^{1}

^{2}

^{3}

^{4}

^{*}

With the prediction of climate change-induced increases in drought frequency and severity in the southeastern USA, it is important to better understand the risks that drought may pose to NO
_{3} accumulation in bermudagrass [
Cynodon dactylon (L.) Pers.] forage. This report offers observations of NO
_{3} concentration in Bermudagrass forage samples submitted to the University of Georgia’s Feed and Environmental Water Lab (FEWL) during the extreme to exceptional drought of 2007, the severe drought of 2008, and the four preceding seasons when drought stress was minimal or absent. The probability (
P) of a sample being at high risk for nitrate toxicosis was the greatest for the extreme to exceptional drought of 2007 (
P = 0.160), slightly lower in the severe drought year of 2008 (
P = 0.105), and the lowest for samples from the 2003-2006 growing seasons (
P = 0.082) when drought stress was minimal or absent.

Global climate change is expected to cause an increase in frequency and severity of drought in the southeastern United States during the next century [_{3}-N) concentrations in the forage crops common to this region [_{3} kg^{−1} are at risk of developing abnormally high concentrations of nitrite in their bloodstream, leading to the formation of methemoglobinemia. This, consequently, can result in condition commonly referred to as nitrate toxicosis. Nitrate levels higher than 4500 mg NO_{3} kg^{−1} pose a risk of chronic nitrate toxicosis and, in some situations, animal death [

Bermudagrass [Cynodon dactylon (L.) Pers.] is the most widely used warm season perennial grass for pasture and hay production in the southern USA, and it is noted for its drought tolerance and high water use efficiency [_{3} [_{3} levels in bermudagrass do frequently occur [

Recent weather conditions in Georgia (USA) have granted a unique opportunity to make observations about the propensity for NO_{3} accumulation in bermudagrass. The National Drought Mitigation Center’s U.S. Drought Monitor classified a large majority of the state of Georgia in their extreme (D3) to exceptional (D4) drought categories during the 2007 growing season and at least a severe (D2) drought during 2008 ( [

This study uses observations obtained from the analyses of bermudagrass forage samples submitted to the University of Georgia’s FEWL from 1 July 2003 through 31 March 2009. The data were first screened to include only those samples that had been analyzed for NO_{3} concentration, which is determined in the FEWL by a colorimetric analysis of the nitration of salicylic acid [

from the USDA―National Agricultural Statistics Service (NASS) indicated that drought conditions were not consistently severe in Georgia’s counties along the Atlantic Coast during 2007 and 2008. Therefore, all observations from the 17 counties in the NASS’s Georgia District-9 were screened from each year used in this analysis.

Next, the data were segregated into growing seasons based on sample submission date, using an assumption that bermudagrass from an individual growing season would be submitted between 1 June of that same year and 31 March of the following year. For example, bermudagrass samples submitted between 1 June 2007 and 31 March 2008 were categorized into the “2007” growing season category. All other observations falling outside of this time frame (i.e., with a sample submission date of 1 April through 31 May) were excluded from the analysis. Samples submitted from the 2003, 2004, 2005, and 2006 seasons were originally subjected to an analysis similar to that described below but were found to not differ (P > 0.10) from one another (data not shown). Because of their similarity and indications from U.S. Drought Monitor data that bermudagrass forage produced during this period was largely free from severe drought stress, the observations from the 2003, 2004, 2005, and 2006 growing seasons were combined into one large category (“2003-2006”).

Using the UNIVARIATE procedure in SAS [_{3} kg^{−1} detection limit. These observations of nitrate levels below the detection limit are known as left-censored data and were included in this analysis by employing a survival analysis technique. In this technique, the censored data are kept in the analysis but are treated differently from actual observed values. Specifically, it is assumed that there is a latent variable Y^{*} for the i-th observation which is the actual NO_{3} concentration in the natural log scale, while Y_{i} is the detected natural log transformed NO_{3} concentration. The model can be expressed using Equations (1) to (3), and used within the LIFEREG procedure in SAS [

where_{3} concentrations in the null, 2007, and 2008 growing season categories, respectively; I{year = 2007} is an indicator variable which is 1 for growing season 2007 and 0 for other growing seasons and I{year = 2008} is an indicator variable which is 1 for growing season 2008 and 0 for other growing seasons (i.e., for the null growing seasons, the two indicator variables are both zero).

In order to compare the frequency of having moderate or highly risky NO_{3} concentrations within the growing season categories, the probabilities of these occurrences during the individual growing seasons were calculated. This was done by first grouping the bermudagrass samples into three categories of toxicity risk based on their NO_{3} concentration: Low Risk (LR), where nitrate levels were less than 2500 mg NO_{3} kg^{−1}; Moderate Risk (MR), where nitrate levels were 2500 - 4500 mg NO_{3} kg^{−1}; and High Risk (HR), where nitrate levels greater than 4500 mg NO_{3} kg^{−1}. Because the three levels are ordinal, a cumulative logistic model [

Growing season | n | Range | Skewness | Kurtosis | W^{†} | D^{‡} |
---|---|---|---|---|---|---|

mg NO_{3} kg^{−1} | ||||||

2003-2006 | 2533 | ND^{§} - 18,647 | 1.969 | 5.385 | - | 0.258^{**} |

2007 | 1010 | ND - 17,804 | 2.503 | 8.507 | 0.808^{***} | 0.181^{**} |

2008 | 869 | ND - 17,012 | 2.981 | 12.623 | 0.716^{***} | 0.238^{**} |

^{**}, ^{***}Significant at the 0.05, 0.01, and 0.001 probability levels, respectively. ^{†}W = the Shapiro-Wilk test statistic, which is appropriate for characterizing datasets when n ≥ 2000. ^{‡}D = Kolmogorov-Smirnov statistic. ^{§}ND = not detectable.

2007, and 2008 growing seasons. Specifically, Equation (4) presents the cumulative logistic model for the log odds that was analyzed using the LOGISTIC procedure in SAS (SAS Institute, 2004).

where the cumulative logits are defined as Equations ((5) and (6)).

where intercepts α_{1} and α_{2} are associated with log odds_{1} and log odds_{2} (as defined above), respectively, and have the interpretation of log odds in the 2003-2006 growing seasons, τ_{1} is the increment to the log odds in the 2007 growing season, and τ_{2} is the increment to the log odds in the 2008 growing season; where I {year = 2007} is an indicator variable that equals 1 for the 2007 growing season and 0 for other growing seasons and I {year = 2008} is an indicator variable that equals 1 for growing season 2008 and 0 for other growing seasons (i.e., for the 2003-2006 growing seasons, the two indicator variables are both zero); and where it is given, as defined in Equation (7), that the probabilities for each risk category in a given year sum to equal 1.

For each growing season category (2003-2006, 2007, and 2008), the estimated parameters from Equation (4) were used in Equation (5) to calculate the probability of MR or greater [P(≥MR)] concentrations of NO_{3} in the submitted forage samples. Similarly, these model parameters were then used in Equation (6) to calculate the probability of HR [P(HR)] concentrations of NO_{3} in the sample submissions for each growing season category. The probability of MR [P(MR)] NO_{3} concentrations was derived by subtracting P(HR) from P(≥MR). Then the probability of LR [P(LR)] NO_{3} concentrations was derived from Equation (7) using the estimates of P(HR) and P(MR).

The results of the analysis of maximum likelihood estimates for β_{0}, β_{1}, and β_{2} indicated that all parameters were significantly (P < 0.01) greater than zero (_{0}, β_{0} + β_{1}, and β_{0} + β_{2} are estimates of the means of the true natural log transformed NO_{3} concentrations during the 2003-2006, 2007, and 2008 growing season categories, respectively, these estimates were calculated and transformed back to the original units (_{3} in samples submitted during the 2003-2006 growing seasons (512 mg NO_{3} kg^{−1}) was significantly lower (P < 0.05) than in samples submitted from the drought-stressed season of 2008 (626 mg NO_{3} kg^{−1}), and both were significantly lower than those from the extreme to exceptional drought-stressed season in 2007 (1170 mg NO_{3} kg^{−1}). Though these mean values are considerably lower than those expected to pose a moderate (2500 mg NO_{3} kg^{−1}) or high (4500 mg NO_{3} kg^{−1}) risk of nitrate toxicosis [_{3} concentrations that would be considered moderate or highly risky (

To compare the frequency of moderate or highly risky NO_{3} concentrations across growing season categories, the probabilities of these occurrences during the individual growing seasons were calculated. The results of the analysis of maximum likelihood estimates for parameters α_{1}, α_{2}, τ_{1}, and τ_{2} used in Equation (4) indicated that each of the parameters was significantly different (P < 0.01) from zero (_{3} in bermudagrass forage sample submissions during the 2003-2006, 2007, and 2008 growing seasons (

In each of the three growing seasons, there existed a high probability (P > 0.65) that a bermudagrass forage sample would be classified as LR. However, this probability differed between each growing season category,

Parameter | Estimate | 95% CI | SE | χ^{2} | P > χ^{2} |
---|---|---|---|---|---|

β_{0} | 6.237 | ±0.066 | 0.034 | 34,303.4 | <0.0001 |

β_{1} | 0.827 | ±0.122 | 0.062 | 177.59 | <0.0001 |

β_{2} | 0.196 | ±0.130 | 0.066 | 8.78 | 0.0030 |

Growing season | Mean | 95% CI | SE |
---|---|---|---|

mg NO_{3} kg^{−1} | |||

2003-2006 | 512^{a†} | ±32.6 | 16.79 |

2007 | 1170^{c} | ±200.1 | 102.09 |

2008 | 626^{b} | ±110.4 | 56.33 |

^{†}Same letters within column are not significantly different at P < 0.05.

Growing season | Low risk (<2500 mg NO_{3} kg^{−1}) | Moderate risk (2500 - 4500 mg NO_{3} kg^{−1}) | High risk (>4500 mg NO_{3} kg^{−1}) |
---|---|---|---|

2003-2006 | 80.3% | 11.2% | 8.5% |

(2034/2533) | (284/2533) | (215/2533) | |

2007 | 65.2% | 19.6% | 15.1% |

(659/1010) | (198/1010) | (153/1010) | |

2008 | 75.8% | 13.3% | 10.8% |

(659/869) | (116/869) | (94/869) |

Parameter | Estimate | 95% CI | SE | Wald χ^{2} | P > χ^{2} |
---|---|---|---|---|---|

α_{1} | −1.402 | ±0.097 | 0.0497 | 793.7724 | <0.0001 |

α_{2} | −2.411 | ±0.120 | 0.0614 | 1541.714 | <0.0001 |

τ_{1} | 0.756 | ±0.160 | 0.0817 | 85.5141 | <0.0001 |

τ_{2} | 0.263 | ±0.182 | 0.0928 | 8.0623 | 0.0045 |

Growing season | Low risk (<2500 mg NO_{3} kg^{−1}) | Moderate risk (2500 - 4500 mg NO_{3} kg^{−1}) | High risk (>4500 mg NO_{3} kg^{−1}) | |||
---|---|---|---|---|---|---|

2003-2006 | 0.802 | ±0.0159 | 0.115 | ±0.0063 | 0.082 | ±0.0086 |

2007 | 0.656 | ±0.0602 | 0.184 | ±0.0213 | 0.160 | ±0.0342 |

2008 | 0.757 | ±0.0549 | 0.138 | ±0.0231 | 0.105 | ±0.0251 |

with the highest probability occurring in the more “typical” growing seasons of 2003-2006 (0.802 ± 0.0159), followed by the severe drought season of 2008 (0.757 ± 0.0549), and with the lowest probability occurring in the extreme to exceptional drought of 2007 (0.656 ± 0.0602). The most concerning issue revealed in this comparison was that samples submitted from the 2007 growing season were 1.95 times (0.160/0.082) more likely to be in the high risk category than in the “normal” growing seasons of 2003-2006. Though this also tended (P < 0.10) to occur in 2008, it was not as likely as in 2007.

It is important to note that our results should not be interpreted as establishing the expected mean NO_{3} concentration of samples from similar seasons or number of observations in LR, MR, and HR categories during drought. Our analysis was performed on producer-submitted bermudagrass forage samples. It is likely that producers were more apt to identify and submit forage samples from bermudagrass lots that may have been at an increased risk of nitrate toxicity during the droughts of 2007 and 2008. This may have increased the observed mean and altered the distribution of NO_{3} concentrations in the samples from these seasons. Even so, this analysis has shown that high risk samples do frequently occur and that this may even occur in samples from growing seasons that were not severely drought stressed (

Moreover, these observations indicate that NO_{3} may accumulate in bermudagrass to a much higher level than previously thought. Of the published work that reports NO_{3} concentration in forage bermudagrass [_{3} kg^{−1} [_{3} concentrations greater than 8350 mg NO_{3} kg^{−1}. In the extreme to exceptional drought of 2007 and the severe drought season of 2008, we observed that 3.47% (35/1010) and 2.30% (20/869) of the bermudagrass samples submitted from those respective growing seasons exceeded this previously reported high.

The degree to which drought stress influences NO_{3} concentrations in forage bermudagrass will need to be assessed in well-designed field trials. For example, this additional research should include an assessment of the influence of N fertilizer application strategies (e.g., splitting the applied N equally across multiple applications, the use of “enhanced-efficiency” and slow/controlled-released fertilizers, etc.) so that an assessment can be made of the degree to which N management interacts with the growing season conditions to influence the development of toxic NO_{3} concentrations in bermudagrass.

The authors thank Dr. Jien Chen, Associate Director of the University of Georgia’s Statistical Consulting Center, for her assistance with the statistical analysis performed for this paper.

Dennis W. Hancock,Uttam K. Saha,Jennifer J. Tucker,R. Lawton Stewart Jr., (2016) Observations of High Nitrate Concentrations in Forage Bermudagrass during Severe to Exceptional Drought Years. American Journal of Plant Sciences,07,695-701. doi: 10.4236/ajps.2016.74062