News

Estimation of canopy parameters using remote sensing in wheat

Author: Lukas Koppensteiner

Every year farmers are confronted with the complex task of managing their crops. The extreme weather conditions of recent years are further complicating decision-making in agriculture.
Here the crucial question is: How can fertilization, plant protection, and irrigation be managed optimally?

An essential basis for optimized crop canopy management is information on the current conditions in the field. Classic methods, e.g., manual measurements of plant height, canopy density, and above-ground dry mass, as well as laboratory analyses on nitrogen and chlorophyll, are time-consuming and costly. A fast and simple alternative is remote sensing. This is defined as collecting information on objects, in our case crop canopies, without direct contact. Exemplary sensors used in remote sensing are hand-held or tractor-mounted sensors as well as aerial or satellite images.

As part of the DiLaAg project, we conducted a field experiment at the experimental farm Groß-Enzersdorf in the growing-season 2019/20. The objective was to collect information on various crop canopy parameters in wheat using remote sensing technologies.
Winter- and spring wheat plots were set up with differing nitrogen fertilization levels. In these plots, data on various crop canopy parameters were continuously collected traditionally. Examples for these parameters are developmental stages, soil coverage, plant height, above-ground dry mass, nitrogen content, chlorophyll content, and water content.

Furthermore, measurements were conducted using various sensors, such as RGB cameras, multispectral and hyperspectral sensors, and smartphone apps. These measurements were performed on different platforms, e.g., hand-held and mounted on a tractor or a drone. Additionally, satellite data was collected in large plots.

Additionally, satellite data was collected in large plots.
In the next step, classically collected data on canopy parameters and sensor measurements will be analyzed. These analyses will be conducted together with colleagues from the University of Technology Vienna. Examples of multispectral and hyperspectral data analysis procedures are vegetation indices, radiative transfer models, and neural networks.

The objective is to set up operational applications for farmers and farm consultants based on our findings. These applications will use the information on the current condition in wheat canopies to support optimized crop management.

Figure 1: Field experiment at the experimental farm Groß-Enzersdorf (modified after Google Maps 2019). Winter wheat was sown in the yellow area. Spring wheat was sown in the white area. The small plot field experiment is located in the center, which is surrounded by large plots.

Figure 2: RGB (left) and CIR images (colored infrared, right) of winter wheat (hand-held) (4.5.2020). RGB images can be used to estimate soil coverage. In the next step, RGB and CIR images can be analyzed regarding relationships with various canopy parameters.

Figure 3: Aerial image of the small plot field experiment using an RGB camera (8.5.2020). The small plot field experiment consists of winter and spring wheat plots. The winter wheat plots are varying visibly according to their respective nitrogen fertilization level. The gaps in the plots indicate the positions at which destructive plant sampling was conducted.

Figure 4: False-color composite of Groß-Enzersdorf based on a Sentinel-2 image (22.4.2020) (modified after ESA 2020). Near-infrared, red, and green are displayed as red, green, and blue. Bare soil and spring crops, which were sown recently, are shown in blue and green. Red indicates well-established vegetation, e.g., healthy winter crops.

Figure 5: NDVI map of Groß-Enzersdorf based on a Sentinel-2 image (compare Figure 4, 22.4.2020) (modified after ESA 2020). In the next step, vegetation indices will be analyzed regarding relationships with various canopy parameters.