Using NDVI to Determine Early Vegetative Growth and Nitrogen Needs

Jarrod Miller, Extension Agronomist, University of Delaware

Precision nitrogen (N) management through the use sensors has been used to improve nitrogen use efficiency (NUE) for field crops by estimating corn needs during early growth stages (Aula et al., 2020; Cao et al., 2017) Indices have been developed using specific wavelengths of light, such as the normalized difference vegetation index (NDVI) to estimate plant biomass and correlate it to N needs (Holland et al., 2012; Raun et al., 2005a; Dellinger et al., 2008). The use of NDVI has allowed for reductions in N application have been performed without reducing crop yield (Barker and Sawyer, 2012; Aula et al., 2020).

Although initial work in crop sensing was performed with passive (sunshine based) sensors (Figure 1), active sensors, using their own light source, were developed for variable rate applications (Barker and Sawyer, 2019, Simbirsk et al., 2009). However, most of this work has been done to address sidedress timing (V5-V6), but has not included any indications of earlier detection of N needs for the crop due to in-situ soil variability.

This project examined both 1) sensor-based management of early N deficiencies (Emergence to V6); and 2) nutrient uptake based on several starter rates. (V1-V4) may limit yield. This was accomplished by applying four starter N rates (0, 15, 30, and 45 lbs N/acre) on a sandy coastal soil in Georgetown, DE. Drone imagery was collected using a multispectral camera to create NDVI images (Figure 1), while a handheld Trimble Greenseeker (Figure 2) was used to collect active sensor NDVI down the rows. Corn was planted on April 29^th and sidedressed on June 6^th.

Comparison of NDVI Across Handheld and Drone Measurements

Although we have more options for sensors than in the past (drones, satellites, or handheld), they may each make different measurements. This is because an index like NDVI is comparing ground coverage, which will vary depending on how close you are to the canopy (plant leaves covering the soil), or how late in the season you are making the measurements (Figure 2). A handheld sensor is just above the canopy, and may have incoprorated limited amounts of soil, while a drone camera may pick up more soil background (Figure 2). To simulate this, we compared drone images by single row [Row] and two row [Plot]. Measurements based on two rows included more soil in the image, and always produced the lowest NDVI measurements (Figure 3). It is important to consider that NDVI measurements should then made with the same resolution sensor when making N recommendations, to ensure accurate recommendations. This is particularly important when considering the use of zero and high N strips to make variable rate N recommendations, where a handheld sensor and drone imagery may be up to 0.1 units apart (Figure 3). Incorporating handheld NDVI into drone based recommendations could be faulty.

Detection of Plant Growth through Drone and Handheld Sensors

Based on NDVI, differences in starter rates were not detected until May 28^th and it did not matter what method was used (Table 1). The only difference observed was between zero and the other rates, meaning that any N applied (15, 30, or 45 lbs) had similar NDVI, compared to no N at all.

However, we could measure differences in plot variability by drone (“St Dev”) with clear separation by May 14^th. This could indicate that corn growth was more variable when N was applied. By the day of sidedress, corn growth variability was greatest for the 45 lb N starter rates (Table 1). Therefore, even though average NDVI across starter rates was similar, there was more variability in plant growth, which could be due to differences in placement, reactions in the soil, or in-situ soil N availability. This greater variability is lost once corn is sidedressed, and the zero N plots become the most variable, based on NDVI.

Nutrient Uptake at Vegetative (V4/5) and Reproductive (VT) Stages Based on Starter N Rates

Early season nutrient uptake varied by starter N rates, with most nutrient increasing in the plant tissue with increase N application (Table 2). For the zero N plots, tissue N was only 3.25%, but reached 4.07% with the 45lb N rate. Other nutrients that were lowest in the zero N plots includes K, Ca, S, Cu, and Mn. Of those, K, Cu, and S were all similar among 15-45 lb N applications, while Ca and Mg had some differences.. Alternatively, Zn was greatest in the 45 lb plots, while all other rates were similar. Phosphorus uptake was actually greatest in the 0 N plots, and lowest in the 45 lb plots. No other nutrients (Mg, Fe, B, Na) were different for early stage corn.

Summary

The use of NDVI to detect early season N deficiencies was not useful until very close to sidedress, when considering NDVI alone. The handheld and drone based measurements both took almost four weeks from planting (V5) to detect any differences. These differences were also not nuanced, simply showing that 0 N in a sandy, coastal plain field had lower NDVI around V3/V4 than plots receiving between 15-45 lbs of N.

Stand and growth variability was detected earlier, closer to the V2 stage, when using the standard deviation of drone NDVI. This study cannot explain why NDVI was variable, but as an index that measures greenness, it was probably picking up on a range in leaf area index (LAI). This could be due to variable fertilizer placement, interactions with soil (e.g. volatilization, leaching, immobilization), or variable soil N contents. The major influence in plot variability does appear to be the application of fertilizer.

The use of two row plots, which includes greater soil reflectance, had the lowest NDVI, when compared to handheld or row-based drone measurements. This would be important if you are considering using a handheld measurement of a zero or high N plot to make variable rate N applications with satellite or drone imagery. Although a handheld meter may be easier, a larger strip that can be detected by drone or satellite imagery may be necessary.

Sponsors

This research was sponsored by the Maryland Grain Producers Utilization Board (http://www.marylandgrain.org/).