25 IREC Farmers' Newsletter No. 198 — Spring 2017 NDVI vs NDRE Remote sensing in rice historically involved the use of normalised difference vegetation index (NDVI) maps of rice fields generated from satellite, aircraft or drone sources. Although these maps can appear to show significant differences within fields, once the rice crop develops a full canopy, which often occurs before PI, NDVI becomes saturated and cannot detect difference in crop biomass or nitrogen uptake. Our research and other research across the world have shown that above a nitrogen uptake of approximately 80 kg N/ha, NDVI cannot detect difference in crop growth at PI (Figure 1a). Many new sensors have become available in the last few years that measure an additional wavelength band called the red edge, which is a narrow band (710 to 740 nm) located between the visible red and near infrared (NIR) regions of the light spectra. This region of the spectra is very sensitive to changes in foliar chlorophyll content, which is strongly related to plant nitrogen concentration. Vegetation indices that include the red edge region, such as the normalised difference red edge (NDRE) have been found to saturate at a much higher nitrogen concentration than the commonly used NDVI. Our research shows that a better relationship exists between NDRE and PI nitrogen uptake (Figure 1b) than NDVI and PI nitrogen uptake (Figure 1a). Once the rice crop gets past mid-tillering and particularly once it has reached PI, NDVI images of rice fields are of little value, however NDRE images are able to show differences in nitrogen uptake of the crop (Figure 3) at this later stage. Although NDRE is much better than NDVI, its value decreases for PI nitrogen uptake values above approximately 100 kg N/ha, and cannot be used as a direct nitrogen uptake prediction tool for rice varieties that require high levels of nitrogen, e.g. Reiziqp , Sherpap and Opusp . Available sources of NDRE imagery In our research we compared NDRE (red edge) imagery from a number of sensors to nitrogen uptake in rice at PI. These sensors included Worldview 3 satellite and a range of drone mounted sensors (the Parrot Sequioa, MicaSense RedEdge and SlantRange). From Figure 2 it can be seen that a strong relationship between NDRE and PI nitrogen uptake exists for each sensor, but they all flatten out and become of limited value above a nitrogen uptake level of around 100 kg N/ha. The other important factor that is evident in Figure 2 is that the relationship between NDRE and PI nitrogen uptake is different for each sensor so an individual calibration needs to be developed for each sensor. The reason for these differences is thought to be due the different location of the red edge waveband in each sensor and how the red edge shifts left or right with varying levels of plant chlorophyll. 0.40 0.50 0.60 0.70 0.80 0.90 1.00 0 50 100 150 200 250 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0 50 100 150 200 250 Figure 1. Relationship between MicaSense RedEdge data collection from 288 sites over 2 seasons and 4 rice varieties with a) NDVI and b) NDRE. NDVI NDRE PI N uptake (kg N/ha) PI N uptake (kg N/ha) R² = 0.7278 R² = 0.8508 a. NDVI b. NDRE 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0 50 100 150 200 250 SlantRange Worldview Sequoia RedEdge Figure 2. Relationship between NDRE and PI nitrogen uptake for remotely sensed data collected from 132 sites in one season from 4 rice varieties with the SlantRange, Parrot Sequoia, MicaSense RedEdge cameras and the Worldview 3 satellite. NDRE PI N uptake (kg N/ha) R² = 0.9095 R² = 0.9221 R² = 0.93 R² = 0.8326 Figure 3. NDRE image of a rice variety by nitrogen experiment at Yanco in the 2016–17 season. Varieties run from left to right across the plots and pre-permanent water nitrogen rates top to bottom.