A global analysis of soil acidification caused by nitrogen addition

We searched the peer-reviewed papers using Web of Science during 1900–2014, whose title, abstract, or keywords related to: N addition, N deposition, N input, N application, N fertilization or N enrichment; soil; and acidification, pH, cation, Ca²⁺, Mg²⁺, K⁺, Na⁺, Mn²⁺, Al³⁺ or Fe³⁺. Then, we selected the appropriate studies with the following criteria. First, N-fertilizer was directly applied to field plots in terrestrial ecosystems with at least one of the above mentioned soil variables included. Second, the experiments were conducted to include an N-added treatment and a control treatment under the same condition. Third, the studies clearly showed N addition levels, experimental duration, and soil depth. Fourth, the measurements of the selected variables explicitly indicated their means, standard deviations or standard errors, and sample sizes, or these parameters could be calculated from the measured data.

To assure the independence among studies, we only collected the data from the latest measurement for each experiment (Liu and Greaver 2010, Lu et al 2011). Here, we considered data from different N addition rates and N-fertilizer forms in the same experiment as independent observations (Curtis and Wang 1998, Liu and Greaver 2009). Data from the figures were extracted using Engauge Digitizer (Free Software Foundation, Inc., Boston, MA, USA). Finally, a database with 106 independent studies was created from 61 published papers (table S1 in the supplementary data available at stacks.iop.org/ERL/10/024019/mmedia). Of them, 97 experiment sites referred to examining soil pH response. Overall, our database covers grassland, boreal forest, temperate forest and tropical forest in most of global land area (figure S1), with a rainfall range from 250 to 3500 mm and an annual temperature range from −3.7 °C to 28 °C. N addition levels vary from 0.93 to 60 g m⁻² yr⁻¹. The measured soils were sampled from a depth range from 2.5 to 30 cm.

In order to examine how N treatments, ecosystem types, and environment conditions influence the responses of soil acidification to N addition, we grouped the data based on N addition levels (<5, 5–10, 10–15 and >15 g m⁻² yr⁻¹), N-fertilizer forms (NH₄-fertilizer, NH₄NO₃-fertilizer and urea) and experimental durations (<5, 5–10, 10–20 and >20 yr). For ecosystem types, we categorized the data into grassland, boreal forest, temperate forest and tropical forest. For environmental conditions, we classified the data according to initial soil pH (3–4, 4–5, 5–6, 6–7 and >7), ambient N deposition (<0.5, 0.5–1, >1 g m⁻² yr⁻¹), rainfall (<1000, 1000–2000 and >2000 mm) and annual temperature (<0, 0–5, 5–10, 10–20 and >20 °C). Due to not enough data for soil total C and total N, the category of data based on these two variables was not conducted.

We analyzed the data by the traditional meta-analysis method offered by Hedges et al (1999). The mean magnitudes of N-added effects on the examined variables were estimated by the log-form of response ratio (R), log₁₀(R) = log₁₀ $\left( {{{\bar{X}}}_{{\rm treatment}}}/{{{\bar{X}}}_{{\rm control}}} \right),$ where ${{\bar{X}}_{{\rm treatment}}}$ and ${{\bar{X}}_{{\rm control}}}$ are the means of a certain variable in N addition and control treatments respectively. The variance (ν) of log₁₀(R) was calculated as following:

$$\begin{eqnarray*}&&\nu =0.1886\times \left( \frac{{{\left( {\rm S}{{{\rm D}}_{{\rm control}}} \right)}^{2}}}{{{N}_{{\rm control}}}\bar{X}_{{\rm control}}^{{\rm 2}}}+\frac{{{\left( {\rm S}{{{\rm D}}_{{\rm treatment}}} \right)}^{2}}}{{{N}_{{\rm treatment}}}\bar{X}_{{\rm treatment}}^{2}} \right),\end{eqnarray*}$$

where SD_control and SD_treatment are the standard deviations of the control treatment and N addition treatment; and N_control and N_treatment are the sample sizes of the control treatment and N addition treatment respectively. Relevant detailed information is presented in Hedges et al (1999). With the MetaWin software (Sinauer Associates, Inc. Sunderland, MA, USA), we analyzed the mean effect size and its variance of N addition. The N-added impacts on the response ratios are considered significant at α = 0.05, if the 95% confidence interval (CI) does not overlap zero.

To further examine how various treatments, ecosystem types and environment factors across studies affected N addition effects, the data was categorized as described in data source section. The mean response ratio for each categorized level and its variance were analyzed with the MetaWin. Total heterogeneity (Q_total) among groups was divided into two components, within-group heterogeneity (Q_within) and between-group heterogeneity (Q_between). If the probability value of Q_between is lower than 0.05, it indicates the significant difference in response ratios among different categorized levels. Mean effect of N application at each level was considered significant, if the 95% CI did not overlap zero. Here, we used two methods to calculate the response ratios of soil acidification. First, we directly computed the response ratio of soil pH. Second, pH was transformed to H⁺ concentration, then the H⁺ response ratio was also calculated. Based on these two methods, we found that the response patterns of soil acidification across different ecosystems, soils and environmental factors were consistent and the magnitudes were quite similar. It is noted that the equation of log₁₀(H⁺_treatment/H⁺_control) can be changed to the form of log₁₀(H⁺_treatment) – log₁₀(H⁺_control), which indicates pH unit change (pH_control – pH_treatment) affected by N addition. Thus, for ease of interpretation, we only presented the mean effect sizes of pH unit changes (pH_treatment – pH_control) in this study.

To clarify the mechanisms of N addition caused soil acidification, we used the linear regression method to analyze the relationships between soil metal cation and pH response ratios. In addition, we employed the regression analysis of linear, power or quadratic function to examine the relationships of pH response ratios with environmental factors. The regression analysis was conducted with the SPSS software (SPSS 11.0 for windows, SPSS Inc., Chicago, IL, USA), and the graphs were drawn with the SigmaPlot software (SigmaPlot 12.5 for windows).

At the global scale, N addition significantly reduced soil pH by 0.26 on average for terrestrial ecosystems (figure 1). Although soil pH in both grassland and forest ecosystems declined in response to N addition, the response magnitude of soil pH in grassland was larger than that in forest. For different types of forest ecosystems, boreal forest did not show a significant response in soil pH, while temperate and tropical forest displayed a significant decrease. Soil pH in the forest ecosystems dominated by conifer species was not affected by N addition, but soil pH in the forest ecosystems dominated by non-conifer species was reduced significantly (figure S2).

Zoom In Zoom Out Reset image size

Soil pH response changed with N addition rate. It was significantly reduced when the added-N amount was more than 5 g m⁻² yr⁻¹ (figure 2). Below this level, there was no significant response. Moreover, soil pH decreased linearly with increased N addition rate (figure S3). For different ecosystem types, a negative linear relationship existed between N addition rate and pH response ratios in tropical forest and grassland, while no relationship was detected in temperate and boreal forests. It was not observed that NH₄-form fertilizer addition decreased soil pH, while NH₄NO₃-form fertilizer and urea addition significantly reduced soil pH (figure 2). Soil pH was reduced significantly in the experiment less than 20 years but was not observed in the experiment longer than 20 years.

Zoom In Zoom Out Reset image size

When initial soil pH ranged from 3 to 4, N addition did not affect soil pH (figure S4). Above this range, N addition reduced soil pH significantly. Further regression analysis revealed that pH response ratios decreased linearly with initial pH (figure 3(a)). The response ratios of soil pH showed a power function curve with the increase of soil C and total N content (figures 3(b) and (c)).

Zoom In Zoom Out Reset image size

Overall, no general relationship was found between ambient N deposition and pH response ratio (figure 3(d)). However, when ambient N deposition was more than 0.5 g m⁻² yr⁻¹, soil pH decreased significantly with N addition (figure S4). Below this level, soil pH did not change significantly. The response ratios of soil pH increased slightly first and then decreased quickly, showing a quadratic response function with the increasing rainfall (figure 3(e)). A weak quadratic relationship (R² = 0.087) was found between pH response ratios and temperature (figure 3(f)). But, when temperature was lower than 0 °C, soil pH was obviously reduced (figure S4).

N addition significantly decreased soil exchangeable Ca²⁺, Mg²⁺ and K⁺ in terrestrial ecosystems (figure 4). The effects of N addition on these base cations were consistent between forest and grassland ecosystems. It was not observed that N addition reduced soil exchangeable Na⁺ in forest, but significantly decreased soil Na⁺ in grassland. For non-base cations, N addition significantly increased free Al³⁺, while it was not observed to influence soil exchangeable Mn²⁺. But for grassland soils, Mn²⁺ showed an increase.

Zoom In Zoom Out Reset image size

The response ratio of soil pH showed a similarly linear relationship with the response ratio of soil Ca²⁺, Mg²⁺, K⁺ (figure 5). However, there was no significant relationship between changes in Na⁺ and pH response ratios. For non-base cations, soil pH decreased linearly with the increase of free Al³⁺ or Mn²⁺.

Zoom In Zoom Out Reset image size

In natural conditions before the Industrial Revolution, the acidification rate in soils is rather low with over hundreds to millions of years. However, at a global scale we found that N addition significantly reduced soil pH by 0.26 in terrestrial ecosystems. This decreased magnitude is lower than a reduced pH of 0.50 reported in Chinese agricultural systems from the 1980s to the 2000s (Guo et al 2010). The different responses of acidification between agricultural systems and non-agricultural ecosystems may be due to more use of N-fertilizer in crop systems (above 50 g m⁻² yr⁻¹) than that in non-crop ecosystems (ranging from 1 to 30 g m⁻² yr⁻¹) (Rothwell et al 2008, Guo et al 2010). Moreover, a recent meta-analysis revealed that repeatedly harvesting plant biomass can lead to soil acidification (Berthrong et al 2009).

Different terrestrial ecosystems exhibited different sensitivities to N addition in soil acidification. Soil pH declined most in grassland by 0.49. This result is in line with the large-scale observations that soil pH decreased by 0.63 in Northern China's grassland during 1980s–2000s, which is mainly induced by N deposition (Yang et al 2012). The pH decline of 0.49 in this study is lower than the decrease of 0.63, which is partly due to the fact that the initial pH range of 3 to 7.7 in this study is lower than the range mostly above 7.5 in Northern China's grassland, since pH response ratio linearly decreased with the increase of initial soil pH (figure 3(a)). These two studies in combination indicate that grassland soils are very sensitive to N-induced acidification. Soil acidification in tropical and temperate forests was more sensitive to N addition than boreal forest (figure 1), which may be due to the influence of the variable abiotic and biotic factors across forest ecosystems. For example, more rainfall (1840 mm) in tropical forests than that (925 mm) in boreal forests can promote soil acidification (Lucas et al 2011). Temperature variation plays a minor role in affecting soil acidification among these forest ecosystems as our result showed that there was no obvious change in pH response with temperature when it was above 0 °C. The annual temperature for tropical forest (24.1 °C), temperature forest (8.3 °C) and boreal forest (4.7 °C) is higher than 0 °C. The less sensitivity of boreal forests to soil acidification is probably due to the dominance of conifer species in this ecosystem because we found that soils dominated by conifer species was less sensitive to N addition in acidification than those dominated by non-conifer species (figure S2). The finding highlights the significant role of conifer species in suppressing soil acidification.

The responses of soil acidification to N addition varied with N addition rate, N fertilizer forms, and treatment duration. N induced acidification became significant when N addition rate was more than 5 g m⁻² yr⁻¹, (figure 2), indicating that soil acidification is sensitive to N input, even at low N addition level. The finding is partly supported by previous studies that pointed out ecosystems usually undergo an accelerating leaching of NO₃⁻ from soils if N addition rate is larger than 2.5 g m⁻² yr⁻¹ (Dise and Wright 1995, Rothwell et al 2008). Furthermore, we found that soil pH decreased linearly with N addition rate, indicating that the acidification will continue to be sensitive to further N deposition.

N fertilizer forms also impacted the responses of soil acidification. We found NH₄NO₃-form fertilizer posed a significant impact on acidification, while NH₄-form fertilizer was not observed (figure 2), suggesting that NO₃⁻ leaching plays a dominant role in affecting acidification (Currie et al 1999, Gundersen et al 2006). N addition experiments with duration of less than twenty years had a significant effect on acidification, while the experiments with duration of more than twenty years were not observed to pose an impact on acidification. It likely suggests that ecosystems may be able to adapt to acidification induced by N addition in long term. There are two potential reasons accounting for this result. First, plant species composition may change to N-demanded species under high-N conditions (Stevens et al 2004, Bai et al 2010), enhancing the ecosystem capacity to sequester excess N. Second, more N-induced deficiency of base cations may stimulate biotic retention and internally recycling of base cations within ecosystems (Perakis et al 2013).

N addition effects on soil acidification varied with soil conditions and environmental factors. We found that the acidification was insensitive to N addition when initial soil pH ranges from 3 to 4, but was sensitive when above this pH range (figure S4). It indicates that soils with different initial pH have different acid-buffering capacity. It is generally expected that absorption ability of metal cations by soils is in the order of: trivalent ions (Fe³⁺ > Al³⁺)> divalent ions (Mn²⁺ > Ca²⁺ > Mg²⁺)> monovalent ions (K⁺ > Na⁺) (Bowman et al 2008). Low valence ions held in soils with high initial pH are most vulnerable to be replaced. Once these cations have been much depleted, it is high valence cations that buffer against acidification. Based on the principle of charge balance, soils of high initial pH should be more sensitive to acidification than those of low initial pH. This explained well for our result that pH response magnitude increased linearly with the increase of initial pH. Inconsistent with our expectation, the pH response ratios tended to increase with soil total N across the study sites. This is probably due to that high N reduced initial soil pH (unpublished data). As we showed above, the lower the initial pH, the less response the pH has. Consistent with our expectation, more C in soils was helpful to suppress acidification, indicating its high cation exchange capacity (Parfitt et al 1995).

We also found that when ambient N deposition was larger than 0.5 m⁻² yr⁻¹, soil pH was significantly decreased even by low levels of N addition. As reported by Penuelas et al (2013), most of land area receives more than 0.5 g m⁻² yr⁻¹ of N deposition, which implies that current N deposition may reach a critical loading to induce soil acidification in terrestrial ecosystems. In accord with our expectation, more rainfall obviously promoted soil acidification when rainfall was above 1500 mm. However, when rainfall was below 1500 mm, only small change was observed for soil pH with precipitation. It suggests that there exists a critical precipitation level of 1500 mm, above which soil acidification is promoted. Consistent with previous studies (Williams et al 1996, Curtis et al 2005), our result also revealed that low temperature promoted soil acidification (figure S4). But, only when the temperature was below 0 °C, soil acidification was obviously promoted. This is due to that low temperature decreases ecosystem N cycles, leading to a low capacity to sequester N. And surplus N leaching promotes depleting base cations. Overall, all these above results highlight the interactive effects of environmental factors and N addition on soil acidification.

Our results showed that N addition significantly reduced the exchangeable base cations of Ca²⁺, Mg²⁺ and K⁺ in soils, which is consistent with a previous study synthesized the relationship between N addition and base cations (Lucas et al 2011). Although this previous study showed N addition above five years had no effect on base cations, it did not have enough data to support. Our study provides sufficient data to reveal that N addition above twenty years did not significantly affect soil pH. Based on the tight coupling of metal cation loss and acidification (Berthrong et al 2009), these two meta-analysis studies hold the point that terrestrial ecosystems are likely to adjust to N-induced acidification in long term. Moreover, soil pH declined with deceasing base cations, suggesting the crucial role of base cations in buffering against acidification. Further analysis showed that there was no significant difference among different cations in their relationships with soil acidification, suggesting that these base cations follow a same buffering rule (linear function).

Additionally, we found that N addition significantly increased free Al³⁺ in soils, and soil pH decreased linearly with the increase of free Al³⁺ (figures 4 and 5). This indicates that N addition has shifted soils in terrestrial ecosystems into the Al³⁺ buffering stage. It was not observed that N addition increased free Mn²⁺ in global soils, but grassland soil Mn²⁺ displayed a significant increase. With the increase of free Mn²⁺, soil pH also reduced linearly, suggesting that grassland soils are simultaneously at a stage of Mn²⁺ buffering stage. Although the content of Mn element in earth crust is not comparable to those of the base cations mentioned above, its toxic impacts on organisms should be paid special attention to (Guo et al 2007, Poschenrieder et al 2008). All these call our caution to care about the danger of the coming buffering range of toxic Al³⁺ and Mn²⁺ with N deposition (Kochian 1995, Poschenrieder et al 2008).