[Comparison of methods for estimating soybean leaf area index].

Area of Science:

• Agricultural Science
• Remote Sensing
• Ecology

Background:

• Leaf Area Index (LAI) is a crucial biophysical parameter for ecological and productivity models.
• Accurate LAI estimation is vital for understanding carbon circulation and plant physiology.
• Soybean LAI retrieval methods require robust evaluation for reliable data.

Purpose of the Study:

• To evaluate and compare the effectiveness of different methods for retrieving soybean Leaf Area Index (LAI).
• To assess the performance of vegetation indices (NDVI, RVI), Principle Component Analysis (PCA), and Neural Networks (NN) for LAI estimation.
• To identify the most precise and reliable method for soybean LAI retrieval.

Main Methods:

• Utilized field experiment data for soybean crops.
• Applied Normalized Difference Vegetation Index (NDVI) and Ratio Vegetation Index (RVI) methods.
• Employed Principle Component Analysis (PCA) and Neural Network (NN) approaches.
• Validated and compared the estimation accuracy using R-squared (R2) and Root Mean Square Error (RMSE).

Main Results:

• All evaluated methods showed ideal effects on LAI estimation.
• PCA (R2=0.883, RMSE=0.202) and NN (R2=0.899, RMSE=0.413) methods exhibited higher precision than vegetation indices (NDVI R2=0.753, RMSE=0.594; RVI R2=0.758, RMSE=0.616).
• PCA was identified as the best performing method overall, with NN showing great potential, especially in its regression slope nearest to 1 (R2=0.949).

Conclusions:

• Principle Component Analysis (PCA) and Neural Network (NN) methods are the preferred choices for soybean LAI estimation due to their higher accuracy.
• The superior performance of PCA and NN is attributed to the utilization of hyperspectral information from multiple bands.
• Vegetation indices performed well in reducing noise for small LAI values, but PCA and NN offer better overall applicability and precision.