Development of a spectral library (CROPSPECPK) for major crops through spectroscopy modeling and in-situ spectrometer data

Identification, mapping and differentiation of one type of crop from another with satellite remote sensing is challenging in areas with small to medium sized patterns of mixed-crop cultivation. Knowledge of the spectral behavior of different crops, during their different growth stages, soil conditions and other environmental settings can therefore play a crucial role in mixed-cultivation areas. This spectral information representative of different crop types during different growth stages and seasons can be recorded by field spectroscopy. This study has developed a comprehensive crop spectral library (CROPSPECPK) for the major crops and vegetables by addressing a critical gap in region-specific crop spectral references for Pakistan. The representative hyperspectral signatures (221 sampling locations), for different crops were collected using GPS and a handheld spectrometer with spectral range from 300 to 1100 nm, during the winter-spring (Rabi) and summer-autumn (Kharif) of 2022. The instrument was calibrated for crop spectral measurements, employing the modified B-spline smoothing and the K-means transformation approach to reduce noise in the spectral curves. Ultimately, each crop's behavior in the spectral library was established for defined atmospheric and seasonal conditions. Distinct reflectance features were found in the visible (450-523 nm, 530-590 nm and 650-680 nm), red edge (698-793 nm) and near-infrared (785-900 nm) regions, with significant variations between specific crop types and times of year. In-situ crop reflectance spectra were compared using eight vegetation indices across Landsat-8, PlanetScope and Sentinel-2 with overall RMSE values of 0.266, 0.296, and 0.272, respectively. This robust research uniquely concentrates on ground-based spectral measurements, offering a distinct advantage over traditional spectral libraries that depend on airborne or satellite imagery, often constrained by atmospheric interference and coarser spectral and spatial resolutions.