Nitrogen nutrition estimation of cotton using greenness and ground cover
Conventionally, plant N status is obtained destructively using tissue leaf N or petiole sap nitrate readings. In recent years, spectral analysis methods, including the chlorophyll meter technique, have gained popularity as an alternative to the conventional technique for estimating plant chlorophyll or N content. In cotton-related research, many previous studies have adopted correlation analysis to determine the best wavelengths for the aforementioned analysis. However, the specific wavelengths that were selected through this approach often differed from one location to another. Likewise, these specific wavelengths were variable at different scales of measurement, times of data acquisition and parameters of interest (i.e., N or chlorophyll). In this study, I propose a new method to estimate chlorophyll or N content based on the physical characteristics of cotton plants as affected by N fertilization. This study is described in three different chapters. In the first chapter, the response of growth parameters (plant height, plant width and percent ground cover) in manifesting differences caused by N fertilizations were investigated. I also examined the relationship between the cotton growth parameters and indicators of N status (leaf tissue N, petiole sap nitrate and chlorophyll meter readings). The main goal is to identify a growth parameter that could be used to complement spectral measurement as a method for estimating plant N status. The responses of the growth parameters were subject to the presence of other growth limiting factors besides N fertilization. Under water stress, plants might fail to display the effects of N treatments. Percent ground cover was found to be the growth parameter best affected by N fertilization. In general, this parameter showed significant effects of N treatments more readily than plant height or plant width. In addition, percent ground cover was better correlated with plant N indicators than the other two growth parameters. These results suggested that percent ground cover could be used along with spectral information acquired from plants for the purpose of estimating different levels of chlorophyll or N content. The strong relationship between percent ground cover and leaf tissue N or chlorophyll meter readings also suggested that it could be used to estimate plant N status when other growth factors were not limited. Leaf tissue N and chlorophyll meter readings were strongly and significantly correlated with growth parameters, especially ground cover. However, none of the growth parameters was significantly correlated with petiole sap nitrate, suggesting its limitation for estimating them. In the second chapter, I examined and compared the reflectance of cotton plants measured at three different spatial scales as related to N treatment effects. The importance of this examination was to select the best spatial scale(s) for estimating chlorophyll or N content. The three spatial scales were the individual leaf, the canopy, and the scene. At the leaf scale, N treatment effects were most apparent at 550 nm and 700 nm. N treatments did not significantly affect the internal leaf structure, and hence the NIR reflectance. Wavelengths sensitive to the N fertilization shifted to 600 nm and 700 nm when the measurements were made at the canopy level. NIR reflectance began to increase with increased N fertilization, as N treatments promoted biomass production and, thus, multiple scattering by the leaves. High LAI had the potential to affect the N treatment signals although its effects were not observed to be great in the data. At the leaf and canopy scales, reflectance measured in the visible region was a function of chlorophyll concentration. On the other hand, measurements made at the canopy scale in the NIR showed the effects of increasing biomass production. At the scene level, the effects of N treatments were most sensitive at wavelengths from 685 nm to 690 nm. NIR reflectance also increased with the amount of N applied. Only measurements made at the scene scale showed a consistent relationship between the amount of N fertilization and reflectance in both the visible and NIR regions. Notably, the primary contribution to this sensitivity was differences in percent ground cover as a result of N fertilization, rather than chlorophyll concentration. Large differences in percent ground cover had a strong influence on the N signals, where it could completely confound the N treatment effects. In comparison, LAI had minimal effects on N signals. Since N fertilization primarily affects cotton chlorophyll content and percent ground cover, selecting only one scale may not be sufficient to show the effects of N treatments. Therefore, combining spatial scales that are most responsive to chlorophyll content (either the leaf or canopy level) with the scale that best explains the variation in percent ground cover (the scene level) might be a practical approach for estimating the chlorophyll or N content of crops such as cotton. Additionally, results derived from this study concerning the inconsistency of spectral reflectance as related to the amount of N fertilization might explain previously unsuccessful attempts to directly transform measurements made at one scale to another. This study also emphasized the significance of accounting for the field of view of a sensor, since spectral signature properties changed significantly with the changing spatial scales. The last chapter describes and tests a novel method of combining measurements made at different spatial scales as an approach to estimate in-season chlorophyll or leaf N content of field-grown cotton. In this study, leaf greenness estimated from spectral measurements made at the individual leaf, canopy and scene levels was combined with percent ground cover to produce three different indices, named TCCLeaf, TCCCanopy, and TCCScene. These indices worked best for estimating leaf tissue N, but did not work well for estimating chlorophyll content. Use of these indices was most effective at early flowering. Prior to this growth stage, the utilization of these indices was not recommended. Of the three indices, TCCLeaf showed the best ability to estimate leaf tissue N at early flowering (r2 = 0.89). These results suggest that the use of green and red-edge wavelengths derived at the leaf scale is best for estimating leaf greenness. TCCCanopy had a slightly lower r2 value than TCCLeaf (0.76), suggesting that the utilization of yellow and red-edge wavelengths obtained at the canopy level could be used as an alternative to estimate leaf tissue N in the absence of leaf spectral information. The relationship between TCCScene and leaf tissue N was the lowest (r2 = 0.50), indicating that the estimation of canopy greenness from scene measurements needs improvement. In general, TCCLeaf and TCCCanopy performed better than the indices developed through correlation analysis and NDVI in discriminating the effects of N rates. Results from this study confirmed the potential of these indices as efficient methods for estimating in-season leaf N status of cotton.