Common bean yield responses to nitrogen fertilization in Brazilian no-till soils: A meta-analysis

: Although numerous studies have been conducted with common bean regarding nitrogen (N) dose, time of application, and source in no-till (NT) soils in Brazil, the heterogeneity of the results makes it difficult to establish technical recommendations based on individualized studies. This meta-analysis aimed to rank the main factors influencing common bean response to N management in NT areas. The database consisted of 99 scientific papers that encompassed 160 trials and 2394 observations. In general, the probability of obtaining a positive response to N application in common bean productivity was 77 %, with an average 18 % increment (358 kg ha -1 ). The main factors that affect the response of common bean grain yield to N fertilization and the choice of the applied N dose are soil organic matter (OM) content and the preceding crop. In soils with OM content >20 g kg -1 , the N dose that determines the highest economic return is 50 kg ha -1 . For soils with OM content <20 g kg -1 , the N dose to be applied is 70 and 100 kg ha -1 for crops preceded by legumes and grasses, respectively. The timing of the N application did not result in a significant difference in common bean yield in 62 % of the trials. However, the results showed that the average yield increase in the common bean went from 15 % (293 kg ha -1 ) with a single application to 32 % (622 kg ha -1 ) with split N applications. All N sources tested showed similar gains to those obtained with N application through urea. Excluding situations with high doses of N application (>100 kg ha -1 ), inoculating common bean seeds increased grain yields by 6 % on average (118 kg ha -1 ). The combined analysis of the available results demonstrated that the crop recommendations used in Brazilian soils under NT could be refined considering the soil OM concentration and preceding crop. Additionally, seed inoculation and N dose splitting in the form of common urea should be recommended.


INTRODUCTION
Common bean (Phaseolus vulgaris L.) is a legume native to the Americas with high climatic adaptation and cultivated throughout Brazil (Salgado et al., 2012a). It is a staple food in developing countries and an important protein, mineral, and vitamin source (Lião et al., 2010;Hungria et al., 2013). In Brazil, small producers generally grow the common bean for subsistence via family labor and low technological level, obtaining low yields (Lobo et al., 2012). In recent years, the possibility of mechanizing the harvest of common bean and improved productive potential has led this crop to be produced at a large scale and part of crop rotation systems to diversify and intensify (off-season) the production (Dal Molin and Ernani, 2017).
Common bean is traditionally grown in central and southern Brazil, while the cowpea (Vigna unguiculata L.) is predominant in the north and northeast. Roughly 1.34 million ha of land is planted with the common bean in central and southern Brazil, with average productivity of 1.72 Mg ha -1 (2019/2020 season; Conab, 2020). Moreover, the average productivity of common bean crops has gradually increased in recent decades in Brazil, although it is far below the potential of the crop, which is above 4.0 Mg ha -1 (Barbosa Filho et al., 2004;Venturini et al., 2005;Arf et al., 2011a). Various factors contribute to the low average productivity in the country, including low technology employed (Lobo et al., 2012), inadequate phytosanitary control, adverse weather conditions (Damian et al., 2018), and inadequate fertilization management (Santi et al., 2006). Regarding fertilization, the common bean requires significant quantities of the nutrient nitrogen (N) (Crusciol et al., 2007), which influences vital plant functions such as photosynthesis and protein synthesis, being a primary constituent of chlorophyll, amino acids, nucleic acids, hormones (e.g., auxin and cytokinin), and alkaloids (Rabelo et al., 2017). Moreover, nitrogen deficiency directly influences biomass production and causes premature leaf senescence, reducing grain yield (Santos and Fageria, 2007).
The main source of N for common beans is from soil organic matter (OM). In no-till (NT) systems, the crop residues are kept on the soil surface, creating a concentration gradient of OM and nutrients in the soil profile that changes the dynamics of N compared to conventional tillage (Veloso et al., 2018). Furthermore, NT system creates a less oxidative environment in the soil that reduces the rates of N mineralization of OM, although OM accumulation compensates for this reduction over time, and the soil begins to provide more N to crops (Lovato et al., 2004). In turn, the increased cation exchange capacity (CEC) in the soil due to higher OM content (pH-dependent loads) favors N-NO 3 losses by leaching (Rojas et al., 2012). The NT areas have increased activity of the denitrifying community and, when not managed properly, they may lead to anaerobiosis due to soil compaction, thereby favoring N 2 losses by denitrification (Wang and Zou, 2020). Due to the dynamics of N in the soil, the various types of losses, and its high mobility in the profile, it is difficult to predict its availability for crops and an even greater challenge to make recommendations for N fertilization.
Through symbiosis with nitrogen-fixing bacteria (rhizobia), common bean converts atmospheric N into ammonium via nitrogenase activity (Pelegrin et al., 2009;Hungria et al., 2013); however, symbiotic N fixation in the common bean is not as efficient as other legumes (e.g., lupine) due to several factors. One of the more notable factors is that common bean perform symbiosis with a high diversity of native bacteria in the soil, some of which are highly competitive and not as efficient as the selected strains (Pasqualini, 2008). Nitrogen fertilization also interferes in the crop nodulation. Soils with a high N concentration make the common bean more prone to absorbing the N available in the soil solution than performing symbiosis, which has a higher energy cost for the plant (Hungria et al., 2013). Thus, biological N fixation is usually not enough to meet the nutritional demands of the common bean, making it is necessary to use mineral fertilization to complement the crops' needs to achieve high yields (Cunha et al., 2011;Mingotte et al., 2014;Soratto et al., 2017). Another component that favors common bean responses to N fertilization is its intrinsic feature of poorly developed and shallow root systems (less than 0.20 m deep), reducing its capacity to absorb N and water in soil subsurface layers (Caires et al., 2016). In addition, the current cultivar cycles are short and commonly vary between 90-100 days (Menegol et al., 2015;Silveira and Gonzaga et al., 2017;Damian et al., 2018), making the crop highly demanding in terms of soil N availability.
Nitrogen recommendations in Brazil are mainly based on estimates of N availability in the soil due to OM mineralization and common bean N requirements (expected yield) (CQFS-RS/SC, 2016) because there is no laboratory analysis capable of accurately identifying N availability in the soil for crops during their cycle (Barbosa Filho et al., 2008). For the southernmost Brazilian states (Rio Grande do Sul and Santa Catarina), the official recommendation for the common bean considers that doses should range from 30 to 70 kg ha -1 according to the soil OM content. In addition, the expected yield should be considered, and when above 2.0 Mg ha -1 , the dose should be increased by 20 kg ha -1 for each ton of grain to be produced (CQFS-RS/SC, 2016). For Minas Gerais State (southeastern Brazil), the recommendation is 40-100 kg ha -1 depending on the technological level employed (Ribeiro et al., 1999). The recommended doses vary from 20 to 90 kg ha -1 in São Paulo State (southeastern Brazil) depending on expected yield, preceding crop, and soil texture (Ambrosano et al., 1997).
Given the significant influence of N fertilization on common bean grain yield, a series of NT studies have been conducted in Brazil to determine the best N fertilization management. Nevertheless, the results varied widely and ranged from research with yield reduction by N application (Silva et al., 2006(Silva et al., , 2009aArf et al., 2014;Guimarães et al., 2017), lack of response (Alves Junior et al., 2009;Lopes et al., 2011;Pacheco et al., 2012Pacheco et al., , 2016Dal Molin and Ernani, 2017;Flôres et al., 2017), and significant increases in yield (Barbosa Filho et al., 2008;Franco et al., 2008;Teixeira et al., 2008;Sant'ana et al., 2010;Cunha et al., 2011). Local variables related to soil properties, climate (Damian et al., 2018), field management history (Arf et al., 2011b;Maia et al., 2012), N management, cultivar genetics (Salgado et al., 2012a), and biological N 2 fixation efficiency (Maia et al., 2012) have also been identified as factors contributing to the high variation in the response of the common bean to N fertilization.
Given the heterogeneity of the responses of the common bean to N fertilization obtained by individual studies conducted in different Brazilian regions, it is difficult to extract information to support the refinement of N recommendations for the crop. In this context, a meta-analysis of the data available in the literature was conducted to answer the following questions: (i) What is the average increment in grain yield by N fertilization and what is the probability of obtaining a positive response? (ii) Which factors influence the magnitude of common bean grain yield response to N fertilization? (iii) Which dose, the timing of application and source of N show the best results? (iv) What is the viability of inoculating common bean seeds?

Data collection
The data covered by the meta-analysis were obtained from scientific papers that evaluated the effects of N fertilization on common bean productivity in NT systems in Brazil. The following search terms in Portuguese and English (separately and together) were used in the databases Web of Science, SciELO, and Google Scholar: "nitrogen", "common bean", "Brazil," and "no-till system". The papers that met one or more of the following criteria were excluded: i) production under conventional tillage soil, ii) studies in greenhouses, iii) lack of control treatment (without N), and iv) lack of grain yield data. After applying the exclusion criteria, the database consisted of 99 scientific papers published from 2000 to 2018, totaling 160 trials (site × year) and 2394 observations (Table 1). (1) Number of observations; (2) Number of the trials evaluated in each primary study (year x location); (3) UR: Urea; AS: Ammonium sulfate; AN: Ammonium nitrate; ASN: Ammonium sulfonitrate; NC: nitrocalcium; CN: calcium nitrate; URUI: urea with urease inhibitor; and URSR: urea of slow-release; (4) DBS: days before sowing; and DAE: days after emergence ; (5) Use of seed inoculation.
Rev Bras Cienc Solo 2022;46:e0220022 The information extracted from the papers was the average grain yield of treatments with N and the control treatment, which did not receive N application, the N dose applied (0-280 kg ha -1 ), N application time (anticipated, at sowing, and topdressing in different phonological stages), number of N application (the total N dose applied at once, or split in two or three times) sowing season [sowing season of the common bean in the primary studies: crop (September to November), off-season (January to March), and winter (May to July)], N source used (urea, ammonium sulfate, ammonium nitrate, urea with urease inhibitors and others), whether or not N was incorporated into the soil (through irrigation soon after N application), soil OM content, soil clay concentration and pH, soil type (Oxisol, Ultisol, Alfisol and Entisol), climate type (tropical or subtropical), use or not of irrigation and seed inoculants, time of adopting (years) the NT system, commercial grain group of the common bean (carioca, buttery, and black), and preceding crop (grass or legume).

Regression tree analysis
Conditioning factors of the effects of N fertilization on common bean yield were determined by regression tree analysis using Jmp® 13 software (SAS Inc, Cary, USA). Categorical (preceding crop, climate zone, soil type, commercial group of grains, N source, inoculant use, N incorporation, irrigation use) and continuous (N dose, clay concentration, OM content, soil pH, NT system adoption year) variables were used in the analysis as variation factors and the effects of N fertilization on common bean grain yield (relationship between treatment yield and control yield) as the response variable.
The regression tree analysis was based on the test of zero variance, that is, the independence between the variation factors and response variable (effect on productivity). When the first hypothesis is rejected, the software selects the variation factor closer to the response variable, repeating the process in each node of the regression tree. Data analysis reliability was maintained and data particularization was avoided by setting the initial and intermediate nodes of the regression tree to contain at least 20 % (479 observations) of the total data.

Data meta-analysis
Meta-analysis was performed by only comparing paired treatments (with and without N fertilization in the same trial), and from the entire database, the effects of N dose, N source, seed inoculation, and timing of N application were evaluated separately. For N dose, the grain yield of the treatments (with N) and the controls were used to determine their differences (treatment -control), which was considered the effect of N fertilization on grain yield. Different N sources were compared with urea, which was considered a control since it is the most used N source in Brazil. For the effects of inoculation, the difference in yield was calculated for treatments with and without inoculant (control). The effects of the timing of N application were not evaluated using meta-analysis because the treatments applied in the primary studies were highly variable. In addition, some studies presented the timing of N application chronologically in days after emergence and/or sowing instead of the phenological stage. Thus, we counted the trials in which there was or was not a significant effect of the application timing treatments. The most promising treatments were manually identified in the studies with a significant difference between the timing applications.
From the values of the effects of treatments on grain yield, the natural logarithm of the values was obtained for the analyses. Since most of the primary studies used in this meta-analysis did not present dispersion measures of their means, the different studies were weighted according to the number of repetitions (n) using equation 1, as suggested by Pittelkow et al. (2015) and Kihara et al. (2017). To privilege results obtained in different trials (year × location), when the trial presented more than one observation for the same subgroup of data, the weight of the trial was divided among the number of observations (Pias et al., 2020).
Rev Bras Cienc Solo 2022;46:e0220022 Weight = {(n control × n treatment) ÷ (n control + n treatment)} Eq. 1 The regression tree selected factors of interest for the meta-analysis, including soil OM content and preceding crops. For each factor in the database, the 95 % confidence intervals for the obtained means of effect of N fertilization on grain yield were calculated by the Bootstrapping procedure using 4999 repetitions. The Bootstrapping procedure is an efficient way to estimate confidence intervals by simulating different datasets from the initial dataset. In addition, the data to be processed need not exhibit a normal distribution, and the results are scarcely influenced by the presence of outliers. A significant effect of the treatments was considered when the confidence intervals did not exceed the value zero and when the factors did not cross their confidence intervals. The meta-analysis data was analyzed using the Statkey 2.0.1 software (Lock et al., 2017). At the end of the analysis, the mean values of the response rate were transformed to the normal scale and presented as the effect on common bean grain yield (kg ha -1 ).
To evaluate the profitability of the different N managements, the average price of the

Characterization of research on nitrogen response by common bean under no-till
Studies evaluating the effects of N fertilization on the productivity of the common bean under NT systems have been conducted throughout Brazil (Figure 1), although they are mainly concentrated in the central-western region, especially Mato Grosso do Sul and Goiás States, accounting for 59 % of the total trials evaluated. Paraná, Minas Gerais, Mato Grosso, and Goiás States were decreasingly the largest producers of common beans in the 2019/2020 season (Conab, 2020). In fact, these states also stand out in conducting N fertilization experiments in the common bean in Brazil (Figure 1).
Of the total trials analyzed, 89 % were grown in tropical climates and only 11 % in subtropical climates (Figure 2a). Irrigation was used in 70 % of the analyzed trials (Figure 2b), and this is because most research was conducted in the winter season (dry season) of midwestern Brazil, requiring irrigation. Carioca common bean was studied in 89 % of the analyzed studies (Figure 2c), and this is also the most commonly consumed cultivar group in Brazil, especially in the southeastern and midwestern regions (Rocha, 2013), that are locations where most studies were conducted. The vast majority (92 %) of the trials that evaluated the effects of N fertilization on the common bean were conducted in Oxisols (Figure 2d), and this is because it is the soil with the highest occurrence in the country, with an even higher proportion in midwestern Brazil (Fageria and Baligar, 2008).
The evaluated database showed a predominance of grasses (88 %) preceding common bean crops, with only 10 % of legume species (Figure 2e). This scenario is theoretically of higher probability of a positive response of common bean to N application because grasses have a high C/N ratio occurring temporary immobilization of soil N. In contrast, legumes crops have a low C/N ratio, which results in net mineralization of N and higher N availability for the following crop (Amado et al., 1999).
The most used sowing season in the experiments was the third season (70 %), whose sowing is done in winter (dry season in midwestern Brazil), followed by the harvest season (18 %), and off-season (12 %) (Figure 2f). Sowing in the third season occurs from May to July, and it is entirely irrigated and represents about 19 % of the national common bean production (Conab, 2019). The first season (crop) is sown from September to November and usually meets its water demand naturally by rainfall. The off-season is sown from January to March soon after harvesting corn or soybean; it is commonly called the dry season because the rainfall does not always meet the water demand, requiring the sporadic use of irrigation to not compromise productivity (Silva, 2002). The crop and off-season represent roughly 45 and 36 % of the Brazilian common bean harvest, respectively (Conab, 2019).
The vast majority of research (94 %) that evaluated the effects of N fertilization on the common bean grain yield does not mention seed inoculation, and it was assumed that inoculation was performed in only 6 % of the studies (Figure 2g).
The most commonly used N source was urea, which accounted for 78 % of the data set ( Figure 2h). Urea is the most used N source in Brazil due to its high N concentration (45 %) and low cost per nutrient unit (Dal Molin and Ernani, 2017). In 83 % of the trials, the effects of N doses were evaluated in only one topdressing application (i.e., with the absence of splitting the doses; Figure 2i).
Soil pH(H 2 O) ranged from 4.5 to 6.8, although it was concentrated between 5.45 and 6.0 (Figure 3a), which is considered suitable for common bean due to the absence or low concentration of exchangeable Al 3+ (CQFS-RS/SC, 2016  (1) Macaé RJ (1) Ribeirão Petro SP (2) Jaboticabal SP (7) Piracicaba SP (1) Botucatu SP (18) Pereira Barreto SP (1) Selvíria MS (56) 150 to 790 g kg -1 , although it was concentrated at the 450 to 530 g kg -1 range (50 % of data; Figure 3b). Soil texture directly affects water retention and soil structure, and the highest the clay concentration, the greatest the water retention capacity, which can be partially made available to crops. The soil OM content found in the experiments ranged from 13 to 75 g kg -1 , and in half of the studies varied between 18 and 29 g kg -1 (Figure 3c). Organic matter is the primary source of N in soil and important for CEC due to the many functional groups in its structure. In addition, it acts as a cementing agent in the soil structure, improving soil porosity and stimulating soil biological activity (Veloso et al., 2019).
Nitrogen doses evaluated ranged from 15 to 280 kg ha -1 , albeit 50 % of the crops applied doses between 50 and 100 kg ha -1 (Figure 3d), which corresponds to N doses recommended by most recommendation systems in different states (

Average increase of common bean grain yield by N fertilization
Nitrogen application increased common bean grain yields in most studies (Figure 4a). The average grain yield increase with N fertilization was 18 % (357 kg ha -1 ), and the average N dose to obtain this increase was 84 kg ha -1 (Figure 3d). The results revealed that the magnitude of the effect of N fertilization on common bean yield depends not only on the dose applied but also on the technological level employed in the crops.
In trials with a low technological level (low soil fertility and poor phytosanitary and crop management), where the control treatments obtained low yields, the effects of N fertilization were more significant than in places with a higher yield potential where the controls treatments obtained higher yields (Figure 4b). These data confirm the Rev Bras Cienc Solo 2022;46:e0220022 assumptions used in official recommendations (e.g., CQFS-RS/SC, 2016), in which the most significant increases by applying fertilizers occur when the soil presents lower nutrient availability to plants caused by inadequate soil management.

Factors influencing the magnitude of the response of common bean to N fertilization
The regression tree data allowed us to determine that the soil OM content, the preceding crop, and the N dose applied were the factors that most influenced common bean responses to N fertilization (data not shown). Organic matter is the main source of N in the soil for plants and soil biota, in this way; soils with higher OM content have higher N availability for plants. In addition, OM improves soil aggregation and increases soil water availability. Therefore, cause plants to respond less to N fertilization. Among the primary references used to recommend N for common bean, only CQFS-RS/SC (2016) recognizes soil OM as a parameter to determine the N dose to be applied. Unlike what is expected and what is currently used by all recommendation manuals, the grain yield or, in the case of recommendation, the expected common bean yield did not influence the response of the crop to N fertilization.
Aiming to predict the response of common bean to nitrogen fertilization, the data compiled from the literature were divided into four different scenarios from the combination of grass and legume preceding common bean and soils with OM content above or below 20 g kg -1 (Figure 5a), a value determined from the regression tree analysis.
In the first scenario, when the preceding crop was leguminous, and the soil had an OM content <20 g kg -1 , there was no significant difference between the doses of N. However, doses above 40 kg ha -1 showed an increase in common bean yield, on average 296 kg ha -1 (Figure 5a). Nonetheless, for crops following legumes in soils with OM content >20 g kg -1 (Figures 5b) Figure 5. Effects of nitrogen (N) doses on common bean grain yield preceding by legumes in soils with organic matter (OM) content below (a) and above (b) 20 g kg -1 and preceded by grasses in soils with OM content below (c) and above (d) 20 g kg -1 . Values in parentheses represent the number of observations and the trials, respectively.
Rev Bras Cienc Solo 2022;46:e0220022 257 kg ha -1 . This result is due to the low C/N ratio of the preceding crop and the OM content being >20 g kg -1 , making significant amounts of N available for plants and microorganisms. However, when the crop is grown after grass in soil with low OM content, the soil N stock is lower, and thus much of the N applied at low doses is immobilized by microorganisms, reducing access to plants and their response to low N doses (Blagodatskaya and Kuzyakov, 2011). In this scenario, where the preceding crop was grass, a more significant potential response of common bean to N doses was observed if compared to growing after legumes, especially in soil with OM concentration <20 g kg -1 (Figure 5c). In this scenario, a linear increase in grain yield occurred with increasing doses of N applied up to doses >120 kg ha -1 . Doses above 80 kg ha -1 of N increased on average 707 kg ha -1 , although this was not the case when the common bean was grown after grasses in soils with OM content >20 g kg -1 , as the crop response was lower and doses between 40-60 kg ha -1 did not differ from larger ones, increasing grain yield by 300 kg ha -1 , on average (Figure 5d).
In common bean preceded by legumes, the probability of a positive effect of N fertilization is 88 and 56 % (Figures 6a and 6b) for soils with OM content below and above 20 g kg -1 , respectively. Nonetheless, the probability of higher common bean grain yield by N fertilization when the preceding crop is grass in soil with OM concentration <20 g kg -1 Figure 6. Probability of increases in common bean yield by nitrogen (N) application in crops preceded by legumes in soils with organic matter (OM) content below (a) and above (b) 20 g kg -1 and preceded by grasses in soils with OM content below (c) and above (d) 20 g kg -1 . is 92 % (Figure 6c), dropping to 71 % when the OM content is >20 g kg -1 (Figure 6d). These findings confirm that the common bean responds to nitrogen fertilization in most cases, evidencing that biological nitrogen fixation and the N mineralized by OM are not enough to meet the nutritional demands of plants in most cases. Particular attention should be given to soils with OM content >20 g kg -1 and preceded by legumes because 44 % of the trials in these environments presented no positive response to N application. In these cases, following the recommendation of CQFS-RS/SC (2016), the observation of active nodulation in the root to the 15-20 days after emergence may be important to analyze the need to add N in topdressing.
In the studies in which there was no positive effect of N application on grain yield (23 % of the data), this lack of response is attributed to several factors: among them the low yield of some cultivars (Salgado et al., 2012b;Damian et al., 2018), high biological N fixation efficiency (Maia et al., 2012;Hungria et al., 2013;Dal Molin and Ernani, 2017), soils with high OM content (Menegol et al., 2015;Dal Molin and Ernani, 2017), and the use of legume crops preceding the common bean (Arf et al., 2011a).
Contrary to expectations, full crop irrigation and N incorporation into the soil by irrigation after urea application to reduce N losses by NH 3 volatilization did not influence common bean response to N fertilization (Figures 7a and 7b). Under the present range soil pH(H 2 O) also did not significantly influence the response of common bean to N fertilization (Figure 7c), although there was a tendency for reduced response in soils with pH(H 2 O) >6.0. This result may be related to the higher efficiency of biological N fixation and higher N losses by volatilization in the form of NH 3 that occur in alkaline soils when using urea (Cunha et al., 2011;Alves et al., 2021), in addition to higher microorganism activity resulting in more significant N mineralization of soil OM.
Common bean cultivated in soils with clay content <500 g kg -1 were more responsive to N doses (Figure 7d), and this may have occurred for several reasons, including the fact that soils with low clay content have higher hydraulic conductivity, favoring N leaching in the form of NO 3 for the subsoil (Sangoi et al., 2003), thereby increasing the dependence on the addition of N via fertilizer. In addition, sandy soils tend to have lower OM content due to the lower physical and chemical protection effect (iron oxide presence) of OM (stabilization mechanisms) thereby reducing the amount of N mineralized by OM and made available in the soil solution. Nevertheless, in soils with clay content >600 g kg -1 , there was also a high increase in common bean grain yield due to the application of N, and this may have been due to the greater productive potential of common bean and/or greater compaction of soils with high clay content, which can limit root growth in the soil subsurface and, consequently, reduce water and nutrient uptake by plants and nodulation (Caires et al., 2016).
Cultivation time under NT of the experimental area also influenced the response of common bean to N fertilizer (Figure 7e). Crops grown in areas with <4 years of NT presented a more significant response to N doses than crops grown in areas with NT consolidated (>4 years) (cut-off level determined in the decision tree regression analysis). This result is possibly due to the slower mineralization of OM under NT (Cantarella, 2007). Research has shown that soils may present low N availability in the first years of NT due to the high immobilization of this nutrient and lower mineralization rates, generating an additional need for fertilization. However, well-managed NT leads to higher OM content and improved soil physical structure and biological activity in the long term, thus reducing the demand for external inputs (Veloso et al., 2018(Veloso et al., , 2019. Common beans grown in the tropical climatic zone showed a more significant response to N doses ( Figure 7f); this result may be related to the lower OM contents (24 vs 32 g kg -1 ) observed in these locations due to the higher temperature and precipitation, which generate increased in microbial activity under higher mineralization rate and soil weathering compared to subtropical soils (Amado et al., 1999).
Among the common bean sowing seasons, we observed that common beans grown in the third season (winter) were more responsive to N fertilization than the first season [rainy season (Figure 7g)] because, in the third season, it is possible to obtain higher yields due to use of irrigation, which increases nutrient demand. Finally, among the commercial classes of common bean, carioca and black bean showed better response to nitrogen fertilization than buttery, although the latter has a restricted number of observations, therefore making this result inconclusive ( Figure 7h). Furthermore, the commercial classes black and carioca are the most consumed in Brazil and, consequently, have undergone improvements to increase their productive potential. The average grain yield of the commercial groups carioca and black was 2260 kg ha -1 while the buttery was only 1420 kg ha -1 , the higher the grain yield, the greater the plant requirement for nutrients.

Common bean response to application times and N sources
The time of N application did not influence common bean grain yield in 62 % of the 39 trials that evaluated this factor. Therefore, the applied dose of N is generally more important than the time of its application because even when testing contrasting application times ranging from 20 days before sowing (Silva et al., 2002), application at sowing (Silva et al., 2002), and applications at various stages of development from V 2 to R 7 (Guimarães et al., 2017), there were no significant differences in most crops. In the studies in which the time of application influenced yield (38 % of the cases), the best results were mostly obtained using doses of up to 30 kg ha -1 at sowing and the rest of the dose between V 3 -V 4 , or 15 to 30 days after plant emergence (Guimarães et al., 2017;Soratto et al., 2003;Gomes Junior et al., 2005), which are similar to what is currently recommended in Brazil (Ribeiro et al., 1999;CQFS-RS/SC, 2016).
The split application of N improved common bean yields compared to the single application ( Figure 8). Regardless of the dose used, the split application of N resulted in twice the Rev Bras Cienc Solo 2022;46:e0220022 increase in yield. On average, there was a 15 % increase (293 kg ha -1 ) in grain yield when N was applied in a single dose, while in two or more applications, the average increase was 32 % when N was applied (622 kg ha -1 ). The positive effects of splitting may be associated with reduced N losses by leaching in the form of NO 3 -, volatilization in the form of NH 3 , and denitrification in the forms of NO and N 2 O, resulting in greater N efficiency by plants (Cantarella, 2007). Notably, N losses in the system are affected mainly by environmental conditions; thus, splitting N application reduces the probability of significant losses (Sangoi, 2003). In addition, splitting provides better synchrony between N availability in the soil and plant demand. Although the number of studies with more than one application was much lower than with only one application, our findings show that splitting, even at lower doses, provided a more significant increase in common bean grain yield (Figure 8). Currently, the main recommendations in use for common bean in Brazil (e.g., Ambrosano et al., 1997;Ribeiro et al., 1999;CQFS-RS/SC, 2016) have already recommended is splitting N doses for common bean, with small amounts at sowing and another application as topdressing, with no recommendations for splitting in two applications in topdressing.
Urea is the most used N source in Brazil due to its high N concentration (46 %) and lower cost per unit (Cunha et al., 2011;Dal Molin and Ernani, 2017). However, urea presents high potential N losses by volatilization [NH 3 ] compared to other sources, including ammonium sulfate and ammonium nitrate, which reduces its use efficiency. Thus, alternative sources of N have been studied and used instead of urea to reduce N losses (Dal Molin and Ernani, 2017). In addition to the sources cited, the use of urea with nitrification and urease inhibitors (stabilized) and coated urea (slow or controlled release) has shown promising results for various crops such as corn and rice (Frazão et al., 2014;Lyu et al., 2015). Results of 21 trials comparing N sources revealed that, compared to urea, the use of ammonium sulfate significantly reduced common bean yields by 40 kg ha -1 on average, while ammonium nitrate increased by 56 kg ha -1 (Figure 9b). Nevertheless, the use of urease inhibitors resulted in similar yields as common urea. Comparing the average effects of alternative N sources and urea showed no significant differences to improve common bean yields (Figure 9a). The lack of differences between N sources and urea has been attributed to climatic conditions favoring the reduction of N losses by urea (Bernardes et al., 2015;Dal Molin and Ernani, 2017), in addition to the fact that most studies were conducted in the dry season and relied on irrigation. Furthermore, the common bean crop uses soil N made available by OM mineralization and biological N fixation. Hence, small reductions in N losses by using alternative sources do not affect crop productivity, and therefore, using the source with the lowest cost per kg of N available on the market should be recommended.

Effects of inoculating common bean seeds on grain yield
The inoculation of common bean seeds is a practice recommended by CQFS-RS/SC (2016) due to the low cost and potential to increase the efficiency of biological N fixation using selected strains. Our results encompassing 37 trials that evaluated paired treatments with and without inoculation showed an average grain yield increment of 6 % (118 kg ha -1 ) but restricted to situations where N application was below 100 kg N ha -1 (Figure 10). At N doses above 100 kg ha -1 , there was no effect of inoculation, corroborating Zeffa et al. (2018) in a meta-analysis of the effects of inoculation on corn yields. According to Zeffa et al. (2018), high N contents inhibit nitrogenase activity in bacteria, which is the enzyme responsible for converting atmospheric N into ammonium.

Profitability of different N management in common bean
Based on the data acquired in this meta-analysis, a simplified estimate was made of the profitability of using N doses between 20 and 180 kg ha -1 under the conditions that most influenced the size of the effects of N on the yield of the common bean crop (preceding crop and soil OM content; Figure 5). The cost of N doses ranged from R$ 52.00 (US$ 13.44) to R$ 467.00 (US$ 120.67) per hectare. The highest profitability occurred for the treatment with 180 kg ha -1 of N with grasses preceding the common bean growing in soils with OM content <20 g kg -1 (Figure 11). The lowest and only negative profitability occurred for soils with OM >20 g kg -1 and legumes as a preceding crop at N dose of 100 kg ha -1 . On average, for every R$ 1.00 invested in N fertilization, R$ 4.00 was obtained; this return increases to R$ 5.50 when common bean are grown after grasses and in soils with OM content <20 g kg -1 . The lower the dose applied, the greater the return per amount spent with N. When using doses of 20 kg ha -1 , for every R$ 1.00 invested in N, there was a R$ 7.00 return, while for the dose of 180 kg ha -1 , the return of investment reduced to R$ 2.30 for every R$ 1.00.
Given the above, it is possible to determine the best doses for each environment. For common bean grown in soils with OM content <20 g kg -1 preceded by legumes, the N dose should be 70 kg ha -1 . However, when common bean are preceded by grass, the dose should be 100 kg N ha -1 . For soils with OM content >20 g kg -1 , the preceding crop has no significant influence on the profitability of N doses, and a dose of approximately 50 kg ha -1 should be recommended. These results show the importance of correct soil management, in which, in addition to minimal soil disturbance (NT), a high contribution of residues mainly with legumes is recommended to increase soil OM content and thus Preceding by legumes with OM concentration <20 g kg -1 Preceding by grass with OM concentration >20 g kg -1   180  160  140  120  100  80  60  40  20 Preceding by legumes with OM concentration >20 g kg -1 Figure 11. Profitability of using different nitrogen doses for common bean cultivated on soils with different organic matter (OM) concentration and preceded by grasses and legumes under no-till in Brazil. Red dashed line indicates the net balance equal to zero.
Rev Bras Cienc Solo 2022;46:e0220022 reduce the N doses applied in crops and, consequently, reduce production costs and increase profitability (Veloso et al., 2018).
An expected advance for future N management in common bean is the large-scale use of sensors (e.g., chlorophyll meters) to determine the N sufficiency index (NSI) and, consequently, N application according to each area's needs and at the correct time; nevertheless, our findings are still incipient and inconclusive. Barbosa Filho et al. (2009) and Maia et al. (2012) observed the need to apply N when the NSI is below 90 %, while Silveira and Gonzaga et al. (2017) concluded it is necessary to apply N as soon as the NSI is below 95 %. Menegol et al. (2015) observed that plants with NSI <90 % also showed no positive response to N application. Thus, although this line of research is promising, further research should be conducted to enable these decision support tools at the farmer scale to improve N recommendations for the common bean crop.

CONCLUSIONS
In general, the probability of increasing common bean yields by applying N is 77 %, with an average increase of 18 % (358 kg ha -1 ). The dose of N applied, soil OM (organic matter) content, and preceding crop are the main factors influencing common bean response to N fertilization in Brazilian NT areas. Higher yield increments as a function of N fertilization were obtained in soils with low OM content (<20 g kg -1 ) compared to soils with OM content >20 g kg -1 (23.5 vs. 15.5 %, respectively) and in crops preceded by grass relative to legumes (18.6 vs. 15.3 %, respectively). The doses at which the highest profitability were obtained were 50 kg ha -1 of N for soils with OM content >20 g kg -1 , regardless of the preceding crop. For soils with OM content <20 g kg -1 , the N doses should be 70 and 100 kg ha -1 for crops preceded by legumes and grasses, respectively.
Timing of N application did not promote significant differences in grain yield in 62 % of the trials. However, splitting the N dose increases the increment in common bean yield from 15 % [(293 kg ha -1 ) single dose] to 32 % [(622 kg ha -1 ) split two or more times]. Alternative sources of N do not result in additional benefits compared to urea; therefore, urea should be the N source of choice due to its lower cost. When combined with mineral N doses below 100 kg ha -1 , seed inoculation provides to common bean an average increase in grain yield of 6 % (118 kg ha -1 ) and can be a good alternative due to its low cost.
This meta-analysis confirmed the importance of N application for the common bean, with an average return of R$ 4.00 for every R$ 1.00 invested in fertilization. Furthermore, the N fertilization recommendations for common bean used in NT Brazilian soils could be refined considering the soil OM content and preceding crop. These data serve as a scientific basis to support updates in the official fertilization recommendations, seeking the rational management of N in the common bean crop under NT areas in Brazil.

ACKOWLEDGEMENTS
Authors are grateful to FAPERGS (Foundation for Research Support of the State of Rio Grande do Sul) for funding the research through of the project "RITE-Carbon Alliance: Low Carbon Agropecuary and adapted to Climate Change in the State of Rio Grande do Sul".