18th Congress of the European Society for Agronomy

From NDVI to variable nitrogen application map – Precision agriculture

Aug 28, 2024, 11:40 AM

15m

Les Horizons (2nd floor) (The Couvent des Jacobins)

Oral Synergies of technologies Sensing & data

Speaker

Emir Memic (University of Hohenheim)

Description

1. Introduction
   In Baden-Württemberg (B-W, Germany) N application for high protein quality winter wheat (E) is commonly split into three application dates: tillering, stem elongation and late booting. Total possible N application amount of a field is defined by B-W N-Düngebedarfsermittlung (N-requirement determination). Farmers in B-W commonly apply homogeneously at first N application date about 30-35%, about 40-45% on second application and the rest at third N application date (of N-Düngebedarfsermittlung total). However, this approach does not take in-field variability into account.
   Identification of in-field variability can be conducted based on normalized difference vegetation index (NDVI) in the form of satellite image-based remote sensing vegetation monitoring. It is expected that “healthy” biomass causes higher reflectance rates in the near infra-red (B08) and high absorption in red (B04) enabling “detection” of biomass with higher chlorophyll content. It is hypothesized that higher N application rates are needed where the index is higher, as it “indicates” good conditions for plant growth.
   For demonstrating a conceptual framework that uses NDVI for developing variable N application maps, free of charge satellite images downloaded with Copernicus Browser (CDSE 2024) were used. NDVI of a specific field was calculated, potential site-specific zones were delineated and the degree of in-field variability, in relative terms depending on NDVI value was quantified. Detailed instructions on downloading satellite images and using QGIS for identifying in-field variability and producing site-specific N application maps can be found in the GitHub repository (https://github.com/memicemir/ndvi_to_variable_N_application)

2. Materials, methods
   In the case of variable N application rates, farmers have to delineate site-specific zones with different yield potential. For this short study only the first N application (at tillering) was tested, where approximately 70 kg N ha-1 can be spread in the field. In the case of site-specific applications, farmers reduce N application amounts in certain site-specific units (with lower yield potential) and apply higher N amounts in areas with higher yield potential without exceeding 70 kg N ha-1 in total. In this study NDVI (Eq.1) was calculated for each site-specific unit (average value of all pixels in one site-specific unit) and used as an indicator of biomass development on 28 February 2021.
   NDVI = (B08-B04)/(B08+B04) (1)
   Because site-specific NDVI was normally distributed in the field, all index values were classified as either low, medium, or high. The NDVI classes were formed by arbitrary splitting min/max range of field NDVI values into three groups with equal intervals.

3. Results
   Figure 1 shows the NDVI index across different site-specific units with three NDVI range categories: low (white), medium (light green), and high (dark green), representing different plant growth potential. Based on the normal distribution of the index, medium category (0.358 ≤ NDVI < 0.423, Figure 1) was set to 70 kg N ha-1 while low and high corresponded to ±20 kg N ha-1, respectively. The ±20 step was used in order to avoid too large differences between min and max N application rates. After translating NDVI index variability into site-specific N applications, the average N amount applied in the field was 68 kg N ha -1.
   Figure1 (insert here!)
   Figure 1 Spatial NDVI index distribution according to three NDVI index categories used to determine N application rates at an early growth state (BBCH 22-25) of winter wheat.

4. Discussion and conclusion
   Depending on the distribution of low, medium, and high NDVI zones, farmers have to be aware of maximum N kg ha-1 allowance. If there are more high-NDVI zones in the field than low (with medium 70), the farmer might end up exceeding the total N allowance limit. It has to be kept in mind that even if the plants are not growing in specific parts of the field, it does not automatically mean that there is no N in the soil or that N is the growth limiting factor.

5. Acknowledgment
   https://diwenkla.uni-hohenheim.de/

6. References
   Copernicus Data Space Ecosystem (CDSE), Modified Copernicus Sentinel data (2024), processed in Copernicus Browser. https://browser.dataspace.copernicus.eu/
   QGIS 2024. QGIS.org, Geographic Information System. QGIS Association. http://www.qgis.org