Abstract :
[en] Biochar, the solid carbon-rich residue of biomass pyrolysis, has been widely promoted in the context of climate-smart agriculture to improve soil nutrient availability, soil acidity, and water holding capacity. The aforementioned biochar impacts on soil quality can, in turn, affect nutrient uptake and crop productivity. Although numerous studies have explored short-term biochar effects on soil-plant system, there exists a general lack of researches investigating the long-term effects. As such, this project (CHAR-ARC) aims to evaluate the effects of century-old biochar (aged in agricultural soils more than 150 years ago) on soil-plant interactions within three different work packages. The first and second work packages study century-old biochar effects on nutrient cycling and water dynamics, respectively. The third work package (this thesis) investigates century-old biochar effects on crop dynamics using remote sensing images allowing for a high spatio-temporal monitoring thorough the entire cropping seasons. The study site is an agricultural field near Isnes (NW corner: 50°31′N 4°44′E; SE corner: 50°31′N 4°45′E) in central Belgium. The field was predominantly covered with oak, hornbeam, beech, and hazel forests and had been turned into cropland since the eighteenth century resulting in eleven biochar patches within the site. The soil type is Luvisol with a silt loam texture in the topsoil. Soil organic carbon and total nitrogen concentrations were determined following a regular sampling scheme by dry combustion using the Leco TruMac element analyser. High-resolution remotely-sensed images captured by unmanned aerial vehicles (UAVs) were deployed (a total of 27 UAV flights) to monitor crop performance across the century-old biochar patches in comparison to their adjacent reference soil patches. Red-Green-Blue (RGB) dataset was acquired using a DJI Phantom 4 Pro. Multispectral and thermal datasets were collected RedEdge-M and FLIR Vue Pro Radiometric sensors respectively, on boar of a DJI Matrice 100.
In Chapter 2, high-resolution UAV images were deployed to monitor chicory crop growth across the century-old biochar patches in comparison to their adjacent reference soil patches over the 2018 cropping season. Chicory canopy cover was retrieved from the RGB sensor as the ratio of the pixels covered by vegetation to the total number of ground pixels over the 2018 cropping season. Statistical paired t-test indicated a significant positive impact of century-old biochar on chicory crop growth during the green-up phase. This was corroborated with greater chicory leaf lengths across the biochar patches measured in the field. In addition, chicory crop water stress was retrieved as the difference between the canopy temperature (derived from the thermal images) and the air temperature (recorded by the on-site weather station). Moreover, the topographical wetness index (TWI) was computed to mitigate the impact of topography on the calculated crop water stress. Interestingly, TWI exhibited a positive relationship with soil organic carbon. Moreover, our results indicated an increase in chicory crop water stress across century-old biochar patches. Lastly, no impact of biochar was observed on the harvested plant biophysical parameters such as chicory yield, root length, or perimeter. The latter underlined the computed chicory canopy cover curves in which no particular impact of biochar was observed at the end of the growing season.
In Chapter 3, winter wheat performance was evaluated using RGB and multispectral images throughout the 2019 cropping season. As such, monitoring winter wheat crop growth and health provided new insights into the alteration in winter wheat crop dynamics at the canopy level associated with century-old biochar presence. Similar to the results of chicory season, a significant positive impact of biochar on the evolution of winter wheat canopy cover was illustrated using the RGB images. Furthermore, winter wheat plant height was computed based on the RGB images by subtracting the digital terrain model from the digital surface model (following the necessary calibration chain). The remotely-sensed winter wheat plant height was slightly higher across the century-old biochar patches. In addition, a total of eight vegetation indices were computed based on the multispectral dataset. Though the contrast between biochar and reference plots was not thoroughly the same for all the vegetation indices, several indices such as optimized soil adjusted vegetation index (OSAVI) highlighted a significantly better winter wheat crop development at the beginning of the growing season. There was however no particular impact of biochar on crop spectral status towards the end of the season. Crop health maps were also computed based on the RGB and multispectral images using principal component analysis and k-means clustering. Both RGB and multispectral crop health maps (comprising a clustering agreement of 74.82%) indicated a better winter wheat crop health across the biochar plots. There was also a significant positive impact of biochar on crop growth stages measured in the field. Finally, the plausible relationship between the multispectral vegetation indices and the harvested crop yield was tested. Simplified canopy chlorophyll content index (s-CCI) and normalized difference red edge index (NDRE) were found to be good linear estimators of harvested crop yield. s-CCI was then used to predict a harvested crop yield map for the entire study field. The predicted crop yield showed no remarkable influence of biochar on winter wheat harvested crop yield.
In Chapter 4, the potential of high-resolution UAV images to monitor plant diseases, and the plausible link between century-old biochar and plant diseases (as raised in generated crop health maps in Chapter 3) was specifically investigated. For this, UAV-based RGB images were first acquired across four winter wheat fields in the Grand Duchy of Luxembourg. At the same time, the percentage of the diseased leaf areas in terms of wheat stripe rust (WSR) and wheat leaf rust (WLR) were visually observed in the field. Then, WSR and WLR severities were determined from the acquired RGB images through additive or subtractive models in visual cryptography. As such, digital numbers (DNs) of green and red spectral bands were used to determine WSR. In addition, the combination of DNs of green, red, and blue spectral bands was used to identify WLR. The results indicated a strong correlation between UAV imagery and in-field observations for the determination of WSR, WLR, and canopy cover for winter wheat crop. The identified indices were then applied across the CHAR study site in Gembloux. The estimated WSR severity was 35.7% on average for the biochar plot in comparison to an average of 42.3% in the reference plots. For WLR severity, there was an average of 22.8% in the biochar plots versus 26.4% in the reference plots. These results were in apparent agreement with the identified crop health percentages in Chapter 3 as a function of century-old biochar enrichment. Furthermore, WSR and WLR showed notable alterations in winter wheat typical reflectance spectra, mostly between the green and red spectral bands, in both RGB and multispectral images available across the study site in Gembloux. This finding may pave the way for future researches developing more robust remote sensing indices to monitor fungal foliar diseases.
In Chapter 5, the capability of fusing remotely-sensed UAV and Landsat-8 images allowing for a high spatio-temporal monitoring of century-old biochar effects on evapotranspiration over the cropping season was explored. The fusion of UAV and Landsat-8 images was performed using additive wavelet transform (AWT), generating sharpened Landsat-8 images with the high spatial resolution as the UAV images. AWT sharpened Landsat-8 images were spatially well-correlated with coarse resolution images, and were well-preserved the spatial details. Surface albedo and normalized difference vegetation index (NDVI) were computed based on reflectance spectra of UAV and Landsat-8 multispectral sensors. We retrieved surface temperature images from FLIR Vue Pro R on board of UAV and Landsat-8 brightness temperature images respectively based on reference ground-based thermal panels and temperature emissivity separation algorithm. The fused images and the meteorological data were used as inputs in the ETLook which is a surface energy balance model based on the Penman-Monteith approach. As such, ETLook model provided high spatio-temporal maps of actual evapotranspiration across the field throughout the 2019 cropping season. The results showed a significant decrease in surface albedo across the biochar plots during the early development stages of winter wheat. Moreover, biochar significantly caused an earlier greening up of wheat plants, and also, stimulated the development of wheat canopies towards the middle of the cropping season. There were however no impacts at the end of the season due to dense wheat canopies covering the aggravated dark colour soil across the biochar patches. Surface temperature was not affected by biochar either at the beginning or towards the end of the season. Neither was there any impact of biochar on actual evapotranspiration over the season. The implemented approach may also develop robust techniques for image fusion of UAV and satellite images to better meet the necessities of the precision agriculture.
