Assessing the predictive capability of N, P, and B diagnosis in cotton crop

Abstract

The compositional nutrient diagnosis—CND method is a standard tool for evaluating plant nutritional status. Adjustments are crucial to elevate accuracy. The effectiveness of such methodological refinements should be rigorously assessed through accuracy tests that are benchmarked against the prescient diagnostic analysis—PDA methodology. The objective of this investigation was to refine the CND technique for a more precise evaluation of N, P, and B nutrient status in cotton. The study’s database encompasses 144 data points pertaining to crop yield and foliar nutrient concentrations from cotton plantations in the Cerrado biome of Brazil. Subsequently, the CND norms were established through rigorous calibration. Three separate nutrient-dose trials, each featuring four levels of N, P and B, were carried out to assess plant true nutritional status. Adjustments were made to the nutrient responsiveness range—NRr (0.5 and 1.0), while yield response—YR were scrutinized at threshold levels (5% and 10%). The prerequisites for achieving high diagnostic accuracy were nutrient specific. For N, maximal accuracy was linked only to the YR parameter (YR = 10%). For P, the most precise outcomes were attained with a NRr = 0.5 and YI = 5%. For B, highest diagnostic accuracy when the NRr = 1.0 and YI = 10%. These insights highlight the need to fine-tune the CND method for reliable nutritional evaluations and cotton crop productivity optimization.

Introduction

Brazil holds a critical position in the global cotton industry, being the fourth-largest producer and the second-largest exporter of cotton (Gossypium hirsutum L.). In the 2022–2023 season, the country produced 4.5 million tons of cottonseed¹.

Several factors play a crucial role in the substantial production of cotton. Notably, the macronutrients nitrogen and phosphorus, along with the micronutrient boron, have been identified as the primary factors limiting cotton productivity^2–4. However, developing effective fertilization programs for cotton cultivation necessitates the establishment of reliable criteria for determining the necessary quantities of nutrients to be applied through fertilization.

Soil fertility testing is a widely adopted method for assessing the nutrient content in the soil and how available these nutrients are for plant absorption. Conversely, evaluating the foliar chemistry of crops allows for the analysis of the interaction between the soil's nutrient supply capacity and the plant’s nutrient uptake. Currently, crop nutritional status assessment can be conducted using various diagnostic methods, such as univariate, bivariate, and multivariate approaches⁵.

The CND method⁶, employs a multivariate approach to evaluate the interrelationships between individual nutrients and the geometric mean of the sample. This method allows for ranking nutrients based on their limitations caused by deficiencies or excesses⁷ as well as for to generate a well-defined covariance matrix and calculate mutually exclusive concentration proportions^7,8.

One aspect of the CND method is their potential to identify nutritional patterns without requiring experimental assays, compared to conventional methods. Instead, the pattern can be obtained through the monitoring of commercial crops, leading to significant savings in terms of time and financial resources^7,9. Moreover, the through the monitoring of commercial crops enables the modeling of a considerably larger number of nutritional and environmental interactions compared to what could be achieved through conventional experimental assays¹⁰.

By employing this methodology, nutritional standards specific to the CND method were established for cotton cultivation in the state of Bahia, Brazil¹¹, as well as in the states of Mato Grosso and Mato Grosso do Sul¹².

However, attaining a high-quality diagnostic process cannot solely rely on nutritional standards obtained through this approach. The quality of diagnoses can be influenced by variations in calculation formulas or different nutritional standards. To address this, was introduced the PDA method¹³. PDA compares the actual nutritional status of the crop, determined by analyzing the plant's productive response to nutrient supply, with the diagnoses generated by the nutritional status evaluation method. This allows for the estimation of the accuracy and performance of the nutritional diagnostic process.

Despite the abundance of literature evaluating various diagnostic methods, only a small fraction is dedicated to assessing the true nutritional condition of crops. This evaluation has been conducted in studies on soybean^14–16, banana^17,18, eucalyptus seedlings¹⁹, and sugarcane^20,21. Additionally, utilizing APD allows for more detailed analyses of the criteria used in nutritional diagnostic methods (NRr) or APD framework (YR).

For example, in sugarcane cultivation, the APD method was employed to evaluate the NRr criteria used in nutritional diagnoses through the responsiveness to fertilization method²². It was found that adjusting the NRr during diagnosis can significantly impact the accuracy of the results, as it can alter the classification of the nutritional status, making them more or less reliable (20, 21). Similarly, research on soybean cultivation in both field-scale operations and experimental plots supports these findings¹⁶. Furthermore, it should be emphasized that optimizing YR variables within the APD framework is essential for a precise assessment of the True Nutritional Status (TNS) of the plant, thus enhancing the method's overall precision^14,16.

The APD method compares the TNS with the nutritional diagnosis to assess its accuracy^13,14. The YR variables help determine the plant’s TNS by measuring its productive response to fertilization, allowing for the classification of the plant as responsive and, consequently, in a state of deficiency. Initially, a 10% response level was set by the author¹³ to classify a plant as having a true nutritional deficiency. However, studies have shown the need to adjust this level depending on the nutrient being analyzed, as macro and micronutrients induce different responses in the plant¹⁶.

Consequently, it can be hypothesized that modifications in NRr could influence the accuracy of nutritional status interpretation through the CND and YR parameters technique could influence the ability to define the plant's TNS. Together, these criteria could increase the efficiency of nutritional diagnoses of the agricultural crops.

To test this hypothesis, this research was conducted to investigate the potential impact of alterations in NRr and YR parameters on the accuracy of N, P, and B nutritional status assessment using the CND technique in cotton.

Results

Among the 144 commercial crops analyzed, 22% were classified as high yielding, with productivity exceeding 5163 kg ha⁻¹ (average plus 0.5 times the standard deviation of the dataset’s productivity). These crops demonstrated a normal distribution, as confirmed by the Shapiro–Wilk test, and were used to establish the CND norms for essential nutrients such as N, P, K, Ca, Mg, S, B, Cu, Fe, Mn, and Zn, as presented (Table 1).

Table 1.

CND norms for N, P, K, Ca, Mg, S, B, Cu, Fe, Mn and Zn from leaf samples of cotton grown in Chapadão do Sul, Brazil.

Norms1	N	P	K	Ca	Mg	S
Mean	3.10	1.27	2.24	3.06	0.96	1.14
SD	0.28	0.18	0.14	0.10	0.36	0.18

	B	Cu	Fe	Mn	Zn
Mean	− 3.43	− 6.15	− 1.86	− 2.80	− 3.95
SD	0.25	0.68	0.22	0.53	0.42

¹Mean and standard deviation (SD) of multivariate relationship in reference crops (high yield). All values expressed in g kg⁻¹.

On average, the concentrations of P, Ca, S, B, Zn, Fe and Cu were higher in the reference subpopulation compared to the low-yield subpopulation. Conversely, Mg had a lower average concentration in the reference subpopulation. There were no significant differences in the average concentrations of N, K and Mn between the two subpopulations. The standard deviation of leaf nutrient concentrations was similar for most nutrients in both subpopulations, except for Fe and Zn, which displayed narrower distributions in the reference subpopulation (Table 2).

Table 2.

Mean and standard deviation (SD) of leaf nutrient content and productivity obtained in the set of low and high productivity plants and in the total of commercial cotton crops, Chapadão do Sul, Brazil.

Nutrient	Low yield		High yield		Total cotton crops1		Test	SR2
	Mean	SD	Mean	SD	Mean	SD	F (10%)
g kg−1
N	34.5a	4.7	33.9a	6.8	34.4	6.8	0.69ns	39.1–43.2
P	5.8a	1.5	5.5b	1.6	5.7	1.6	0.93ns	2.3–2.8
K	13.9a	1.7	14.1a	2.2	14.0	2.2	0.76ns	13.7–18.2
Ca	34.3a	7.1	32.4b	7.0	33.9	7.0	1.02ns	16.9–20.8
Mg	3.9b	0.7	4.0a	0.9	3.9	0.9	0.74ns	2.7–3.7
S	4.9a	0.8	4.7b	0.8	4.9	0.8	1.01ns	3.8–5.4
mg kg−1
B	50.0a	14.5	49.3b	14.2	49.9	14.2	1.02 ns	28.0–38.0
Cu	5.3a	2.7	4.2b	3.0	5.1	3.0	0.92 ns	7.0–11.0
Fe	287.2a	150.6	246.6b	112.4	278.7	112.4	1.34*	50.0–70.0
Mn	89.5a	51.0	106.0a	72.1	92.9	72.1	0.71 ns	44.0–65.0
Zn	39.8a	35.2	34.7b	27.8	38.7	27.8	1.27*	17.0–27.0
Yield (kg ha−1)	4182.9	661.8	6166.2	1162.3	4599.0	1162.3	0.57 ns	–

¹High-yielding and low-yielding croplands, as well as the aggregate of all croplands, were evaluated.

²The sufficiency range recommended by Kurihara et al. (2013). Distinct letters between the means of high-yielding and low-yielding croplands indicate a statistically significant difference as determined by the t-test (p = 0.10).

Annotations: ‘ns’ denotes non-significant, ‘*’ indicates significance as per the F-test (p = 0.05). Means followed bya common letter are not significantly different by the t-test at the 5% level of significance.

In the experimental trials, the application of 147 kg of N per hectare to the soil resulted in the highest cottonseed yield, reaching 5601 kg ha⁻¹. This value signified a 22% increase in productivity compared to the lowest N application rate. When N applications exceeded the ideal rate, productivity was negatively impacted, while the leaf N concentration continued to exhibit a linear increase (Fig. 1a).

Figure 1.

Productivity and leaf content of nitrogen (a); phosphorus (b) and boron (c), depending on the application of the respective nutrients to the soil, in Chapadão do Sul, Brazil. **significant at the 5% probability level (p ≤ 0.05 from the F test).

The highest yield observed was 5601 kg ha⁻¹, and 90% of this value amounts to 5041 kg ha⁻¹. By analyzing the correlation between leaf N content and 90% of the yield, a calibrated sufficiency range (calibrated SR) was established with lower and upper limits of 25 and 45 g kg⁻¹ leaf N, respectively. The optimal N level was identified as 39 g kg⁻¹ leaf N (Fig. 1a).

The application of phosphate fertilizer at a rate of 67 kg P₂O₅ per hectare resulted in the highest cottonseed yield, reaching 4020 kg ha⁻¹, which represented a 12% increase compared to the lowest dosage (Fig. 1b). This rate yielded a leaf P concentration of 4 g kg⁻¹. Plant productivity declined when P applications exceeded this rate, even though the leaf P concentration kept increasing in proportion to the applied dosage (Fig. 1b).

Considering the highest recorded yield of 4020 kg ha⁻¹, 90% of this value translates to 3618 kg ha⁻¹. Through an analysis of the relationship between leaf P content and 90% of the yield, the calibrated SR was established with lower and upper limits of 2.6 and 4.8 g kg⁻¹ leaf P, respectively. The optimal leaf P level was determined to be 4.0 g kg⁻¹ P (Fig. 1b).

Unlike the macronutrients N and P, B fertilization did not reach maximum productivity within the range of applied dosages, as productivity increased directly in proportion to the dosages applied (Fig. 1c). Consequently, the SR for B was not established.

The leaf B concentration exhibited a similar pattern of proportional increase in response to the applied B dosage in the soil. The highest B dosage applied, which was 4 kg ha⁻¹, resulted in a leaf B concentration of 52 mg kg⁻¹ and a peak cottonseed productivity of 6679 kg ha⁻¹, representing a 40% increase in productivity compared to the lowest dosage (Fig. 1c). Although cotton crops exhibited an average response to N, P and B fertilization, the concentrations of P, Ca, Mg, B, Zn, Fe and Mn, surpassed the upper limit of the recommended SR in the literature¹². In contrast, average concentrations of N and Cu fell below the SR’s lower limit. Only K and S concentrations were found to align within the established SR limits (Table 2).

When examining the TNS of N, it was found that 40% of the experimental plots were found to be responsive to N (true deficiency), based on a criterion of a YR above 5% (Table 3). The largest positive response to N addition took place between dose 2 and dose 3, while the smallest increase was noted between dose 3 and dose 4. Utilizing a criterion of a 10% YR, only 33% of plots were deemed responsive to N. For treatments ranging from dose 1 to dose 2 and dose 2 to dose 3, the proportion of responsive plots stood at 40%, with the lowest proportion observed from dose 3 to dose 4 (Table 3).

Table 3.

Comparison of yield and nutritional status in test and reference plots: nitrogen fertilizer effects assessed using Compositional Nutrient Diagnosis—CND in Cotton Leaf Samples from Chapadão do Sul, Brazil.

Blocks	Trial	PT	PR	YIc%	TNS		CND
					YI		NRr
					5%	10%	0.5	1.0
1	N2-N1	5939	3368	− 43%	TS	TS	INS	INS
	N3-N2	3368	4159	23%	TD	TD	BAL	BAL
	N4-N3	4159	4014	− 3%	TS	TS	BAL	BAL
2	N2-N1	4540	6354	40%	TD	TD	INS	INS
	N3-N2	6354	5731	− 10%	TS	TS	BAL	BAL
	N4-N3	5732	5418	− 5%	TS	TS	BAL	BAL
3	N2-N1	4361	5889	35%	TD	TD	INS	INS
	N3-N2	5889	6212	5%	TD	TS	BAL	BAL
	N4-N3	6212	3474	− 44%	TS	TS	BAL	BAL
4	N2-N1	5731	5479	− 4%	TS	TS	BAL	BAL
	N3-N2	5479	4960	− 9%	TS	TS	BAL	BAL
	N4-N3	4960	6888	39%	TD	TD	BAL	BAL
5	N2-N1	5341	4253	− 20%	TS	TS	BAL	BAL
	N3-N2	4253	6201	46%	TD	TD	BAL	BAL
	N4-N3	6201	4251	− 31%	TS	TS	BAL	BAL
TD					40%	33%
INS							20%	20%

Doses for nitrogen tests: N1 = 0 kg ha⁻¹, N2 = 63 kg ha⁻¹, N3 = 125 kg ha⁻¹ e N4 = 250 kg ha⁻¹.

PT plot test, PR plot reference, YIc% yield increases change (in percentage)—denotes the change in yield associated with each trial conducted, TNS true nutritional status, YI yield increases, NRr nutrient responsiveness range, TD true deficiency, INS diagnosis of insufficient, BAL balanced, TS true sufficiency.

On the other hand, when using the CND method for N diagnosis, only 20% of the experimental plots were determined to be insufficient, irrespective of the criterion applied for NRr. All these insufficient plots were found between dose 1 and dose 2. The remaining plots (dose 2 to dose 3, and dose 3 to dose 4) were classified as nutritionally balanced (Table 3).

For P, when assessing the true nutritional status and considering responsive plots with a YR above 5%, 47% of the plots were identified as responsive to P. The highest proportion of positive response to P addition took place from dose 2 to dose 3, while the lowest proportions were observed from dose 1 to dose 2 and dose 3 to dose 4. When the criterion for determining the true nutritional status involved a YR of 10%, 40% of the plots were considered responsive to P, with an equal proportion maintained for all comparisons between doses (Table 4).

Table 4.

Comparison of yield and nutritional status in test and reference plots: phosphorus fertilizer effects assessed using Compositional Nutrient Diagnosis—CND in cotton leaf samples from Chapadão do Sul, Brazil.

Blocks	Trial	TP	PR	YIc%	TNS		CND Diagnosis
					YR		NRr
					5%	10%	0.5	1.0
1	P2-P1	3061	5899	93%	TD	TD	INS	INS
	P3-P2	5899	3688	− 37%	TS	TS	INS	INS
	P4-P3	3688	3019	− 18%	TS	TS	INS	INS
2	P2-P1	4912	3803	− 23%	TS	TS	BAL	BAL
	P3-P2	3803	3090	− 19%	TS	TS	INS	BAL
	P4-P3	3090	4549	47%	TD	TD	INS	INS
3	P2-P1	4283,3	3296	− 23%	TS	TS	INS	INS
	P3-P2	3295,8	3601	9%	TD	TS	INS	INS
	P4-P3	3601,2	4636	29%	TD	TD	INS	INS
4	P2-P1	3051	4267	40%	TD	TD	INS	BAL
	P3-P2	4267	5044	18%	TD	TD	INS	BAL
	P4-P3	5044	4903	− 3%	TS	TS	BAL	BAL
5	P2-P1	2896	2882	0%	TS	TS	INS	INS
	P3-P4	2882	4048	40%	TD	TD	INS	INS
	P4-P3	4048	3164	− 22%	TS	TS	INS	INS
TD					47%	40%
INS							87%	67%

Doses for phosphorus tests: P1 = 0 kg ha⁻¹ P₂O₅, P2 = 30 kg ha⁻¹ P₂O₅, P3 = 57 kg ha⁻¹ P_2O₅ e P4 =  114 kg ha⁻¹ P₂O₅.

PT plot test, PR plot reference, YIc% yield increases change (in percentage)—denotes the change in yield associated with each trial conducted, TNS true nutritional status, YI yield increases, NRr nutrient responsiveness range, TD true deficiency, INS diagnosis of insufficient, BAL balanced, TS true sufficiency.

Concerning the diagnosis produced by the CND method and using a NRr of 0.5, 86% of the experimental plots were classified as insufficient. Only one plot within the range of dose 1 to dose 2, and another between dose 3 and dose 4, were diagnosed as nutritionally balanced. When adopting a NRr of 1.0, 67% of the experimental plots were identified as insufficient, with the largest proportion of insufficient diagnoses found between dose 3 and dose 4. In this case, employing a NRr of 1.0 led to a decrease in the number of insufficient diagnoses from dose 1 to dose 2 and from dose 2 to dose 3 (Table 4).

Considering responsive plots with a yield increase exceeding 5%, 47% of the plots were identified as responsive to B. The highest proportion of responsive plots took place between dose 1 and dose 2, and between dose 3 and dose 4. When using the 10% criterion to establish the extent of responsiveness, the proportion of responsive plots for B was 33%, with the largest proportions of responsive plots found between dose 1 and dose 2, and between dose 3 and dose 4 (Table 5).

Table 5.

Comparison of yield and nutritional status in test and reference plots: boron fertilizer effects assessed using Compositional Nutrient Diagnosis—CND in cotton leaf samples from Chapadão do Sul, Brazil.

Blocks	Trial	TP	PR	YIc%	TNS		CND diagnosis
					YR		NRr
					5%	10%	0.5	1.0
1	B2-B1	4998	4212	− 16%	TS	TS	INS	INS
	B3-B2	4212	3426	− 19%	TS	TS	INS	BAL
	B4-B3	3426	4472	31%	TD	TD	BAL	BAL
2	B2-B1	4736	4834	2%	TS	TS	INS	INS
	B3-B2	4834	4936	2%	TS	TS	BAL	BAL
	B4-B3	4936	3754	− 24%	TS	TS	BAL	BAL
3	B2-B1	4322	4770	10%	TD	TD	INS	INS
	B3-B2	4770	5519	16%	TD	TD	BAL	BAL
	B4-B3	5519	3853	− 30%	TS	TS	BAL	BAL
4	B2-B1	4905	8539	74%	TD	TD	INS	INS
	B3-B2	8539	5831	− 32%	TS	TS	BAL	BAL
	B4-B3	5831	6368	9%	TD	TS	BAL	BAL
5	B2-B1	4361	4699	8%	TD	TS	INS	BAL
	B3-B2	4699	4344	− 8%	TS	TS	BAL	BAL
	B4-B3	4344	10,459	141%	TD	TD	BAL	BAL
TD					47%	33%
INS							40%	27%

Doses for boron tests: B1 = 0 kg ha⁻¹, B2 = 1 kg ha⁻¹, B3 = 2 kg ha⁻¹ e B4 = 4 kg ha⁻¹.

PT plot test, PR plot reference, YIc% yield increases change (in percentage)—denotes the change in yield associated with each trial conducted, TNS true nutritional status, YI yield increases, NRr nutrient responsiveness range, TD true deficiency, INS diagnosis of insufficient, BAL balanced, TS true sufficiency.

For B diagnoses with the CND method, when applying a NRr of 0.5, 40% of the experimental plots were identified as insufficient for B. In this scenario, 100% of the plots from dose 1 to dose 2 were B-insufficient, while 100% of the plots from dose 3 to dose 4 were considered balanced. Upon altering the NRr to 1.0, 26% of the plots were considered insufficient in B. For these plots, 100% of those from dose 2 to dose 3 and from dose 3 to dose 4 were categorized as balanced (Table 5).

The accuracy of net yield productivity (NYP), in terms of cottonseed yield, varied from 21 kg ha⁻¹ for B accuracy to 304 kg ha⁻¹ for P accuracy. When considering the diagnosis of all three nutrients together, the minimum accuracy amounted to 341 kg ha⁻¹ (sum of 227 + 93 + 21 kg ha⁻¹), while the maximum was 772 kg ha⁻¹ (sum of 270 + 304 + 197 kg ha⁻¹). The smallest range of NYP was observed for N accuracy, corresponding to 43 kg ha⁻¹, followed by B accuracy (176 kg ha⁻¹) and P accuracy (211 kg ha⁻¹) (Table 6).

Table 6.

Percentage of cases identified for diagnosis true insufficiency (Tins) or balanced (Tbal) and diagnosis false insufficiency (F Ins) or balanced (Fbal); overall accuracy (Ac); accurate insufficiency (Ac Ins); e accurate balanced (AcBal) e general average (A-g) of the net yield productivity (NYP), using the compositional nutrient diagnosis (CND) method to the diagnosis of the N, P e B in cotton leaf samples, in Chapadão do Sul, Brazil.

CND	YI	NRr	NYP1						Case count (%)				Accuracy measurements (%)
			T Ins	F Ins	T Bal	F Bal	Final	A-g	T Ins	F Ins	T Bal	F Bal	Ac	Ac Ins	Ac Bal
N	5	0.5	3342	− 2571	7627	− 4990	3408	227	13%	7%	53%	27%	67%	33%	89%
		1.0	3342	− 2571	7627	− 4990	3408	227	13%	7%	53%	27%	67%	33%	89%
	10	0.5	3342	− 2571	7950	− 4667	4055	270	13%	7%	60%	20%	73%	40%	90%
		1.0	3342	− 2571	7950	− 4667	4055	270	13%	7%	60%	20%	73%	40%	90%
P	5	0.5	8796	− 5479	1249	0	4567	304	47%	40%	13%	0%	60%	100%	25%
		1.0	6803	− 4766	1963	− 1993	2007	134	33%	33%	20%	13%	53%	71%	38%
	10	0.5	8491	− 5784	1249	0	3956	264	33%	47%	13%	7%	47%	83%	22%
		1.0	6498	− 5071	1963	− 1993	1398	93	20%	40%	20%	20%	40%	50%	33%
B	5	0.5	4419	− 1669	6014	− 8447	317	21	20%	20%	33%	27%	53%	43%	63%
		1.0	4082	− 884	6800	− 8785	1213	81	13%	13%	40%	33%	53%	29%	75%
	10	0.5	4082	− 2007	6551	− 7910	715	48	13%	27%	40%	20%	53%	40%	60%
		1.0	4082	− 884	7675	− 7910	2963	197	13%	13%	53%	20%	67%	40%	80%

¹NYP values expressed in kg ha⁻¹, value obtained for the fifteen confrontations between control plot and response plot.

N nitrogen, P phosphorus, B boron, YI yield increases, NRr nutrient responsiveness range.

In the evaluation of the N status of cotton, a balanced accuracy of 89–90% was achieved, considering that 53% to 60% of the plants were classified as sufficient for nitrogen nutrition. Conversely, the accuracy for insufficiency was lower, ranging from 33 to 40% of cases, due to the small number of identified cotton plants deficient in N, which accounted for only 13% (Table 6).

In assessing the nutritional status of cotton for B, both the insufficiency and balanced accuracies consistently remained below 75%. This resulted from the low number of true deficiency and sufficiency cases, representing less than 20% and 53% respectively, based on the applied criteria (Table 6).

When evaluating the quality of P nutritional diagnoses, the accuracy for insufficiency ranged from 50 to 100%, being higher when a greater frequency of true deficiency cases was present. The balanced accuracy for P was the lowest among the three evaluated nutrients across all comparisons conducted (Table 6).

The adoption of the YR criterion of 10% raised the frequency of true sufficiency cases for N, P, and B by 7%, 7%, and 14% respectively. Similarly, adjusting the NRr from 1.0 to 0.5 increased the frequency of nutritional deficiency diagnoses for P and B by 20% and 13% respectively. However, the NRr adjustment had no effect on N (Tables 3, 4, and 5).

The conditions and criteria for achieving higher diagnostic accuracy varied among the nutrients. For N, the highest deficiency accuracy depended solely on the criterion adopted for the NRr factor and was attained by improving the identification of true sufficiency cases with an NRr of 1.0. For P, the best accuracy was associated with identifying true deficiency cases, using an NRr of 0.5 and a YR of 5%. For B, the optimal performance was related to improving the identification of true sufficiency cases, using an NRr of 1.0 and a YR of 10% (Table 6). This combination resulted in better Net Yield Productivity to the cotton (Fig. 2).

Figure 2.

General average of the Net yield productivity (NYP) attributed to the diagnoses of N, P and B, using NRr of 0.5 and 1.0 and YI of 5% and 10%, given by the CND method, in a cotton, in Chapadão do Sul, Brazil.

Discussion

The monitored commercial cotton crops exhibited an average yield higher than the national estimate (4267 kg ha⁻¹) and same the estimate for the state of Mato Grosso do Sul (4500 kg ha⁻¹) for cottonseed, referring to the 2017/2018 harvest²³. This is due to the prevalence of optimal edaphoclimatic conditions and a high level of technology in the study region.

The higher foliar nutrient concentrations observed in low-yield crops can be explained by the concentration effect, which occurs when the relative growth rate of dry matter is lower than the relative nutrient absorption rate²⁴.

Observing higher standard deviations among foliar nutrient levels in the low-yield population is consistent with the literature, which suggests increased variability among foliar levels in low-yield populations²¹. Consequently, there is a greater likelihood that nutrients such as Ca, S, B, Fe, and Zn are experiencing nutritional deficiency due to their increased variability.

Nonetheless, none of these nutrients fell below the recommended SR for the crop¹², indicating no nutritional deficiencies. This suggests a need to update reference values for foliar concentrations, as the current values in the literature are based on samples collected from 1998 to 2005, and cultivation conditions have likely changed since then¹². When comparing nutrient levels of N and P, the calibrated SR, obtained in the present study, showed a wider range than the literature SR¹², revealing distinct patterns. This indicates that the existing reference values no longer accurately represent current cultivation conditions.

In contrast, for B, the nutrient level exceeded its corresponding value cited in the literature, suggesting an over-utilization of this nutrient by the plant. However, contradictorily, the data proposes that the plant remains in a state of nutritional deficiency. This is evident, as even with the application of the maximum boron dose to the soil, there was a consistent, linear increase in cotton productivity, with no signs of growth saturation.

Various elements can dictate the absorption and utilization of nutrients in plants, with the choice of genetic materials used for cultivation being a major influence²⁵. This study observed a boron content range of 36–43 mg kg⁻¹ across the five different genetic materials evaluated. This discrepancy could be linked to the quantum of nutrient absorption by the plant, either due to a comprehensive root system or a higher absorption rate per unit root length²⁶.

Moreover, the application of hormonal growth regulators, a widespread practice in commercial cotton cultivation, is another contributing factor. Research finding suggest that these regulators can lead to changes in the nutrient composition of plant leaves and induce a nutrient concentration effect, as they limit the vegetative growth of the cotton plants²⁷.

In the calibration trials conducted for N and P, we were able to identify the fertilizer dosage that led to the highest cotton crop yield in the Brazilian Cerrado, aligning with the recommendation²⁸. However, in the case of B, we did not find a quadratic relationship indicating the point at which crop productivity decreases due to nutrient toxicity.

The findings concerning N showcased the crop’s sensitivity to this nutrient, which resulted in a boost in productivity. Nevertheless, the continued administration of N, followed by its overabsorption, may have caused toxicity in the plant, triggering a reduction in productivity. This downturn doesn’t stem directly from N toxicity. Instead, it is primarily due to the plant's excessive energy consumption for vegetative growth, which leads to neglecting reproductive organs²⁹.

The precise calibration of N allocation is vital in striking the perfect equilibrium between vegetative and reproductive development in cotton cultivation. Previous studies (3, 30–32) have indicated a range of optimal dosage quantities. This range underscores the profound impact of edaphoclimatic factors across varying research locales.

Research from the Brazilian Central-West region indicated that an optimal N dosage of 47 kg ha⁻¹ would yield 3396 kg ha⁻¹ of seed cotton³¹. A larger dose of 200 kg ha⁻¹ of N resulted in a yield of 2953 kg ha⁻¹ of seed cotton³⁰. In the Amazon region, a dose of 68 kg ha⁻¹ of N resulted in a productive yield of 3393 kg ha⁻¹of seed cotton³. However, in the Brazilian semi-arid region, a substantially higher optimal N dose of 210 kg ha⁻¹ was observed, leading to a peak yield of 5707 kg ha⁻¹ of seed cotton³².

As for P, its significant contribution to the yield of cotton is unmistakable, signifying the crop’s response to this essential macronutrient. This denotes the feasibility of a widespread P application, especially on soil well-suited for stable no-till farming with strong fertility.

Several investigations have identified a rise in productivity in relation to P application. In a study conducted in Chin³³, a considerable yield of 5216 kg ha⁻¹ of seed cotton was obtained with the application of 65 kg ha⁻¹ of P, on soil with a P content of 11.2 mg dm⁻³.

Meanwhile in Brazil, was observed a 27% boost in productivity, achieving 4894 kg ha⁻¹ of seed cotton, by applying 52 kg ha⁻¹ of P on soil containing 22.0 mg kg⁻¹ of P (Mehlich) in Januária, MG³⁴. In Chapadão do Sul—MS, the implementation of 52 kg ha⁻¹ of P yielded 2507 kg ha⁻¹ of seed cotton, marking a 20% increment when compared to non-P applications, in soil with a P content of 55 mg dm^-3 (Resin) ³¹. In the Cerrado territories of Bahia, where the soil demonstrated an average P content of 18 mg dm⁻³ (Mehlich), the use of 104 kg ha⁻¹ of P led to a remarkable yield of 5820 kg ha⁻¹ of seed cotton³⁵.

It is remarkable to note in this study³⁵ observed a trivial 2% dip in productivity in the absence of P application. Furthermore, the yield procured without nutrient application outperformed that of our current investigation that included nutrient application. According to the author³⁵, this can be attributed to the rigorous fertilization practices customary in commercial cotton cultivation, where deploying 43–52 kg ha⁻¹ of P is typical. Excessive P application in cotton farming has been reported in other studies, reflecting yield limitations due to P nutritional excess¹¹. The authors emphasize this insight as essential for optimizing nutrient application and averting unnecessary deployment when the plant does not require it.

The proposal to apply augmented doses of B is supported by comparable observations in study on Turkey³⁶. Their study involved the application of 3 kg ha⁻¹ of B to soils with an average B content of 0.18 mg kg⁻¹. Remarkably, they reported a 30% enhancement in cotton yield, realizing an average yield of 4320 kg ha⁻¹ of seed cotton.

These findings indicate that cotton plants, unlike those of other nutrients, may endure heightened boron concentrations without impacting their productivity. However, it’s paramount to underscore that different regions and cotton varieties may exhibit unique characteristics pertaining to the absorption and utilization of boron.

These results, which reveal a boost in crop yields owing to the targeted fertilization of N, P, and B, underline the rate-limiting role these nutrients play in cotton cultivation within tropical ecosystems. This observation aligns well with the plant's extant nutritional status, as evidenced by more than one-third of the experimental plots showing NYP upon the application of N, P, or B nutrients.

The findings suggest that a higher YR value is requisite for accurately identifying plants that are truly deficient in specific nutrients. This trend has been corroborated by previous studies on sugarcane^20,21 and soybean¹⁶, employing the diagnosis and recommendation integrated system (DRIS) and CND methodologies, respectively. For instance ^20,21, investigated a YR range spanning from 5 to 40%, while¹⁶ scrutinized YR values within a 1–10% interval. Both studies concurred that optimizing YR values facilitated enhanced methodological accuracy in their respective diagnostic approaches.

Contrary to the findings of our current study¹⁶ demonstrated that NRr values below 0.5 yielded greater diagnostic accuracy for plants that were either nutritionally deficient or balanced. Conversely²¹, observed that elevated NRr values enhanced the precision of phosphorus (P) nutritional assessments, thereby corroborating our own conclusions.

In evaluating the efficacy of different diagnostic methods for assessing the nutritional levels of N, P, and B in cotton crops, the study found a marked divergence in results. This discrepancy was magnified by the variability in the applied YR and NRr parameters.

Previous research has also pointed to variations in accuracy levels across different nutrients while assessing the nutritional status of banana plants using the DRIS^17,18. Was reported an overall accuracy of 63% for K and 69% for N¹⁷ and of the 51% (N) e 33% (K)¹⁸. The method displayed exceptional competence in the diagnosis of N. This led to a significant enhancement in crop yield, peaking at 2.14 metric tons per hectare¹⁷.

In a distinct study, a diagnostic accuracy of 80% was attained for B nutrient status in soybean cultivation using the CND approach¹⁵. This high level of accuracy was attributed to a perfect accuracy for balance score of 100%, which resulted in a negative NYP of − 46 kg ha⁻¹ for soybean grains¹⁵. Subsequent, research was confirmed that fine-tuning the NRr values led to substantial improvements in the methodological performance (− 167 to 36 kg ha⁻¹ for soybean grains)¹⁶.

Utilizing the CND methodology for nutritional assessment in sugarcane²¹ was recorded diagnostic accuracies between 40 and 71%. The true-to-false ratios were observed to span from 0.41 to 3.40. In alignment with this, a preponderance of NYP values manifested as negative, ranging from − 15 to 5 metric tons ha⁻¹ for nearly all levels of YR and NRr adopted in the study.

Under the specialized conditions governing this research, optimal combinations of NRr and YR were delineated, yielding the most propitious results. Such configurations underscored that accurate nutritional profiling of the plant can result in marked elevations in crop productivity. Concretely, the cotton yield saw a baseline increase of 16% with an accurate N diagnosis (YR = 10%, regardless of NRr), 13% in the case of P (NRr = 0.5 and YR = 5%), and a substantial 59% surge for B (NRr = 1.0 and YR = 10%).

The findings emphasize the necessity of continuously reviewing and adapting diagnostic methodologies. Maintaining flexibility in using criteria (NRr and YR) is crucial for ensuring accurate diagnoses and optimized results. However, it's important to note that the research on NRr and YR remains limited, underscoring the need for further exploration and studies in this field.

Methods

Nutritional monitoring and experiments were conducted on cotton crops in Chapadão do Sul, Mato Grosso do Sul, Brazil, situated in the Cerrado biome. The region has an Aw climate (Köppen system), with a 2 month dry season and 1550 mm average annual precipitation³⁷. Both the experimental site and the municipality have Rhodic Ferrosols soils³⁸. The distribution of precipitation and temperature distribution throughout the calibration experiment period (2017–2018 harvest) is illustrated (Fig. 3).

Figure 3.

Precipitation and temperature distribution throughout the calibration experiment period (2017–2018 harvest) in Chapadão do Sul, Brazil.

In the experimental area at Chapadão do Sul Foundation, three calibration experiments assessed N, P, and B nutrient fertilization. Treatments included varying doses of N (0, 63, 125 and 250 kg ha⁻¹), P₂O₅ (0, 30, 57 and 114 kg ha⁻¹), and B (0, 1, 2 and 4 kg ha⁻¹). These treatments were randomly arranged in blocks with five replications. Doses represented 0%, 50%, 100%, and 200% of recommended levels for cotton cultivation in the Brazilian Cerrado region²⁸. Experimental plots comprised five rows, each 5.5 m long and 4.5 m wide, with 0.9 m between rows (resulting in a density of seven plants per meter).

In the experimental area, soil acidity correction was performed before sowing. This involved applying 2 tons ha⁻¹ of lime to the entire area in August 2017, followed by 1 ton ha⁻¹ of agricultural gypsum 30 days later. Cotton, FM983 GLT cultivar, was sown on February 7, 2018. Soil test results showed high P (60.0 mg dm⁻³) and low B (0.08 mg dm⁻³) levels, interpreted following²⁸. In the Table 7 was described the chemical analysis of the experimental cotton plantation area.

Table 7.

Chemical analysis of the experimental cotton plantation area, Chapadão do Sul—MS, Brazil.

pH		P-Melich	P-Resina	S	MO	K	Ca	Mg
CaCl2	H2O	mg dm−3			g dm−3	cmol dm−3
4.88	5.86	60.02	55.80	13.1	30.98	0.2	3.1	0.6
Al	H	Al + H	Fe	Micronutrients (mg dm−3)
cmol dm−3				Mn	Zn	Cu	B
0.08	4.95	5.65	142.67	13.4	4.1	0.5	0.08

The N doses, applied as urea, were split into two broadcast applications. The first application consisted of 25% of the total dose, administered 14 days after emergence, while the remaining 75% was applied at 54 days after emergence. B doses, in the form of water-soluble boric acid, were applied between the planting rows at 27 days after emergence, ensuring no contact with the leaves. The application was carried out using a CO₂-pressurized sprayer with a spray volume of 150 L per hectare. P doses, in the form of single superphosphate granules, were broadcasted at 31 days after emergence. Additionally, 57 kg ha⁻¹ of K were applied as a broadcast of potassium chloride at 47 days after crop emergence. Calcium sulfate was employed to balance the sulfur content across all treatment conditions.

To control weed growth, desiccation was performed using glyphosate with a concentration of 792.5 g kg⁻¹ (2.5 kg ha⁻¹), along with the pre-emergence application of S-metolachlor at a concentration of 960 g L⁻¹ (1.2 L ha⁻¹), followed by post-emergence application of ammonium–glufosinate salt at a concentration of 200 g L⁻¹ (1.2 L ha⁻¹).

For the management of cotton pests, various combinations of insecticides were utilized, including: carbosulfan (700 g L⁻¹, 1 L ha⁻¹), flubendiamide (480 g L⁻¹, 150 mL ha⁻¹), etiprole (200 g L⁻¹, 1 L ha⁻¹), thiamethoxam (141 g L⁻¹) combined with lambda-cyhalothrin (106 g L⁻¹, 300 mL ha⁻¹), fipronil (800 g kg⁻¹, 100 mL ha⁻¹), thiodicarb (800 g kg⁻¹, 500 g ha⁻¹), malathion (100 g L⁻¹, 1.8 L ha⁻¹), chlorantraniliprole (100 g L⁻¹) combined with lambda-cyhalothrin (50 g L⁻¹, 400 mL ha⁻¹), indoxacarb (150 g L⁻¹, 600 mL ha⁻¹), methoxyfenozide (240 g L⁻¹) combined with spinetoram (120 g L⁻¹, 300 mL ha⁻¹), tebufenozide (150 g L⁻¹, 200 mL ha⁻¹), spiromesifen (240 g L⁻¹, 500 mL ha⁻¹), chlorfenapyr (240 g L⁻¹, 1.5 L ha⁻¹), diafenthiuron (500 g L⁻¹, 800 mL ha⁻¹), and novaluron (100 g L⁻¹, 400 mL ha⁻¹).

For cotton disease management, preventive treatments with various fungicides were used: a combination of fluxapyroxad (167 g L⁻¹) and pyraclostrobin (333 g L⁻¹L, 350 mL ha⁻¹), fenhexamid (400 g L⁻¹, 500 mL ha⁻¹), chlorothalonil (720 g L⁻¹, 1.5 L ha⁻¹), and difenoconazole (250 g L⁻¹, 500 mL ha⁻¹).

To sustain a daily growth rate of around 1.1 cm, growth regulators were administered, commencing at 24 days after emergence. Weekly applications of mepiquat chloride at a concentration of 250 g L⁻¹ (200 mL ha⁻¹) were employed. The treatment regimen concluded at 65 days after emergence with the application of chloromequat chloride at a concentration of 100 g L⁻¹ (3 L ha⁻¹).

For nutritional monitoring, a one-hectare area was designated in 144 commercial cotton crops around the municipality. In these areas, distinct fertilization methods, cultural practices, and phytosanitary management approaches were implemented in the various areas, tailored to the specific requirements of each planting location. The cultivated cotton varieties encompassed FM975, TMG81, FM 983 and FM983 GLT. The cultivars are recorded and can consulted in Brazilian National Registry of Cultivars (RNC).

Standardized foliar sampling occurred during the full bloom stage for both commercial crops and experimental plots. The fifth leaf from the top of the plant was collected³⁹. In commercial fields, 30 leaves were gathered per hectare, while experimental plots yielded 10 leaves.

Following the sampling process, the leaves were subjected to decontamination in deionized water, then washed in a detergent solution (0.1%) combined with hydrochloric acid (0.3%), and rinsed again in deionized water, following the methodology⁴⁰. Subsequently, the leaves were dried in a forced air oven at a temperature range of 60–70 ºC until a constant mass was achieved. Once dried, they were ground in a mill and analyzed for their N, P, K, Ca, Mg, S, B, Zn, Fe, Mn, and Cu content, employing the procedures⁴¹.

Cotton was harvested in July–August 2018, coinciding with boll maturation. In each commercial field, one square meter was marked per plot, 10 plants were randomly selected, and all bolls were collected. In calibration experiments, the two central rows of each treatment plot were harvested. Cotton yield, measured in kg ha⁻¹, was recorded as seed cotton.

To determine the calibrated SR for N and P levels, as well as the calibrated CL for B, leaf content values were considered, specifically those that corresponded to 90% of the maximum yield.

Using productivity data, the combined group was split into low and high yield subpopulations. The high yield subpopulation set the CND norms. The inclusion of the population in the high yield group required productivity exceeding the productive threshold, which is the average plus 0.5 times the standard deviation of the dataset's productivity⁴².

To determine the CND standard, was used the cottonseed yield and nutrient content values (N, P, and B) from the dataset, which included both experimental plots and commercial farms. The CND method was calculated as described below⁶.

First, calculate the R component (Eq. 1), which represents unquantified dry matter (in grams). For this, it is necessary that all the nutrients are in the same unit of measure (g kg⁻¹).

$$\text{R}=1000-\left(\text{N + P + K + Ca + Mg + S + B + Cu + Fe + Mn + Zn}\right)$$ 1

The geometric mean (mGeo) is the nth root of the product of nutrient contents and R, where “n” is the number of components, including R (Eq. 2):

$$\text{mGeo =}{\left(\text{N}\times \text{P}\times \text{K}\times \text{Ca}\times \text{Mg}\times \text{S}\times \text{B}\times \text{Cu}\times \text{Fe}\times \text{Mn}\times \text{Zn}\times \text{R}\right)}^{\text{(1/n)}}$$ 2

The multivariate relationship between nutrients is obtained by taking the natural logarithm (nl) of the ratio of each nutrient's value (vNutr) to the mGeo of the nutritional composition in the leaf samples (Eq. 3).

$$\text{zNutrj = nl}\left(\text{vNutr/mGeo}\right)$$ 3

Reference values (CND norms) consist of the mean and standard deviation of these relationships in reference crops (high-yielding). This was used to obtain the CND index for each nutrient (I_Nutrj).

$$\text{I\_Nutrj =}\left(\text{zNutrj - mNutrj}\right)/\text{sNutri}$$ 4

where, zNutrj is the nutrient relationship, and mNutrj and sNutrj are the mean and standard deviation for the nutrient, respectively (Eq. 4).

To interpret the CND indices, the responsiveness to fertilization method was employed, utilizing the average nutrient balance index (aNBI) as the classification criterion²². The nutrient balance index (NBI) was calculated by summing the absolute values of the individual nutrient balance indices⁴³. The aNBI is derived by dividing the NBI by the number (n) of nutrients evaluated²² (Eq. 5).

$$\text{aNBI:}\left[\text{I\_N}\right]+\left[\text{I\_K}\right]+\left[\text{I\_P}\right]+\left[\text{I\_Ca}\right]+\left[\text{I\_Mg}\right]+\left[\text{I\_S}\right]+\left[\text{I\_B}\right]+\left[\text{I\_Fe}\right]+\left[\text{I\_Mn}\right]+\left[\text{I\_Zn}\right]+\left[\text{I\_Cu}\right]\text{/n}\left(2\right)$$ 5

Interpretation of the nutritional status was done by the responsiveness to fertilization method²², where each nutritional diagnosis uses index values with the aNBI value. The variable “NRr” is a factor introduced to adjust the scaling the aNBI value on the interpretation of nutritional status²¹ thereby guiding the diagnostic outcomes. Were employed “NRr” of the 0.50 and 1.00.

So, to interpretation of the nutritional status, if the index is negative and exceeds f × aNBI in absolute value, it's categorized as insufficient (Eq. 6). In all other cases, the Index was considered balanced.

$$\text{Insufficient:}\text{Index}<0\text{and}\left[\text{Index}\right]>\text{NRr}\times \text{aNBI}$$ 6

The nutritional statuses classified as insufficient or balanced for N, P or B (by CND method) were evaluated in comparison with the TNS of the cotton plants¹³. The calibration test was used to define the TNS. In this context, cotton was deemed to be in a true deficient (TD) state if the application of the nutrient (N, P, or B) resulted in a significant yield increase (responsive) compared to the experimental plot without the nutrient or with a lower dosage than that applied in the test plot. Conversely, cotton was in a true sufficient (TS) state if the application of the nutrient (N, P or B) did not lead to a yield increase (nonresponsive) (Table 8). The YR were regarded as responsive when the yield increase of the test plot was at least 5% or 10% higher than that of the reference plot¹⁶.

Table 8.

Diagnosis of the nutritional status obtained by the interpretation method and the physiological true nutritional status (TNS) defined by the response to fertilization on the crop cotton.

Nutritional status interpretation	True nutritional status (TNS)
	Responsive|true deficiency (TD)	Nonresponsive|true sufficient (TS)
Insufficient	T Ins	F Ins
Balanced	F Bal	T Bal
Subtotais	∑Def	∑Suf

T Ins true insufficiency, T Bal balanced, F Ins false insufficiency, F Bal false balanced, ∑Def sum of deficiency, ∑Suf sum of sufficiency.

The diagnoses were regarded as true balanced (T Bal) ou true insufficient (T Ins) when cotton was categorized as insufficient (Ins) a true deficiency (TD) state or as balanced (Bal) in a true sufficiency (TS) state, respectively (Table 8). Conversely, the diagnoses were deemed false insufficiency (F Ins) or false balanced (F Bal) when cotton was classified as insufficient despite being in a true sufficiency (TS) state, or when classified as balanced while being true deficient (TD), respectively^13,15.

The accuracy of the nutritional diagnoses for N, P and B in cotton was evaluated using several measures: overall accuracy (Ac), net yield productivity (NYP) ¹³, accuracy for insufficiency (AcIns), and accuracy for balance (AcBal)^13,14.

The Acc is given by the percentage sum the of true diagnosis cases and obtained by the “Eq. (7)”, where n is the total number of comparisons performed.

$$\text{Acc}=100\ast \left(\text{T Ins}/\text{n}+\text{T Bal}/\text{n}\right)$$ 7

AccDef and AccSuf correspond to the percentage of correct answers in relation to the total of deficiency or sufficiency diagnoses, respectively, and were obtained by the “Eq. (8) and (9)”, where VDef and ∑Suf are the sum of cases of deficiency and sufficiency, respectively.

$$\text{AccDef}=100\times \text{T}\text{Ins}/\sum \text{Def}$$ 8

$$\text{AccSuf}=100\times \text{T}\text{Bal}/\sum \text{Suf}$$ 9

We calculated the Average Net Yield Productivity (A-g NYP) using the “Eq. (10)”, with adaptation of the original author’s formula^13,44.

$$\text{A - g}\text{NYP}=\left[\sum \left(\text{P}\times \text{TIns}\right)+\sum \left(\text{P}\times \text{TBal}\right)-\sum \left(\text{P}\times \text{FIns}\right)-\sum \left(\text{P}\times \text{FBal}\right)\right]/\text{n}$$ 10

In this equation, the mathematical operation involves adding or subtracting the absolute productivity value (|P|) (which denotes the change in yield associated with each trial conducted). The value is added if the diagnosis is considered true (|P_TIns| or |P_TBal|) and subtracted if the diagnosis is considered false (|P_FIns| or |P_FBal|) for each state of nutritional deficiency/insufficiency or sufficiency/balanced. “n” represents the total number of plots evaluated.

We also computed the average net yield productivity for nutrients (A-n NYP) for each nutrient (N, P, or B) by averaging the NYP values obtained between the tested factors (NRr and YR). Furthermore, we determined the range of net yield productivity (r-NYP) by subtracting the lowest NYP value from the highest NYP value obtained from the tested factors (NRr and YR) for each nutrient (N, P, or B).

The Shapiro–Wilk test was employed to assess the normality of the data distribution obtained from commercial crops. Was conducted an F-test at a significance level of 10% to examine the variation (standard deviation) in leaf nutrient contents between crops with low and high productivity. The one-sample t-test was performed at a significance level of 1% to assess whether the mean nutrient contents of all sampled crops were equivalent to literature reference values (upper limit of the SR). The two-sample t-test was carried out at a significance level of 5% to determine if the average nutrient contents of low yield crops were equal to those of high yield crops. Yield variability associated with the nutritional contents (N, P or B) underwent assessment via regression and correlation analyses. Within the experimental plots, the means of the nutritional components (N, P, and B) and yield across the experimental treatments, based on the applied nutrient doses (N, P or B), were scrutinized using Tukey’s test at a 5% probability level, conducted with the AgroEstat statistical software⁴⁴.

This research was not conducted with endangered species and all methods in this study were carried out in compliance/accordance with relevant institutional, national, and international guidelines and legislation.

Conclusions

All tested nutrients effectively increased cotton productivity, yet the CND method’s success in accurately diagnosing nutrient deficiencies varied depending on the nutrient and the criteria used to differentiate deficient plants from non-deficient ones.

Regarding nutritional diagnostics for cotton plants using the CND method, the most favorable outcomes for N and B were achieved with a nutrient responsiveness range of 1 and a yield response of 10%, with an average gain of 43 kg ha⁻¹ and 176 kg ha⁻¹, respectively, compared to the result obtained with the worst combination. For P, the optimal combination was a nutrient responsiveness range of 0.5 and a yield response of 5%, with an average gain of 211 kg ha⁻¹, respectively, compared to the result obtained with the worst combination.

Author contributions

E.I.F.T. contributed to data collection, data analysis, graph creation, and the writing and revision of the manuscript. P.G.S.W. assisted with data analysis and the writing and revision of the manuscript. R.M.P. also contributed to the writing and revision of the manuscript. D.F. conducted the experiments and collected the data. C.N.S.C. was responsible for conducting the experiments.

Data availability

The datasets generated and/or analyzed during this study are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.