2020). Furthermore, to explore the seasonal variation in soil N, soil samples at different depths were collected at station S1 during the dry season (n = 2, November 2019 and January 2021) and the wet season (n = 5, August 2018, July 2019, March 2021, April 2021, and August 2021) using a portable impact drill-corer (Jujiang, Shandong, China). Because the WTD was mainly distributed at ~1 m depth (Supplementary Table 2), a slightly greater depth (1.2 m) was considered for all soil sampling. The soil cores were divided at 0.2 m intervals. Here, the soils in both March and April of 2021 were further used to quantify the nitrification rate and microorganisms. Soil samples of ~1 kg for each sample were placed on ice and transported to the laboratory for further analysis.

Soil water and groundwater were monitored at the outlet of the watershed (Supplementary Fig. 8). To explore the formation of ENL, soil water was monitored in three adjacent stations at a depth of 0.4–0.5 m (no water), 0.7–0.8 m (n = 14), and 1.1–1.2 m (n = 34) using φ 7 cm PVC pipes. The bottom of the pipes had 5 rows of φ 5 mm holes (0.1 m high), which were covered with a 0.1 mm nylon screen to allow the inflow of soil water. The pipes had a 0.5 m extension above the ground and were covered with a removable cap. Water samples were collected by a portable peristaltic pump on a monthly scale. Before sampling, the old water was drained to obtain freshwater samples. Adjacent to these pipes, a larger-diameter monitoring well (site #1) was installed in 2021 for real-time WTD measurements. Groundwater (n = 278) was sampled approximately every 4 days from 2019 to 2021 at a riparian well (~3 m depth, site #2) for nutrient analysis. A pressure transducer in this well-recorded real-time WTD from 2019 to 2020, supplemented by monthly manual measurements. To track groundwater N sources, monthly rainfall (n = 7) and groundwater (n = 14) samples for isotope analysis were collected from April 2020 to June 2021.

Hourly rainfall data were obtained from the National Meteorological Science Data Center (Pinghe weather station, http://data.cma.cn/). Hourly river flow data were provided by the local Hydrology and Water Resources Survey Sub-Center. These data were aggregated to daily or weekly scales as needed. Daily baseflow was estimated from daily river flow using an automated separation procedure within the Hydrological Utility Package (http://www.yibaipin.com/).

Laboratory analysis

Soil NH4+-N and NO3--N contents were measured by segmented flow automated colorimetry (San + + analyzer, Germany) after extraction with 2 mol L-1 KCl. Soil particle size composition was determined by the Mastersizer 2000 particle size analyzer (Malvern Instruments, Malvern, UK). The soil gross nitrification rate was determined by the N isotope dilution method with δ15N-NO3- additions56. The gross nitrification rate was further multiplied by a coefficient of 0.76 (the proportion of autotrophic nitrification to gross nitrification) to represent the autotrophic nitrification rate22.

Soil genomic DNA was extracted using the E.Z.N.A. Soil DNA Kit (Omega Bio-Tek, Inc., USA). The extracted DNA was dissolved in 50 μL of TE buffer and stored at -20 °C until analysis. DNA concentration was determined using a Nanodrop 2000 (Thermo Fisher Scientific, Inc., USA) for subsequent normalization in quantitative PCR (qPCR) assays. The abundance of N-cycle functional genes was quantified by qPCR using a Bio-Rad CFX96 system (Bio-Rad Laboratories, Inc., USA). Detailed qPCR conditions are provided in Supplementary Table 9.

Temperature, pH, dissolved oxygen (DO), and conductivity in the watershed water environment were measured using a WTW multiparameter (Multi 3430, Germany). The contents of NH4+-N and NO3--N in water samples were determined by segmented flow automated colorimetry (San + + analyzer, Germany). Dual isotopes of nitrate (δ18O-NO3- and δ15N-NO3-) were determined following a chemical reduction method57. The detection of isotope samples was done by the Analytical Test Center of the Third Institute of Oceanography, Ministry of Natural Resources, China.

Statistical analysis

To obtain the lag time of groundwater N to fertilization, we first needed to determine N input at a daily scale. However, information on fertilization on a daily scale is largely unavailable in watersheds. Here, the annual fertilizer N was distributed on rainy days according to the proportion of weekly rainfall to annual total rainfall in fertilizing months (Supplementary Figs. 9 and 11a). This method was demonstrated to be optimal and consistent with agricultural practices in simulating the N output of the present study area58. Here, both R2 and the significance (p value) of linear regression were used to determine the lag time in the FSW-CZO. A t-test was used to compare differences in N and other parameters across seasons. Stable isotope analysis based on the SIAR model was used to calculate the contribution of major N sources from chemical fertilizer, manure and sewage, and rainfall to the groundwater NO3--N. The ranges of δ15N-NO3- and δ18O-NO3- values of these sources and fractionation factors were shown in Supplementary Table 5.