Estimating Forage Biomass and Nitrogen Concentration Using False Color Infrared Photography

remote sensing;close range photography

Morgan, Sarah Paige ; James T. Green, Jr., Committee Co-Chair,Ronnie W. Heiniger, Committee Member,Noah N. Ranells, Committee Member,Jeffrey G. White, Committee Co-Chair,Morgan, Sarah Paige ; James T. Green ; Jr. ; Committee Co-Chair ; Ronnie W. Heiniger ; Committee Member ; Noah N. Ranells ; Committee Member ; Jeffrey G. White ; Committee Co-Chair

University:North Carolina State University

关键词: remote sensing;    close range photography;   

Others  :  https://repository.lib.ncsu.edu/bitstream/handle/1840.16/906/etd.pdf?sequence=1&isAllowed=y

美国|英语

来源: null

【 摘 要 】

The objective of this research was to investigate the utility of using nonnormalized (raw) digital counts and vegetation indices (VIs) derived from false color infrared (FCIR) photography to estimate biomass (dry), nitrogen (N) concentration, and N uptake of several warm season forage canopies at several locations. In July 2000, FCIR aerial photography was obtained at an altitude of 854 meters from an experiment established in 1998 at the Caswell Farm in Kinston, NC to investigate realistic yield expectations (RYE) from warm season forages fertilized with swine (Sus scrofa domesticus) effluent and ammonium nitrate (NH₄NO₃). The experiment consisted of three forage canopies (bermudagrass [Cynodon dactylon L. 'Coastal'], crabgrass [Digitaria sanguinalis L. 'Red River'], and volunteer warm season [80% native crabgrass, 20% forbs]) fertilized at five N rates (0, 224, 449, and 674 kg ha⁻¹ yr⁻¹) with either effluent or NH₄NO₃ in a stripped split plot design. Biomass, N concentration, and N uptake were measured and regressed against green (G [490 to 550 nm]), red (R [550 to 700 nm]), and near infrared (NIR [700 to 900 nm]) digital counts and seven VIs (NDVI, Green NDVI [GNDVI], DVI, RVI, Normalized NIR [NormNIR], Normalized Green [NG], and Normalized Red [NR]). There was an N source x N rate interaction for N uptake in bermudagrass (BG) and crabgrass (CG) canopies and for biomass and N concentration in BG. Differences due to N source (N source x VI) affected the relationship between biomass and GNDVI in BG canopies and many of the relationships between crop response variables and VIs in VWS canopies. Biomass was best estimated by NIR digital counts in BG (R² = 0.82), NDVI in CG (R² = 0.54), and NormNIR in VWS (R2 = 0.86). Nitrogen concentration was best estimated by NDVI in BG (R² = 0.62), NIR digital counts in CG (R² = 0.56), and G digital counts in VWS (R² = 0.63). Green NDVI was a consistently strong estimator (R² > 0.76) of N uptake for all forage canopies and was unaffected by N source.In September, 2000, an experiment was established in Raleigh, NC to (i) test the utility of FCIR ground-based photography in estimating N concentration differences in bermudagrass ('Coastal') canopies grown at similar biomass, and (ii) to investigate how different soil moisture levels would affect image interpretation. The experiment consisted of three irrigation levels (0, 25 minutes, and 90 minutes) applied 24 hours before harvest split across three replications of five N rates (0, 11, 22, 45, and 90 kg ha⁻¹) applied eleven days before harvest on a well-established bermudagrass sod at early heading. False color infrared photographs were obtained from a height of 1.83 meters above the ground and represented a harvest area of 0.25 m². Biomass (dry), N concentration, N uptake, and soil moisture were measured at each harvest area and regressed against nonnormalized raw digital counts and VIs. Differences among N rates were found for N concentration and N uptake but not for biomass, indicating that biomass levels were similar across all treatments. Irrigation rates only affected soil moisture. Significant, but weak correlations (R² < 0.28) were found for the relationships among N concentration, N uptake, NIR digital counts, NormNIR, DVI, NG and GNDVI. When replications were analyzed as 'sites' and irrigation blocks as 'replications within sites', there was a site x VI interaction for N concentration, NG, and GNDVI, whereby two of the three 'sites' were more strongly correlated (R² = 0.38 to 0.57) with a VI than the combined relationship. Relationships were generally stronger within site 1 versus site 2 and site 3.In July and August of 2000 and 2001, ground-based FCIR photographs were acquired from six harvests of bermudagrass canopies ('Coastal') from four different locations throughout eastern North Carolina which were part of a larger experiment examining realistic yield expectations (RYE) on three soil types fertilized at five rates of nitrogen. Similar photographic methods were used, however, each harvest area consisted of an average of three photographs, each representing a ground area of 0.25 m². Relationships between Red, Green, and NIR digital counts and crop response variables for most of the sites were weak, however, normalization generally improved correlations. Moderate or strong correlations between spectral and crop response variables, such as between Green NDVI and N uptake (R² = 0.89), could be found among all sites and cuttings except one. There were cutting, year, and site interactions with VIs for all three comparisons among sites, cuttings, and years. Despite statistics indicating that harvests were best modeled individually, combined relationships usually resulted in higher coefficients of determination (R² = 0.66). Taking into account location, photography method, and environmental conditions, Green NDVI and DVI were best to estimate N concentration across four sites (R² = 0.40).