Simulating spaceborne imaging to retrieve grassland nitrogen concentration

Abstract

Spaceborne optical imaging enables continuous monitoring of nitrogen concentration (N%) in grasslands. However, the differences in instrumental setup, image pre-processing, wavelength coverage, and sampling rate pose a challenge when attempting to leverage data from multiple spaceborne instruments. We develop a method that makes use of field spectroscopy and the Soil–Plant–Atmosphere Radiative Transfer (SPART) model to simulate any Top Of Atmosphere (TOA) spaceborne sensor. The purpose of this method is to allow better integration of field spectroscopy data with spaceborne optical imagery. We develop a hybrid model using simulated data and a Random Forest Regressor, which is then independently validated using real Sentinel-2 TOA reflectance collected during 2016 and 2020. Our model achieves an independent validation accuracy of 0.44 (RMSE), MPIW of 0.11 and R2 of 0.55, demonstrating the potential of this methodology for monitoring grassland N% using field spectroscopy.