研究目的
To predict pasture quality maps and associated uncertainties using hyperspectral and environmental data, specifically for copper concentration in pasture.
研究成果
QRF provides accurate prediction maps of copper concentration in pasture with associated uncertainties, enabling better farm management decisions. Further research is needed to optimize band selection for improved accuracy.
研究不足
The study did not select the most appropriate spectral bands sensitive to copper concentration, which could decrease uncertainty. The prediction accuracy may vary with factors like topography and pasture species composition.
1：Experimental Design and Method Selection:
The study used a Quantile Regression Forest (QRF) approach to handle non-linear and non-parametric data for predicting copper concentration from hyperspectral, environmental, and topographical data. This method estimates prediction intervals to account for uncertainty.
2：Sample Selection and Data Sources:
150 pasture samples were collected randomly from a hill country farm in New Zealand. Hyperspectral data was acquired using an airborne survey with the AisaFENIX sensor, covering 350 to 2500 nm. Environmental and topographical data were also integrated.
3：List of Experimental Equipment and Materials:
AisaFENIX hyperspectral imaging system (Specim Ltd), GPS and navigation system (RT Oxford Survey+ Ltd.), GNSS, IMU, and laboratory equipment for chemical analysis (Analytical Research Laboratories Ltd.).
4：Experimental Procedures and Operational Workflow:
Conducted an airborne survey at 660 m altitude for 1 m GSD. Processed raw hyperspectral images to radiance, performed geometric correction and atmospheric compensation using ATCOR4. Collected pasture samples, analyzed for copper concentration in the lab. Applied QRF algorithm to combine data and generate prediction maps with uncertainties.
5：Collected pasture samples, analyzed for copper concentration in the lab. Applied QRF algorithm to combine data and generate prediction maps with uncertainties.
Data Analysis Methods:
5. Data Analysis Methods: Used QRF for regression, optimizing the number of trees with out-of-bag predictions. Evaluated accuracy with R2 and RMSE metrics.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
