I am an agricultural engineer and currently doing a PhD in precision agriculture. Last year, I was collecting yield data from a silage corn field at harvesting time. The field was quite big (50 hectares) and during the harvesting process, I noticed different zones infested by weeds. The weeds become more clear to recognise after the mowing of corn where only about 15cm of corn stem remains in the soil and the whole plant collected for silage. At this stage, I found some zones which are infested by weeds in the form of patches and covering the area between corn rows. These weeds compete with corn in nutrition, water and decrease the yield quantity and quality. Also, these weed patches will continue for the next year in the same place or will spread in the whole field if not treated. In this situation, I wondered, can we monitor these weeds by remote sensing?
Nowadays, Remote Sensing (RS) role in agriculture is improving continually with different applications and ideas. One of the common applications of RS is to monitor crop health and predict crop yield. Currently, a high resolution and free satellite images are available every five days, thanks to the Sentinel-2 satellite platform. There are different vegetation indices to describe crop health from satellite images such as the Normalized Difference Vegetation Index (NDVI). The NDVI has a range from zero to one, and the higher NDVI value means more healthy crop compared to lower values in the same field.
My idea is to process sentinel-2 or drone images after harvesting directly and develop the NDVI map. At this stage, the NDVI value will describe the weed distribution through the field where higher NDVI value will refer to higher weed infestation zone compared to lower NDVI zones. Here, the NDVI map will act as a prescription map for herbicide application. Now, the herbicide sprayer will spray only the weed patches instead of the whole field. This method will reduce the herbicide application and improve crop yield quantity and quality.
In order to perform an experiment to evaluate this idea, I designed these steps:
- Select some corn fields which are more than 10 ha to act as experimental fields.
- At harvesting time, developing a weed infestation map as a ground truthing data.
- Compare NDVI images from Sentinel-2 or drone images with the weed infestation map.
As a preliminary work, I tried to do a visual interpretation between the infested zones in my previous field, and Sentinel-2 images and I assume that it will work. The previous steps will evaluate the proposed idea in a proper way and it may be repeated with other crops such as wheat.
This simple method will support farmers to improve their yield and reduce their expenses. On the other hand, reducing herbicide application and improving the spraying efficiency will lead to healthy food, preserve our environment and support sustainable agriculture.
The attached photos show different weed infestation degree from the same field after mowing corn to prove the idea.