Phosphorus extractants for soils in the humid tropical region of Brazil

ABSTRACT Given the heterogeneity of soils in Roraima state, Brazil, and the need for a phosphorus (P) extraction protocol, this study aimed to assess the efficiency of the Mehlich-1, Mehlich-3, Bray-1 extractants and anion exchange resin (AER) in determining available P in soils representative of the state. The chemical, physical and mineralogical attributes of seven soil classes were determined. The experiment was conducted in a greenhouse, using a randomized block design in a 7 x 5 factorial scheme with four repetitions. The first factor was the soils (LA (ITÃ), LV (ITÃ), LA (SERRA DA PRATA), LA (CCA), PA (CCA), FT (BONFIM), SN (SURUMU)), and the second P doses, estimated based on the maximum P-adsorption capacity (MPAC). A polyethylene pot containing 5 dm3 of soil and five plants was defined as the experimental unit. The indicator species was corn grown in two 30-day phases. The P doses were applied using monopotassium phosphate. Soil samples were removed before planting to determine available P using the extractants. Corn dry weight was assessed at the end of the two-phase experiment. Correlation tests were performed for dry weight versus extractant and extractant versus extractant. The Mehlich-3 and Bray-1 extractants were the most sensitive to variations in soil attributes. The resin exhibited the greatest correlations with the different soils individually and combined, showing potential in determining available P in soils from the humid tropical region of the Brazilian Amazon.


INTRODUCTION
Soil available phosphorus (P) is typically evaluated using acid extractants, such as Mehlich-1, Mehlich-3 and Bray-1.In addition to chemical extractants, anion exchange resins (AERs) are also used in some Brazilian laboratories.The choice of extractant to determine available P depends on the degree of correlation between the P content of the extractant and that of the plant, whereby the greater the correlation, the more suitable the extractant (STEINER et al., 2012).However, there is no consensus regarding the extractant that best expresses plant-available P, largely due to soil attributes that infl uence adsorption of this element (NUNES et al., 2021;ROSA;SILVA;MALUF, 2018;WUENSCHER et al., 2015;ZEHETNER et al., 2018).
Acid extractants, such as Mehlich-1, facilitate the acquisition of clear liquids by decantation, enable lowcost analyses and are more feasible for routine use in laboratories (ZEHETNER et al., 2018).These extractants act diff erently depending on soil type because of their diff erent capacities to dissolve and solubilize P from phosphate compounds (NUNES et al., 2021).
Mehlich-1 is restricted for use in clayey soils, especially when phosphates that are poorly soluble in water have been applied, due to excessive P-Ca extraction, which is unavailable to plants (MEDEIROS et al., 2021;MUMBACH et al., 2018).Additionally, the extractant may underestimate available P in clayey soils with a high pH because of extractant exhaustion resulting from high cation exchange capacity and ineffective extraction of P bound to Fe (P-Fe) and Al (P-Al) (ROY, 2017;VINHA et al., 2021).
Bray-1 is recommended for soils with a wide range of chemical characteristics due to Al 3+ fl uoride complexation (DARI et al., 2019;ZEHETNER et al., 2018).For acid soils, Mehlich-3 is more widely used for available P extraction because it is more economical and versatile than Bray-1 in that it can simultaneously extract other nutrients in addition to P (NUNES et al., 2021).However, Penn et al. (2018) used this method and observed a decline in P extraction with increasing pH due to changes in P forms and the consumption of extractant fluoride by calcium minerals.
AERs have a number of favorable characteristics, particularly the high correlation between P uptake by the plant and that extracted from the soil and better theoretical basis for determining soil P content (SILVA;RAIJ, 1999).These resins simulate the behavior of root systems at P uptake and do not affect natural soil chemistry (GONÇALVES et al., 2012).
Results diff er in terms of extractant effi ciency in determining soil-available P and demonstrate the importance of these studies in establishing an extraction method for each region (DARI et al., 2019;MEDEIROS et al., 2021;RAIJ;FEITOSA, 1980;SANTOS;KLIEMANN, 2005;SILVA;RAIJ, 1999;STEINER et al., 2012;WUENSCHER et al., 2015).Moreover, although most research has concluded that ARTs exhibit the best correlation with Brazilian soils, chemical extractants have been used to estimate available P in humid tropical regions of the country.
As such, given the diff erent fi ndings obtained in establishing a P extractant and the small number of studies on this topic under humid tropical conditions in Brazil, this investigation aimed to assess Mehlich-1, Mehlich-3, Bray-1 and AER in determining available P in seven soil classes representative of Roraima state.

MATERIAL AND METHODS
Seven soil classes were selected (Table 1) based on their agricultural importance to Roraima and classifi ed according to the Brazilian Soil Classifi cation System (EMBRAPA, 2018).Samples of each soil type were collected from the 0-0.20 m layer for chemical, physical, mineralogical and P adsorption analyses.

Soil chemical, physical and mineralogical analysis
Chemical analysis of the soils consisted of pH in water, calcium, exchangeable magnesium and aluminum in 1 mol L -1 KCl, available phosphorus and potassium using Mehlich-1, potential acidity with 0.5 mol L -1 of calcium acetate and organic matter by determining organic carbon in 0.2 mol L -1 potassium dichromate, in accordance with Embrapa (2009).The soil attributed sum of bases -SB (Ca 2+ + Mg 2+ + K + + Na + ), total cation exchange capacity -T (SB + H + Al), eff ective cation exchange capacity -t (SB + Al), clay activity index (T/clay x 1000), base saturation -V% (SB/T x 100) and aluminum saturation -m% (Al 3+ /t x 100) were obtained from the results of the abovementioned analyses.The soil characterizations LA (CCA) and PA (CCA) were taken from Benedetti et al. (2011) and SN (SURUMU) from Schaefer et al. (1993).The results of these analyses are presented in Table 2.
Physical characterization consisted of determining the texture class of the soils based on granulometry, using the pipette method (EMBRAPA, 1997) and the soil texture triangle of the Brazilian Soil Classifi cation System -SiBCS (EMBRAPA, 2018).The silt-to-clay ratio (SCR) was calculated by dividing the two soil fractions.The results are shown in Table 3.
Mineralogical analysis of the sand, silt and clay fractions (Table 4) was performed via X-ray diffraction (XRD), by adding around 100 mL of water Phosphorus extractors for soils in the humid tropical region of Brazil    ---------------------------g kg -1 --------------------------- (1993) and 10 mL of 1 N NaOH solution to soil samples of approximately 100 g.After 24 hours, the samples were broken up (disaggregation) in a mechanical shaker for 10 min.The fractions were separated using 0.053 mm mesh sieves for sand and silt and 0.002 mm for silt and clay, with constant washing to remove any traces of NaOH that might infl uence the diff ractograms.After drying at ambient temperature, the soil fractions were macerated and fi xed on glass slides with mineral oil for sand, and water for silt and clay.The X-ray diff ractograms were obtained using a Shimadzu diff ractometer with a cobalt tube, in a range of 7 to 70º and scan speed of 0.02º/sec.

Preparing the experimental unit
A 5 kg sample of each soil class, collected from the 0-0.20 m layer, was previously sieved (4 mm mesh) and placed in 5 L polyethylene pots.Soil chemical analysis indicated the need for liming to raise base saturation to 60% for corn cultivation (RIBEIRO; GUIMARÃES; ALVAREZ, 1999).The soils were incubated with CaCO 3 (TNP 100%) and irrigated daily with distilled  water to maintain 70% field capacity.Incubation ended after 20 days, when the soils reached pH greater than 6.
Corrective fertilization was performed for N, K, S and Zn for the corn crop, in accordance with Ribeiro, Guimarães and Alvarez (1999).
The P doses for each treatment in the diff erent soil types was incorporated into the soil in the pots in the form of a monopotassium phosphate (KH 2 PO 4 ) solution.
Five days later, 200 g of soil was collected from each experimental unit using a Dutch auger to remove soil across the entire profi le.The samples were taken to the drying area and prepared for chemical analysis.

Greenhouse experiment
Corn (Zea mays L.) was cultivated in two successive 30-day cycles, totaling 60 days.The 30F35YH corn hybrid was used (Pioneer.Brazil).Eights seeds were planted in each experimental unit, leaving fi ve plants per pot after thinning.
Topdressing containing N and K in the form of solution was applied 12 days after emergence in both cycles (RIBEIRO; GUIMARÃES; ALVAREZ, 1999).No pest control was needed.Irrigation was performed daily with distilled water to maintain fi eld capacity at 70%.The plants were cut at ground level 30 days after emergence, with replanting fi ve days after the fi rst phase.The plant material collected in both cycles was placed in paper bags and dried in a forced air oven at 60 ºC until constant weight and its mass measured for use in correlation analyses with P extracted via the Mehlich-1, Mehlich-3, Bray-1 and AER methods.

Statistical analysis
The linear correlation coeffi cient was determined for extractant x dry weight and extractant x extractant, considering the soils individually and grouped according to the amount of P extracted with the increase in dry weight (SANTOS; KLIEMANN, 2005).All the parameters were compared using the t-test at 5% probability and classifi ed according to Larson and Farber (2015).

Relationship between the P doses applied and P extracted using the diff erent methods
The relationship between the P doses applied to the soils and the maximum P content extracted by Mehlich-1, Mehlich-3, Bray-1 and AER was explained by linear functions.The maximum P contents extracted, angular and linear coeffi cients and coeffi cients of determination of each function are shown in Table 6.
With the exception of NP, which exhibited the highest angular coeffi cient, AER was the method that extracted the most P from the soils, with 373.50 mg dm -3 in LA (ITÃ), 427.25 mg dm -3 in LV (ITÃ), 286.75 mg dm -3 in LA (SERRA DA PRATA) and 395.50 mg dm -3 in LA (CCA) (Table 6).The factors that most infl uenced this variation were MPAC, soil texture and pH.Due to the increase in pH from liming, part of the P in the soil bonds to calcium (PENN; CAMBERATO, 2019), a non-labile form that is not available to plants (MATOS et al.,  SILVA; RAIJ, 1999;VINHA et al., 2021).AER cannot extract this form of P, justifying the smaller amounts extracted in LA (SERRA DA PRATA) when compared to the other yellow latosols with similar MPAC.In Itã soils, AER was the most sensitive to texture variations in latosols (Table 6), with a lower available P content in sandy clay loam (373.50 mg dm -3 ) and higher in clayey soil (427.25 mg dm -3 ), as also reported by Santos and Kliemann (2005) and Mumbach et al. (2018Mumbach et al. ( , 2020)).Mehlich-3 extracted similar values to those observed for AER at most doses (Table 6), with maximum values of 323.35 mg dm -3 in LA (ITÃ), 288.68 mg dm -3 in LV (ITÃ), 246.40 mg dm -3 in LA (SERRA DA PRATA) and 351.79 mg dm -3 in LA (CCA).Both methods extract similar P forms in these classes, particularly those bound to Fe and Al (ROY, 2017).Mumbach et al. (2018) observed similar P extraction effi ciency with Mehlich-3 and AER in four diff erent soil texture classes.Steiner et al. (2012) reported greater extraction capacity for Mehlich-3 than AER, with a strong correlation between the P extracted and that absorbed by soybean.Nunes et al. (2021) studied P extractants in yellow latosol and haplic gleysol and found that although AER was a more effi cient extractant, Mehlich-3 is less costly and faster, justifying its use.Dari et al. (2019) observed greater extraction capacity for Mehlich-3 compared to other extractants in alkaline soils from Idaho in the United States.
The Mehlich-1 and Bray-1 extracted the least P from latosols (Table 6), with values of 197.04 and 293.36 mg dm -3 , respectively, in LV (ITÃ).The clayey texture of LV (ITÃ), concomitant to the presence of Fe-and Al-bound P, compromised the efficiency of Mehlich-1, which has a low extraction capacity for these P forms.This result is corroborated by other authors who analyzed the P extraction efficiency of Mehlich-1 (MEDEIROS et al., 2021;NUNES et al., 2021).On the other hand, Bray-1, which preferentially extracts Fe-and Al-bound P (RAIJ;FEITOSA, 1980;SILVA;RAIJ, 1999), showed increased extraction.In LA (ITÃ) without hematite (iron oxide), Mehlich-1 exhibited greater extraction than Bray-1. Wuenscher et al. (2015) obtained different results, with Bray-1 performing better is soils from Central Europe than the other chemical methods studied.Dari et al. (2019) reported that Bray-1 exhibits limited P extraction in alkaline soils from Idaho, unlike the acidic soils studied here.
In sandy soils with a clay content lower than 150 g kg -1 in the surface horizon, extraction was infl uenced by factors such as MPAC, mineralogy and soil pH (Table 3, 4 and 6).In FT (BONFIM), which has the highest MPAC of the soils studied and contains 2:1 clay, extractants based on strong (Mehlich-1) or weak acids (Mehlich-3 and Bray-1) are hampered by the high cation exchange capacity of these soils, thus exhausting the extractants due to the consumption of anions, sulphates and fl uorides by aluminum or calcium present in the soil not bound to P (GONÇALVES et al., 2012).As a result, P extraction using these methods was inferior to AER, with Phosphorus extractors for soils in the humid tropical region of Brazil maximum values between 134.20 and 355.67 mg dm -3 for the acid extractants and 852.50 mg dm -3 for AER, which simulates the P uptake behavior of the root, extracting large amounts of the element through ligand exchange between bicarbonate and dihydrogen phosphate (SILVA; RAIJ, 1999;VINHA et al., 2021).
PA (CCA) and SN (SURUMU) exhibited similar maximum P extraction values for the diff erent methods tested due to their similar clay contents and basically kaolinitic mineralogy (Tables 3, 4 and 6).Thus, phosphate in its diff erent forms has lower binding energy with its active sites, making it easily dissociated even by extractants with a low dissociation capacity for P-Al, P-Fe (Mehlich-1) and P-Ca (Mehlich-3 and Bray-1).The maximum P extracted remained below 75.77mg dm -3 for Mehlich-3 in PA (CCA) and 14.39 mg dm -3 for Mehlich-1 in SN (SURUMU).Similar results were reported by Raij and Feitosa (1980) in red podzolic soils (red argisols according to SiBCS (EMBRAPA, 2018)) with 14% clay.Rosa, Silva and Maluf (2018) reported high extractable P with the addition of humic acids to quartzarenic neosols containing 4% clay, due to the increase in P adsorption sites.In turn, Mumbach et al. (2020) reported superior extraction power for Mehlich-1 when compared to Mehlich-3 in soils with diff erent clay contents.

Correlation between the amount of P extracted via the diff erent methods and corn dry weight produced in the diff erent soil classes
In this section, the soils were assessed individually and together considering the amount of P extracted with the increase in dry weight.Correlations between P extracted by Mehlich-1, Mehlich-3, Bray-1 and AER and corn dry weight produced in the soil classes individually and grouped together according to the amount extracted are shown in Tables 7 and 8.
Considering each soil type individually, the amount of P extracted by all the methods shows a strong correlation with the dry weight produced (Table 8).Strong correlations (> 0.78) were observed between the extractants and dry weight in the medium-textured and clayey soils.Bray-1 exhibited greater amplitude between soil correlations, possibly due to its increased sensitivity to iron and aluminum oxide, clay and calcium contents (DARI et al.;2019;ZEHETNER et al., 2018).Matos et al. (2021) found that the combination of diff erent soil chemical, physical and mineralogical characteristics in humid tropical regions may explain the high correlations and small amplitudes between soil classes.
Despite the superior P extraction power of Mehlich-3 when compared to Mehlich-1, the correlation with dry weight was similar for all the latosols, indicating that both methods are suitable for estimating available P in this soil class (Tables 7 and 8).Several authors have reported similar results (MUMBACH et al., 2018;NUNES et al., 2021;STEINER et al., 2012).Dari et al. (2019) concluded that Mehlich-3 is most adequate in the correlation between soil Considering the soils with a sandy textured surface (PA (CCA), FT (BONFIM) and SN (SURUMU)), AER exhibited a higher correlation when compared to the other extractants, with a value of 0.90 for these three soil classes (Tables 7 and 8).With the exception of AER, similar correlations were observed for all the extractants, ranging from 0.72 to 0.80 to Mehlich-1, 0.72 to 0.79 for Mehlich-3 and 0.67 to 0.78 for Bray-1.The high CV% of the data obtained for these extractants demonstrates the variable extraction capacity for a same P dose applied to the soil, compromising the correlation between these data with dry weight.On the other hand, Wuenscher et al. (2015) studied 14 P extraction methods in 50 soils from Central Europe and observed a greater correlation for Mehlich-3 and Bray-1 when compared to AER.Culman et al. (2020) also observed a strong correlation between Mehlich-3 and Bray-1 and dry weight, although the former extracted 35% less P than the latter.
In a bibliographic review of extractant efficiency analyses, Silva and Raij (1999) found that AER was the most suitable method for a wide variety of soils.Several recent studies have reported similar findings (MEDEIROS et al., 2021;MUMBACH et al., 2018;NUNES et al., 2021).By contrast, the results obtained here differ from those of other authors who used AER (DARI et al.;2019;MUMBACH et al., 2020), which indicates a unique dynamic for phosphate in terms of the heterogeneity of the physical, chemical and mineralogical properties of Brazilian soils and the importance of these studies in different regions.
With regard to soils grouped according to the amount of P extracted in the diff erent classes (Table 8), Group 1 showed a strong correlation only for AER (0.73), with lower values for the remaining extractants (0.64, 0.55 and 0.60 for Mehlich-1, Mehlich-3 and Bray-1, respectively).For Group 2, a strong correlation was observed for all the extractants (0.83 to 0.92), with the highest value recorded for AER.According to Santos and Kliemann (2005), the factors that influence this correlation are unclear, since soils with different characteristics that affect extraction are grouped together.However, the authors defend clustering as a means of validating the correlation indices, given the increased number of points in determining the linear correlation for each group.These results make AER an adequate extractant for determining available P considering the soil classes separated by common characteristics, such as clay content, OM and total cation exchange capacity (T) or the amount of P extracted.Given the difficulties involved in determining fertilization levels for each soil class and its variations, in order to be feasible an extractant should exhibit a correlation with soil groups as opposed to a single class.
The correlations between Mehlich-1, Mehlich-3, Bray-1 and AER for each soil class are presented in Table 9. acidity of Mehlich-3 indicates that this extractant is not always effi cient at extracting P bound to Ca, with reasonable amounts of the latter element present in the soil in question.These fi ndings are corroborated by Penn et al. (2018).The fl uoride ion present in both solutions forms a complex with Al 3+ in particular and, to a lesser extent, Ca 2+ , releasing phosphate which precipitates in the form of calcium fluoride (MEDEIROS et al., 2021;ZEHETNER et al., 2018).Despite the presence of fluoride in both extractants, its higher concentration in Bray-1 (0.03 N NH4F) in relation to Mehlich-3 (0.015 N NH4F) maintained the complexing power of Ca 2+ and Al 3+ in the former, an important feature in soils with a high concentration of expandable clay minerals, which consume a large part of the fl uoride ions due to their elevated T. Diff erent results were reported by Culman et al. (2020), with a strong correlation between extraction by Mehlich-3 and Bray-1 in alkaline soils from Ohio, Indiana and Michigan.Mehlich-1, in turn, preferentially removes P-Ca due to the presence of two strong acidic solutions, namely 0.05 mol L -1 HCl and 0.0125 mol L -1 H 2 SO 4 , with Mehlich-3 extracting higher values in soils containing P-Ca (MEDEIROS et al., 2021).
Correlations between the chemical extractants and AER were low for SN (SURUMU).The selectivity of chemical extractants means they remove phosphate bound to diff erent cations in labile and nonlabile forms (SOUZA; PEGORARO;REIS, 2017;ZEHETNER et al., 2018), whereas AER uses a similar process to that of roots, via porous material with a high bicarbonate-form anion exchange, responsible for P dissolution and subsequent adsorption to AER (GONÇALVES et al., 2012;SILVA;RAIJ, 1999).In less acidic soils with low T, chemical extractants may overestimate plant-available P, extracting nonlabile forms present in the soil.Under these conditions, AER is better suited to extracting P amounts consistent with those available to plants.

CONCLUSIONS
1. Extraction by Mehlich-1, Mehlich-3, Bray-1 and RTA was strongly correlated with the dry weight produced in the latosols and argisols analyzed, with superior results for AER in FT (BONFIM) and SN (SURUMU);

Table 1 -
Soil classes of the samples collected in the 0-0.20 m layer for P extraction using diff erent extractants

Table 2 -
Chemical characterization of soil samples (0-0.20 m layer) from seven classes for P extraction using diff erent extractants

Table 6 -
Coeffi cients of the linear equations for P extraction by Mehlich-1, Mehlich-3, Bray-1 and AER for the diff erent P doses applied to the soil classes 2021;C.H. L.Matos et al.

Table 8 -
Coeffi cient of linear extractant x dry weight correlation for each soil individually and grouped according to the amount of P extracted