26 IREC Farmers' Newsletter No. 198 — Spring 2017 The project team will also monitor and test new and rapidly evolving technology as it becomes available. RIRDC Project PRJ-009772 Moving forward with NIR and remote sensing Acknowledgements This research was co-funded by the NSW Department of Primary Industries and the Rural Industries Research & Development Corporation. Excellent technical support from Chris Dawe and Craig Hodges has contributed significantly to this project. Further information Brian Dunn Research Agronomist (Irrigation) T: 02 6951 2621 E: [email protected] 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0 50 100 150 200 250 Figure 4. Relationship between NDRE and PI N uptake for data collected from the MicaSense RedEdge camera from 288 sites over 2 seasons with 4 rice varieties. The red box shows NDRE values above 0.60 where prediction of PI N uptake is not sufficiently accurate. Figure 5. PI N topdressing recommendations for variety and water depth used in the NIR Tissue Test calibrated against NDRE using the MicaSense RedEdge relationship between NDRE and PI N uptake shown in Figure 3. The red box shows NDRE values above 0.60 where prediction of PI N uptake is not sufficiently accurate. From the NDRE and PI N uptake relationship data shown in Figure 2 it can be seen that both the MicaSense RedEdge and Worldview 3 relationships are relatively steep at lower PI nitrogen uptake values, while the SlantRange and Sequioa relationships are much flatter at the same levels. The slope of these relationships and spread of the data highlights the possibility that the MicaSense RedEdge and Worldview 3 sensors show potential for PI nitrogen uptake predictions below a nitrogen uptake value of approximately 100 kg N/ha (Figure 2). Using MicaSense RedEdge NDRE values Our research showed that below an NDRE of 0.6, the MicaSense RedEdge camera has what we consider an acceptable level of accuracy for predicting PI nitrogen uptake (Figure 4). Above an NDRE of 0.6, the curve flattens significantly and the prediction accuracy is poor and not suitable for the prediction of PI nitrogen uptake, as shown by the red box in Figure 4. The NIR tissue test has nitrogen top dressing recommendations in the program that are based on many years of research and provide recommendations based on crop PI nitrogen uptake for each variety and water depth at microspore. The top dressing recommendations have been converted into a direct relationship with NDRE (Figure 5) based on the relationship between NDRE and PI nitrogen uptake shown in Figure 4. When the area of the graph with NDRE above 0.6 is shaded out (the red box) due to its poor prediction accuracy of PI nitrogen uptake, there is still potential for Koshihikari, Doongara and YRK5 PI top dressing requirements to be predicted using MicaSense RedEdge NDRE values directly (Figure 5). Drone and satellite red edge options From these results, drones and red edge sensors clearly have a role for rice growers looking at spatial variability of nitrogen in individual rice fields. However, as an industry-wide option the use of drones is limited by short flying time (battery life), line of sight regulations, wind and cloud conditions, often 70% image overlap requirements and the short daily data collection period (only 3 hours either side of solar noon). Satellite imagery is really the only practical option for covering all of the rice crops in the industry in the short PI timing window. The Worldview 3 satellite based sensor has shown considerable potential with good correlations with PI nitrogen uptake, daily revisit time and 1.25 m resolution. However images are very expensive, i.e. approximately $55 per km2 with a minimum capture of 100 km2 . The Sentinel 2 sensor offers free imagery, a 10-day revisit time and a red edge waveband (705 nm), but the resolution of the red edge band is 20 m, which is too coarse for measuring zonal variability in rice fields that contain contour banks. Another option, which is yet to be investigated, is the RapidEye satellites that have a red edge waveband, daily revisit time and 5 m resolution. Data from RapidEye is much more cost effective than Worldview 3 but future research needs to determine its accuracy at predicting PI nitrogen uptake and if the 5 m pixel size is suitable. Remotely sensed images of rice crops can highlight within field variability of crop nitrogen uptake. Through targeted agronomy this information can greatly improve crop management decisions leading to improved grain yield and quality and ensure sustainability of rice growing in Australia. Where to now? In the 2017–18 rice season a calibration will be developed between NDRE and PI nitrogen uptake for the RapidEye satellites testing its accuracy and potential for commercial use. Options will be investigated into making satellite based red edge images available to growers in a timely manner at an acceptable price. Calibrations for the MicaSense RedEdge sensor will continue to be updated across more rice varieties. NDRE PI N uptake (kg N/ha) 0 20 40 60 80 100 120 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 Reiziq, Sherpa - Deep Reiziq, Sherpa - Shallow Koshi, YRK5 - Deep Koshi, YRK5 - Shallow PI N topdressing rate (kg N/ha) NDRE (MicaSense RedEdge)