GAI and chlorophyll content monitoring over wheat fields from continuous sub-hourly spectroradiometer measurements
Abstract
In-situ daily monitoring of the crop status along the growth cycle is one of the major concerns for an efficient management of agricultural fields. Green area index (GAI) and leaf chlorophyll content (Cab) are among the most essential crop variables since they relate to canopy photosynthetic capacity. Currently, they are measured by different hand-held instruments, e.g., LAI-2200, ceptometer and Digital Hemispherical Photos for GAI, and SPAD leaf clip for Cab. These instruments are expensive and time consuming since they require human intervention.
In this study, we present an autonomous monitoring Field Sensor system that is specifically designed for agricultural applications. The system is composed of a RGB camera, a miniature multispectral spectrometer and a PAR sensor that are fixed on a pole in the field. While the PAR sensor points vertically upwards, the RGB camera and spectrometer point towards the canopy with a 45° zenithal angle in a direction perpendicular to the row. The spectrometer has a field of view of ± 20° to take into consideration the heterogeneity of the vegetation cover. The spectrometer is sensitive to visible, red edge and near infrared light. Canopy reflected radiation and downwelling PAR are measured every 15 minutes. The measurements are transferred automatically to server through mobile network. The estimation of daily GAI and Cab was composed of two steps. We firstly trained one neural network to estimate the PAR diffuse fraction based on measured total PAR, using more than 10,000 historic field-measured diffuse PAR and total PAR values. The PAR diffuse fraction was then exploited to estimate the diffuse fraction in each band of the spectrometer using simulations of the 6S atmospheric correction model. In the second step, we retrieved GAI and Cab by inverting the PROSAIL radiative transfer model using as inputs the reflectances measured by the spectrometer and normalized by the average reflectance values across the six bands, and the diffuse fraction. This allowed minimizing the uncertainties associated to the absolute calibration of the spectrometer measurements that was used to convert the measured signal into reflectance values. The algorithm was evaluated over 43 Field Sensor systems that were installed in wheat fields in France from March to May 2019. GAI was measured using a RGB camera and estimated by inversion of Possion gap fraction model, and Cab was measured with a SPAD chlorophyll-meter. Results obtained over 40 samples showed that: (1) the canopy reflected flux measurements present diurnal variation that depends on the PAR diffuse fraction ; (2) the estimated daily GAI showed a good consistency with ground truth, with RMSE=0.53; (3) Cab was also retrieved accurately with a RMSE=14.05 μg/cm2.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMB040...06L
- Keywords:
-
- 0410 Biodiversity;
- BIOGEOSCIENCES;
- 0439 Ecosystems;
- structure and dynamics;
- BIOGEOSCIENCES;
- 0466 Modeling;
- BIOGEOSCIENCES;
- 0480 Remote sensing;
- BIOGEOSCIENCES