34321

Hyperspectral and NDVI measurements of plants that are heated to simulate global warming

Favoritar este trabalho

Introduction
Predictions for a global temperature increase within the next 100 years range anywhere from 2 to 10°C. An experiment was set up in 2014 at the Bioforsk complex in the Pasvik valley (Norway) to assess the impact of climate change and pollution-induced stress on a meadow field. Plots of meadow are heated up by three degrees Celsius and monitored by a complex system of sensors and imaging devices in the field, remote satellite imagery from above and laboratory tests of soil samples to keep track of plant production, soil moisture, nutrients, microbial communities, heavy metal concentrations and greenhouse gas emissions in the soil and in the air above the heated sites. The present work aims at a monitoring of the Normalized Difference Vegetation Index (NDVI) of the plants during the whole growth season. This is carried out using a dedicated NDVI camera combined with full hyperspectral imaging.

Materials and methods
A total of 20 plots of 4m diameter are included in the experiment. Half of these are heated up by three degrees Celsius. Moreover, every plot is divided in two, and biochar has been added to the soil on one of the halves. All the plots are being monitored with a NDVI camera (Canon EOS700D, Maxmax.com) with a frequency of two to three times per week during the growth season. Hyperspectral images in VIS-NIR (400 – 1000 nm, Specim) and NIR wavelengths (900 – 1700 nm, NEO) were obtained from the plots in two campaigns during the growth season. Images were obtained of whole plots at 10m above ground and of close-up regions approximately 1m above the ground. A white reference Spectralon plate was included in all the images for flat field correction. The images were analyzed using MATLAB (Mathworks) and PLS Toolbox (Eigenvector) as software platforms. Different NVDI equations for the NDVI camera were evaluated to data from the hyperspectral camera.

Results and discussion
The verification of the NDVI measurements was successfully done by bringing a full hyperspectral laboratory setup out to the field. The effect on the NDVI values by compensating the RGB values for the overlapping spectras in the NDVI camera is shown.