The K-testing Paradox

Numerous flaws inherent to soil testing for exchangeable K (Exch-K) have been thoroughly documented on pages 6–11 of our paperReference Khan, Mulvaney and Ellsworth ², and leave no doubt that such testing is of no interpretive value for evaluating soil K buildup/depletion or as a basis for fertilizer K recommendations. This view could hardly be avoided considering that: (1) soil K test levels were extremely variable when assessed by a fixed protocol that involved 4 years of biweekly sampling, and increased substantially during this period despite the absence of fertilization (see figure 1 of our paperReference Khan, Mulvaney and Ellsworth ² ); (2) a far greater increase occurred for the Morrow Plots following 51 years of continuous corn (Zea mays L.) that removed almost 900 kg K ha⁻¹ without K fertilization (see table 1 of our paperReference Khan, Mulvaney and Ellsworth ² ); and (3) Exch-K was not sensitive to K addition or removal in 68 field trials throughout the world involving seven soil orders, 20 cropping systems and a wide range of ecological conditions (see table 2 of our paperReference Khan, Mulvaney and Ellsworth ² ). Bar-Yosef et al.Reference Bar-Yosef, Magen, Johnston and Kirkby ¹ agree that ‘an increase in Exch-K over time under zero K fertilization occurred’, cite further evidence in support of our interpretation and recognize the need to improve estimation of plant-available K by utilizing intensity and capacity factors. Regardless, they cling to the belief that ‘measurement of Exch-K is an essential and valuable tool and its use should be continued’. Such resistance to scientific reality surely qualifies as a paradox.

Figure 1. Relation of 1993 corn yield with K test, as reported by PeckReference Peck and Hoeft ¹² for a 40-acre (16-ha) field systematically sampled for both parameters using a 16 × 16 grid.

The Paradox of KCl Fertilization

If crop K uptake originates from huge K reserves throughout the soil profile, the question naturally arises as to whether producers can expect a profitable return from annual or biennial applications of KCl. The answer is a resounding NO, according to table S4 in the online supplement of our paper that compares yield data for NP and NPK treatments (P and PK for legumes) in more than 2100 short-term university field trials mostly conducted since fertilizer KCl became widely available in the early 1960s. The vast majority of these trials showed no profitable yield gain from KCl fertilization, and this was even more striking for grain production in North America, in which case KCl was of no agronomic value for 93% of 774 trials surveyed and more often led to yield loss than gain.

In our survey of fertilizer K response, considerable care was taken to avoid data sets confounded by growth-limiting factors such as diseases or drought, but apparently Bar-Yosef et al.Reference Bar-Yosef, Magen, Johnston and Kirkby ¹ are unwilling to face the prospect that KCl is often of no value to agricultural producers, and instead have resorted to obfuscation for justifying the intensive use of this fertilizer. Two long-term studies with K fertilization are cited to strengthen their case, one in England by Johnston et al.Reference Johnston, Warren and Penny ³ and another in Australia by Li et al.Reference Li, Helyar, Conyers, Cregan, Cullis, Poile, Fisher and Castleman ⁴ . In contrast to the survey reported in our paper, such studies are of no relevance to fields in a production setting, as yield differences between fertilized and unfertilized treatments reflect nutrient depletion that intensifies over time. Examination of figure 1 in Johnston et al.Reference Johnston, Warren and Penny ³ does not substantiate the interpretation of Bar-Yosef et al.Reference Bar-Yosef, Magen, Johnston and Kirkby ¹ , that ‘wheat and barley responded by enhanced grain yield to K application of 70 kg K ha⁻¹ in starved soils, but not on previously K-enriched plots’. In fact, Johnston et al.Reference Johnston, Warren and Penny ³ reported for both the Rothamsted Exhaustion Land study and the Woburn study that ‘the responses to new K fertilizer were too small and the yields on starved and enriched soils were too variable’. Interestingly, Bar-Yosef et al.Reference Bar-Yosef, Magen, Johnston and Kirkby ¹ neglected to mention that the K source for both sites had always been K₂SO₄, not KCl.

The study by Li et al.Reference Li, Helyar, Conyers, Cregan, Cullis, Poile, Fisher and Castleman ⁴ is no more useful than that of Johnston et al.Reference Johnston, Warren and Penny ³ for defending the value of soil testing for Exch-K and the need for K fertilization. We note, for example, that Exch-K was estimated by Li et al.Reference Li, Helyar, Conyers, Cregan, Cullis, Poile, Fisher and Castleman ⁴ using 0.1 M BaCl₂–0.1 M NH₄Cl rather than 1 M NH₄C₂H₃O₂, which has implications for extraction efficiency. More importantly, Li et al.Reference Li, Helyar, Conyers, Cregan, Cullis, Poile, Fisher and Castleman ⁴ were unable to establish a critical level in relating soil test K to dry matter or grain production by wheat (Triticum aestivum L.) in a mixed pasture, nor was there any meaningful relationship between plant K concentration and pasture yield despite soil K-supplying power inherently limited by a sodic subsoil in a semi-arid region where K fertilization is not generally recommended for cereal crops. Bar-Yosef et al.Reference Bar-Yosef, Magen, Johnston and Kirkby ¹ have misinterpreted both findings.

In their attempt to avoid the obvious evidence in supplementary table S4 that KCl is often ineffective for yield response, Bar-Yosef et al.Reference Bar-Yosef, Magen, Johnston and Kirkby ¹ are discounting a global database of more than 2100 field studies that encompass diverse cropping systems, tropical to semi-arid climates and a wide array of soils representing Mollisols (525 site-years), Alfisols (423 site-years), Ultisols (403 site-years), Inceptisols (261 site-years), Entisols (177 site-years), Spodosols (53 site-years), Oxisols (50 site-years), Aridisols (23 site-years), Vertisols (23 site-years) and Histosols (3 site-years). Instead, they prefer empirical modeling by Mueller et al.Reference Mueller, Gerber, Johnston, Ray, Ramankutty and Foley ⁵ that predicted the need for a 35% increase in K₂O usage to close the yield gap in underachieving areas; however, Bar-Yosef et al.Reference Bar-Yosef, Magen, Johnston and Kirkby ¹ failed to mention a major concern that the model was insensitive to soil and slope parametersReference Mueller, Gerber, Johnston, Ray, Ramankutty and Foley ⁵ . This is an important limitation, because yield responses to K fertilization predominately occur on soils inherently low in K-supplying power, as discussed on page 11 of our paper.

In discounting the adverse effects of intensive KCl fertilization on soil properties, Bar-Yosef et al.Reference Bar-Yosef, Magen, Johnston and Kirkby ¹ focus on hydraulic conductivity and aggregate stability without addressing the loss of cation-exchange capacity (CEC) as documented on page 18 of our paper in regard to cumulative KCl usage over decades. As justification, they cite short-term data from a pot study by Chen et al.Reference Chen, Banin and Borochovitch ⁶ and a column study by Levy and TorrentoReference Levy and Torrento ⁷ . The latter publication, and several others that could have been citedReference Quirk and Schofield ^{8 –} Reference Levy and van der Watt ¹⁰ , actually show a decrease in hydraulic conductivity rather than the benefits broadly claimed by Bar-Yosef et al.Reference Bar-Yosef, Magen, Johnston and Kirkby ¹ in concluding that ‘there is no possibility that KCl application adversely affects soil structure’. Critical examination of figure 1 in Chen et al.Reference Chen, Banin and Borochovitch ⁶ clearly refutes this conclusion, because Bar-Yosef et al.Reference Bar-Yosef, Magen, Johnston and Kirkby ¹ neglected to mention a marked decrease in hydraulic conductivity that occurred with one of the three soils studied. More importantly, data reported by Chen et al.Reference Chen, Banin and Borochovitch ⁶ show that the three soils decreased 4–13% in CEC as K saturation increased.

Despite extensive coverage given to the consequences of KCl fertilization for crop quality in a global survey of more than 1000 field studies (see pages 15–17 and table S5 of the online supplement), Bar-Yosef et al.Reference Bar-Yosef, Magen, Johnston and Kirkby ¹ consider this no more than ‘scant evidence’ and instead declare that there is ‘no evidence of a detrimental effect of potassium chloride (KCl) on crop yield or quality’. This claim cannot be reconciled with the references they cite to document the antagonistic effect of K on plant uptake of Ca and Mg, which has important ramifications for human and animal health. Bar-Yosef et al.Reference Bar-Yosef, Magen, Johnston and Kirkby ¹ consider Cl an essential nutrient for controlling plant diseases, while avoiding considerable evidence that Cl enhances bioaccumulation of Cd in food and feed.

Seven Decades of the K Paradox

Soil K testing and commercial K fertilization originated from work in the 1940s by Dr Roger Bray at the University of IllinoisReference Bray ¹¹ , utilizing static plot trials with and without K fertilization for up to four decades. Even after this prolonged period of soil K depletion, BrayReference Bray ¹¹ was unable to establish a direct relationship between the Exch-K test level and corn yield, and the same reality was apparent five decades later in a 1993 study by PeckReference Peck and Hoeft ¹² (Fig. 1). The latter finding exposes the futility of current K management, considering that: (1) crop K removal reflected much higher yields in 1993 than in the 1940s; (2) more than half of the test values were below 300 lb per acre (336 kg ha⁻¹), which is supposed to be the critical threshold for yield reductionReference Fernández, Hoeft and Nafziger ¹³ ; and (3) no KCl had been applied in the previous 6 years. The reality represented by Figure 1 is consistent with the overwhelming majority of K response trials we surveyed for North America, and is further substantiated by Figure 2, which shows no evidence of a relationship between annual K fertilizer consumption and crop K removal (estimated from yield data) in Illinois since 1950. This disparity is reflected in fertilizer K inputs that have exceeded crop K removal by almost 60,000 Mg yr⁻¹ on average for the past 40 years, representing an unsound investment at current prices of US$60 million per year in the purchase of KCl by Illinois producers. In reality, the annual loss would easily total several hundred million dollars, considering both direct and indirect costs and the widespread lack of K response cited previously for grain crops in North America.

Figure 2. Illinois data for fertilizer K inputReference Fernández, Hoeft and Nafziger ^{13 ,} Reference Brown ¹⁴ and K removal in corn, soybean (Glycine max L. Merr.), wheat, oats (Avena sativa L.), barley (Hordeum vulgare L.), rye (Secale cereale L.) and hay, plotted with lines that represent a 5-yr moving average. Crop K removal was estimated using composition data ^{15 ,} Reference Mengel, Barker and Pilbeam ¹⁶ in conjunction with online production records ^{17 , 18} .

As documented by our supplementary table S4, KCl consumption can often be decreased without sacrificing yield, but this opportunity has not been realized because of buildup–maintenance recommendations that intensify fertilizer consumption at the expense of agronomic uptake efficiency and thus accentuate the economic interests of the fertilizer industry over those of the producer. These recommendations have been effectively presented through ergonomic diagrams such as figure 8.8 in the Illinois Agronomy Handbook Reference Fernández, Hoeft and Nafziger ¹³ , and are widely publicized by the extension community as a best management practice. Thus continues the K paradox.