Sustainable Sidedress Nitrogen Applications for Early Corn and Cotton Using Unmanned Aerial Systems

Parker_Presentation_Spring2022

Please see presentation for yield maps and graphs.

Crop Canopy Sensing to Detect Early N Status

Results from crop canopy sensing missions, conducted between 2020-21 in corn and cotton, are discussed in relationship to flight speed trials, biophysical parameters relating crop canopy N status to imagery outputs, and the fitting of vegetation indices (VIs) to sensitivity equivalents (SEqs). Data are presented in tabular format detailing the statistical outcomes that comprise the resulting quotient (linear regression slope divided by the root mean square error) from relating a biophysical parameter to extracted imagery VIs. It is worth noting the canopy sensing missions featured both corn (erectophile) and cotton (planophile) canopy types. Variations in canopy reflectance across many other crops may be strategically related to the data discussed herein.

Because of the unique nature of weather and technical events affecting mission quality during each year of this study, data will be presented by year instead of by crop variety. The crop canopy section concludes with a synopsis of the general findings and assumptions.

Cotton 2020

The 2020 cotton sensing mission was flown on July 11, 2020. A quad-copter was employed to fly four separate flights at speeds of 16 km h­^-1 (10 mi h^-1), 24 km h^-1 (15 mi h^-1), 32 km h^-1 (20 mi h^-1) and 40 km h^-1 (25 mi h^-1). From the July 11 sensing mission, both fixed-rate N and VRN applications were made on July 15, 2020.

            There was no consistently high-performing VIs among speeds 16 km h­^-1 (10 mi h^-1) and 24 km h^-1 (15 mi h^-1) in the 2020 North Cotton plot. FENDVI produced the highest sensitivity equivalents for the 16 km h­^-1 (10 mi h^-1) when modeled with SPAD. SPAD sensor do not sample for green (550 nm) reflectance as is incorporated into the FENDVI but, instead, measures the absorbance in red and NIR wavelengths. Therefore, it is not possible to state FENDVI and SPAD are highly relatable other than through SEq modeling. The FENDVI algorithm incorporates NDRE, NDVI, and green bands of a sensor. Opposite of red-edge detection and shift relating to canopy N status, green band reflectance is indirectly related with canopy N status during photosynthesis and, therefore, inadvertently detectable by SPAD meters (Mutanga and Skidmore, 2004). However, research is needed to decouple green, red, red-edge, and NIR band reflectance behavior from other phenomenon especially considering SPAD does not account for red-colored soil albedo detected with the above-canopy sensing tools. Development of a SPAD tool that includes the green (550 nm) band may increase understanding of individual band reflectance’s and improve canopy N detection through proximal and sUAS sensing methods.

 The highest sensitivity equivalents for 16 km h­^-1 (10 mi h^-1) when modeled with Leaf N was achieved by BENDVI, a VI that incorporates NIR, blue, and green bands. For the 24 km h^-1 (15 mi h^-1) speed, SCCCI had the highest sensitivity equivalents when modeled with SPAD. The 24 km h^-1 (15 mi h^-1) speed for VIs modeled with leaf N resulted in FENDVI, a VI incorporating a modified green band with SCCCI, having the highest sensitivity equivalents.

When canopy N status was assessed using SPAD, SCCCI attained the highest sensitivity equivalents value for the 32 km h­^-1 (20 mi h^-1) speed. The highest sensitivity equivalents result for the 32 km h­^-1 (20 mi h^-1) speed was FENDVI when modeled with Leaf N. SCCCI modeled with SPAD achieved the highest sensitivity equivalents for 40 km h^-1 (25 mi h^-1). An increase of 0.13081 was noted in SCCCI with SPAD sensitivity equivalents among speed increase from 32 km h­^-1 (20 mi h^-1) to 40 km h^-1 (25 mi h^-1). Inconsistent with the 32 km h­^-1 (20 mi h^-1) speed, BENDVI modeled with Leaf N outperformed other VIs in terms of sensitivity equivalents.

Sensitivity Equivalents results across all speeds did not reveal a consistent ideal VI for modeling with SPAD or Leaf N. A probable cause for the inconsistent results is related to the previously mentioned “tilt” of the quad-copter coupled with the inability to remove outliers due to small sample count.

The 2020 south cotton plot results imply that NDVI modeled with SPAD produce the highest sensitivity equivalents value at a speed of 16 km h­^-1 (10 mi h^-1). Both SPAD and NDVI employ red and NIR bands to estimate chlorophyll (and thus, related N status). BENDVI modeled with Leaf N sensitivity equivalents results revealed that BENDVI may be preferred to other VIs at the 16 km h­^-1 (10 mi h^-1) speed. 24 km h^-1 (15 mi h^-1) results indicated that BENDVI modeled with SPAD may be favored compared to alternative VIs.

At a value of 0.03267, NDVI modeled with Leaf N sensitivity equivalents results achieved a higher value than comparable VIs. One single VI sensitivity equivalents value was not identified as consistently higher at speeds of 16 km h­^-1 (10 mi h^-1) and 24 km h^-1 (15 mi h^-1). Subsequent studies using sUAS and sensing technologies necessitate an increase in sample size collection that will likely mitigate sampling errors related to on-farm topographic changes. Moreover, a highly consistent and uniform off-farm research methodology made up of similar small-plot research sites on limited soil substrate may also improve VI sensitivity.

NDVI modeled with SPAD produced a higher sensitivity equivalents value than other VIs at a speed of 32 km h­^-1 (20 mi h^-1). A value of 2.27549 was produced by BENDVI modeled with SPAD, thus, implying that BENDVI is optimal to other VIs at a speed of 32 km h­^-1 (20 mi h^-1). For the 40 km h^-1 (25 mi h^-1) speed, FENDVI had the highest sensitivity equivalents when modeled with SPAD. NDVI modeled with Leaf N sensitivity equivalents resulted in higher values than compared VIs for the 40 km h^-1 (25 mi h^-1) speed.

            Consistent with the 2020 cotton plot north results, the 2020 cotton plot south results did not have a persistent VI that ranked above other VI sensitivity equivalents. The quad-copter tilt and small sample count may factor into inconsistent results.

Corn 2021

The sensing mission for the 2021 corn crop was flown on May 1, 2021. Two separate flights at speeds of 32 km h­^-1 (20 mi h^-1) and 64 km h^-1 (40 mi h^-1) were conducted using both quad-copter and fixed-wing flight platforms, respectively.

From the May 1 sensing mission, both fixed-rate N and VRN applications were made on May 8, 2021, utilizing quad-copter data.

The 2021 corn plot revealed that SCCCI, when modeled with SPAD, produced the highest sensitivity equivalents value at a speed of 32 km h­^-1 (20 mi h^-1). FENDVI, a modified green band divided by the SCCCI, performed best when modeled with Leaf N at the 32 km h­^-1 (20 mi h^-1) speed. A decrease of 0.0813 was noted in SCCCI modeled with SPAD sensitivity equivalents among an increase of speed from 32 km h­^-1 (20 mi h^-1) to 64 km h^-1 (40 mi h^-1). FENDVI modeled with Leaf N experienced a slight increase in sensitivity equivalents of 0.04157 when speed increased from 32 km h­^-1 (20 mi h^-1) to 64 km h^-1 (40 mi h^-1).

Among both quad-copter (32 km h­^-1 (20 mi h^-1)) and fixed-wing platforms (64 km h^-1 (40 mi h^-1)), SCCCI modeled with SPAD experienced a decrease. The 0.0813 decrease in SCCCI modeled with SPAD sensitivity equivalence is likely due to the unreliability of SPAD. SPAD requires an elevated level of precision and consistency when collecting samples that is not typically present when conducting on-farm sampling. Xiong et al. (2015) found that SPAD readings are highly variable depending upon leaf and environmental characteristics.

When performing on-farm research, SPAD is not a preferred biophysical parameter by which to assay canopy leaf N status. The physically demanding and time-consuming effort required to sample using SPAD are, in themselves, tremendously limiting factors. Added effects of end-user subjectivity and inconsistent sampling method necessitates the development of objective on-the-fly field measurements techniques that can, in almost real time, determine crop canopy status. Moreover, a SPAD lacking green band assessment may also reduce its overall efficacy for use on-farm. A reduction in green band reflectance is highly and indirectly relatable to increasing canopy N status.

FENDVI modeled with Leaf N produced an increase of 0.04157 in sensitivity equivalents with an increase in speed from quad-copter (32 km h­^-1 (20 mi h^-1)) to fixed-wing platforms (64 km h^-1 (40 mi h^-1)). The 0.04157 increase expresses that fixed-wing flight may produce higher sensitivity equivalents than quad-copter flight.

Cotton 2021

On June 18, 2021, the first sensing mission was completed for the 2021 cotton crop during pinhead square development stage. Due to a fixed-wing platform malfunction, only a quad-copter flight was executed at a speed of 32 km h­^-1 (20 mi h^-1) on June 18. From the June 18 quad-copter only sensing mission, both fixed-rate N and VRN applications were made on June 20, 2021, utilizing quad-copter data. A second sensing mission was accomplished on July 28, 2021, at mid-bloom development stage, employing both quad-copter (32 km h­^-1 (20 mi h^-1)) and fixed-wing (64 km h^-1 (40 mi h^-1)) flight systems. The July 28 mission was necessary for comparison of quad-copter and fixed-wing flight data in similar environmental conditions.

The VRN quad-copter flight for the 2021 cotton crop resulted in the NDRE modeled with SPAD sensitivity equivalents value rising above compared VIs. NDRE may be ideal when modeled with SPAD at a speed of 32 km h^-1 (20 mi h^-1). When modeled with leaf N, SCCCI contained a higher sensitivity equivalents value higher than other VIs. At a speed of 32 km h^-1 (20 mi h^-1), SCCCI modeled with SPAD may be preferred in terms of sensitivity equivalents.

When modeled with SPAD, SCCCI possessed a higher sensitivity equivalents value than other VIs at a speed of 32 km h^-1 (20 mi h^-1). Consistent with SCCCI modeled with SPAD at 32 km h­^-1 (20 mi h^-1), SCCCI modeled with Leaf N attained a higher sensitivity equivalents value than compared VIs at 32 km h­^-1 (20 mi h^-1). An increase of 0.07259 was noted in SCCCI modeled with SPAD when speed was increased from 32 km h­^-1 (20 mi h^-1) to 64 km h^-1 (40 mi h^-1). Inconsistent with Leaf N sensitivity equivalents results at 32 km h­^-1 (20 mi h^-1), NDRE achieved the highest sensitivity equivalents value when modeled with SPAD for the 64 km h^-1 (40 mi h^-1) speed.

The mid-bloom sensing mission resulted in SCCCI earning the highest sensitivity equivalents when modeled with SPAD. The increase in speed from quad-copter (32 km h­^-1 (20 mi h^-1)) to fixed-wing (64 km h^-1 (40 mi h^-1)) led to an increase in SCCCI modeled with SPAD sensitivity equivalents of 0.07259. The increase of 0.07259 may infer that fixed-wing is ideal to quad-copter for sensitivity equivalents in VIs modeled with SPAD. When speed increased from 32 km h­^-1 (20 mi h^-1) to 64 km h^-1 (40 mi h^-1), VIs modeled with Leaf N’s sensitivity equivalents values resulted in the higher values for SCCCI and NDRE, respectively.

Concluding Remarks on Crop Canopy Sensing

Sensing missions for this study resulted in numerous novel findings. The first being that NDRE-based indices performed well in comparison to other VIs in all individual studies. Multiple studies support similar findings revealing NDRE-based VIs produced the strongest relationship with leaf N content (Varco et al., 2013; Raper and Varco, 2014; Fox, 2015; Sumner et al., 2021). Among all VIs, NDRE-based VIs may yield the highest relation to crop canopy N levels. This study does not suggest there is verifiable causation between canopy N status and VI efficiency without further study into the complex relationship of lateral, right red-edge shift and accompanying reduction in green reflectance with increasing canopy N status. However, noted to date herein, no SPAD canopy analysis tools have successfully incorporated the green band, which may further elucidate canopy N status in-field.

A loss in VI efficiency was not evident among an increase in speed from quad-copter (32 km h­^-1 (20 mi h^-1)) to fixed-wing (64 km h^-1 (40 mi h^-1)). In select cases, faster sensing speeds may result in a higher level of accuracy. Fixed-wing sensing platforms provided more efficacy than quad-copter platforms in these studies, but further studies are needed. Overall, fixed-wing platforms may provide a faster alternative to quad-copter platforms without a loss of data accuracy. However, the results from this study suggest the sensor attitude onboard the aircraft may affect sensing results, but relevant research is needed to decouple speed and attitude effects.

2020 Cotton Plot North

The 2020 cotton crop sensing flight was performed on July 11, 2020. A quad-copter was employed at a speed of 32 km h^-1 (20 mi hr^-1) and altitude of 121.92 m (400 ft) for the flight. Multispectral sensing was conducted by a MSRE aboard the quad-copter. Both fixed- and variable-rate nitrogen applications were made on July 15, 2020, using data from the 32 km h^-1 (20 mi hr^-1) flight on July 11.

The mean N rate was 13.45 kg ha^-1 (12 lb ac^-1) for the VRN application dataset. The mode for the VRN dataset was 15.69 kg ha^-1 (14 lb ac^-1). A rate of 14.57 kg ha^-1 (13 lb ac^-1) was found to be the median of the VRN dataset. The total season variable-rate N section of the plot ranged from 89.67 kg ha^-1 (80 lb ac^-1) to 123.29 kg ha^-1 (110 lb ac^-1).

The overall mean yield for the 2020 cotton plot north was 1,445.89 kg ha^-1 (1,290 lb ac^-1). An average yield of 1,448.14 kg ha^-1 (1292 lb ac^-1) was observed for the VRN application area of the plot. The fixed-rate application portion of the plot resulted in an average yield of 1,440.29 kg ha^-1 (1,285 lb ac^-1). 

2020 cotton plot North NUE by N rate class

Refer to Materials and Methods for unit conversions

An increase in yield was evident among an increase in VRN application from 0 kg ha^-1 (0 lb ac^-1) to 17 kg ha^-1 (15 lb ac^-1). Once yield peaked at the 10 kg ha^-1 (9 lb ac^-1) to 17 kg ha^-1 (15 lb ac^-1) stage, yield experienced a decrease as VRN application increased from 18 kg ha^-1 (16 lb ac^-1) to 34 kg ha^-1 (30 lb ac^-1). The high yield at the 10 kg ha^-1 to 17 kg ha^-1 class may suggest that 10 kg ha^-1 to 17 kg ha^-1 class is the optimal rate for cotton N application. Both VRN and total available NUE experienced an increase when VRN application increased from 0 kg ha^-1 (0 lb ac^-1) to 26 kg ha^-1 (23 lb ac^-1). VRN NUE continued to rise as VRN application rates increased to 34 kg ha^-1 (30 lb ac^-1) while total available NUE decreased by 0.68. 

2020 cotton plot North Tukey Studentized range of yield means by treatment class

LSMeans with same letter are not significantly different

Yield units are kg ha^-1

Classes are as follows: 1,2,3,4 VRN Treatments = 1-4. Fixed-rate Treatment = 5

Tukey’s studentized range test revealed the yield for treatment class 1,2,3,4, and 5 were not significantly different. According to this test, treatment class one may be preferred to classes 2,3,4, and 5 due to lower input of N without a significant loss in yield. 

The final, high rate, VRN class resulted in a 61.71 kg ha^-1 (55 lb ac^-1) lower yield than the fixed-rate N application. Edge-effect is a probable cause for the decrease in yield since all locations for the high rate VRN applications were located on the outer edge of the plot. Future on-farm research should integrate the edge-effect and spatial dynamics related to outer rows and consider possible removal of marginal edge production from results.

2020 Cotton Plot South

The 2020 cotton plot south employed the same platform and payload as the 2020 cotton plot north did. Fertilizer date and rates for the 2020 cotton plot south are consistent with 2020 cotton plot north.

A season total of 123.29 kg ha-1 (110 lb ac-1) was applied for the fixed-rate N amount. A rate of 21.29 kg ha^-1 (19 lb ac^-1) was the mean for the VRN application dataset. The mode for the VRN dataset was calculated to be 26.90 kg ha^-1 (24 lb ac^-1). A rate of 23.54 kg ha^-1 (21 lb ac^-1) was found to be the median of the VRN dataset. The variable-rate N section of the plot received a total season amount ranging from 89.67 kg ha^-1 (80 lb ac^-1) to 123.29 kg ha^-1 (110 lb ac^-1). 

The 2020 cotton plot south attained an overall average yield of 1,173.53 kg ha^-1 (1,047 lb ac^-1). A mean yield of 1,185.86 kg ha^-1 (1,058 lb ac^-1) was noted for the fixed-rate portion of the plot. The VRN applicated portion of the plot achieved an average yield of 1,163.92 kg ha^-1 (1,038 lb ac^-1). 

2020 cotton plot South NUE by N rate class

Refer to Materials and Methods for unit conversions

As VRN application increased from 0 kg ha^-1 (0 lb ac^-1) to 34 kg ha^-1 (30 lb ac^-1), yield decreased from 1350.86 kg ha^-1 (1205 lb ac^-1) to 1149.32 kg ha^-1 (1025 lb ac^-1). The reason for the decrease in yield was due to heavy rain from Hurricane Laura. The heavier bolls loaded with N were shed from the plant during the heavy wind and rain. The NUE for both VRN and total available sections increased as VRN application increased from 0 kg ha^-1 (0 lb ac^-1) to 16 kg ha^-1 (14 lb ac^-1). The VRN NUE increase corresponded with a decrease in N rate applied as VRN application progressed from 17 kg ha^-1 (15 lb ac^-1) to 34 kg ha^-1 (30 lb ac^-1). 

2020 cotton plot South Tukey Studentized range of yield means by treatment class

LSMeans with same letter are not significantly different

Yield units are kg ha^-1

Classes are as follows: 1,2,3,4 VRN Treatments = 1-4. Fixed-rate Treatment = 5

Yield means for treatment classes 1,2,3,4, and 5 were not significantly different based on Tukey’s studentized range test. Treatment class 4 may be preferred due to the decrease in N input without significantly decreasing yield. 

As fertilizer N rates progressed, yield encountered a drastic decrease for both the VRN applications and the fixed-rate application. Hurricane Laura produced a unique scenario where the fixed-rate N application performed poor. The heavy bolls possessed by the plants that received the high N application rate were too cumbersome to remain attached to the plant amidst high winds and heavy rain.

2021 Corn

The 2021 corn crop sensing mission was conducted on May 1, 2021. A quad-copter platform was equipped with a MicaSense RedEdge multispectral camera and flown at 32 km h^-1 (20 mi h^-1).

An overall average and fixed-rate average yield of 14.66 Mt ha^-1 (217 bu ac^-1) was observed for the 2021 corn plot. The VRN application portion of the plot exceeded the overall average with a yield of 14.69 Mt ha^-1 (218 bu ac^-1). 

2021 corn plot NUE by N rate class

Refer to Materials and Methods for unit conversions

Corn yield increased from 14.90 Mt ha^-1 (221 bu ac^-1) to 15.23 Mt ha^-1 (226 bu ac^-1) as VRN application increased from 85 kg ha^-1 (75 lb ac^-1) to 122 kg ha^-1 (108 lb ac^-1). The peak in yield at 15.23 Mt ha^-1 (226 bu ac^-1) may suggest that the VRN application range of 111 kg ha^-1 (99 lb ac^-1) to 122 kg ha^-1 (108 lb ac^-1) is the optimal rate. Furthermore, corn yield decreased as VRN application reached the 123 kg ha^-1 (110 lb ac^-1) to 134 kg ha^-1 (120 lb ac^-1) range. The VRN NUE exhibited a decrease as VRN application increased. The total available N exhibited a linear decrease as VRN application increased. 

2021 Corn plot Tukey Studentized range of yield means by treatment class

LSMeans with same letter are not significantly different

Yield units are Mt ha^-1

Classes are as follows: 1,2,3,4 VRN Treatments = 1-4. Fixed-rate Treatment = 5

Tukey’s studentized range test discovered that the yield of treatments 1,2,3,4, and 5 were not significantly different. A gap of 0.33 Mt ha^-1 separated treatment 1 from treatment 3. Treatment 1 may be favored to treatments 2,3,4, and 5 due to reduced N input without significant decrease in yield. 

Yield peaked at 15.23 Mt ha^-1 (226 bu ac^-1) followed by a decline to 15.16 Mt ha^-1 (225 bu ac^-1). The fixed-rate portion of the plot performed lower than the highest VRN application rate at a yield of 15.11 (224 bu ac^-1). Edge-effect may be a likely culprit to the decline of yield among the highest VRN application and the fixed-rate application.

2021 Cotton

The crop sensing mission for the 2021 cotton crop was conducted on June 18, 2021. A quad-copter platform was equipped with a MicaSense RedEdge multispectral camera and flown at 32 km h^-1 (20 mi h^-1).

The fixed-rate total season N application that the crop received was 89.67 kg ha^-1 (80 lb ac^-1). A rate of 84.87 kg ha^-1 (75 lb ac^-1) served as the mean for the VRN application dataset. The mode and median for the VRN application dataset were 85.34 kg ha^-1 (76 lb ac^-1). Total season range of the VRN applicated portion of the plot was 56 kg ha^-1 (50 lb ac^-1) to 89.67 kg ha^-1 (80 lb ac^-1).

For the 2020 season, a heavier, early season N treatment of 89.67 kg ha^-1 (80 lb ac^-1) was applied as the starter N rate. The 2021 cotton crop only received a starter N rate of 33.63 kg ha^-1 (30 lb ac^-1). The decrease in VRN application variability for the 2021 season may be linked to a reduced need for N at pinhead square as a result of the heavier starter N rate. 

An overall average yield of 91.93 Mt ha^-1 (1,366 bu ac^-1) was noted for the 2021 cotton plot. The fixed-rate only portion of the plot achieved a mean yield of 92.66 Mt ha^-1 (1,377 bu ac^-1). Plants that received the VRN application achieved an average yield of 91.01 Mt ha^-1 (1,352 bu ac^-1). Low topographic field areas are noted to have a significantly reduced yield. 

2021 cotton plot NUE by N rate classes

Yield increased from 1586.50 kg ha^-1 (1415 lb ac^-1) to 1663.44 kg ha^-1 (1483 lb ac^-1) as VRN application rate increased from 73 kg ha^-1 (65 lb ac^-1) to 84 kg ha^-1 (75 lb ac^-1). A decrease in yield was found as VRN application rate increased from 85 kg ha^-1 (76 lb ac^-1) to 89 kg ha^-1 (79 lb ac^-1). Total available NUE increased from 8.31 to 8.62 with transition from the first VRN application class to the second. After peaking at 8.62, total available NUE decreased to 7.41 as VRN application increased. VRN NUE consistently decreased as VRN application increased from 73 kg ha^-1 (65 lb ac^-1) to 89 kg ha^-1 (79 lb ac^-1). 

2021 cotton plot Tukey Studentized range of yield means by treatment class.

LSMeans with same letter are not significantly different

Yield units are kg ha^-1

Classes are as follows: 1,2,3,4 VRN Treatments = 1-4. Fixed-rate Treatment = 5

According to Tukey’s studentized range test, treatment classes 1,2,3,4, and 5 are not significantly different. A yield range of 97 kg ha^-1 exists between treatment class 3 and treatment class 4. Treatment class 1 may be the preferred rate due to decreased input N without a significant yield loss. 2021 cotton yield modeled against variable- and fixed-rate application rate and is noted.

Yield increased as fertilizer rate increased from 73 kg ha^-1 (65 lb ac^-1) to 84 kg ha^-1 (75 lb ac^-1). A decrease in yield occurred as fertilizer rate increased from 85 kg ha^-1 (76 lb ac^-1) to 89 kg ha^-1 (79 lb ac^-1). The highest VRN application (89 kg ha^-1 (79 lb ac^-1)) resulted in an 83 kg ha^-1 (74 lb ac^-1) lower yield than the fixed-rate applied area. The low yield of the final VRN application point may be due to limited sample locations that fell within the highest VRN applied area.

Discussion on Variable Rate N Demonstration Using sUAS

The variable rate N demonstration using sUAS provided insightful results. Variable rate N demonstration data from 2020 and 2021 found that farmers may reduce N input without a significant reduction in yield. Reduced N input would improve sustainability for the future of our crop land and the wildlife that surrounds the land. This study confirmed that sUAS can successfully be used to implement variable rate N practices in an on-farm situation.

Results Conclusions

Through the results of this study, the following is inferred:

1. sUAS sensing mission data provided sufficient evidence suggesting that fixed-wing flight platforms may be successfully deployed to estimate crop canopy N status most effectively, and the results of fixed-wing missions were not significantly different than that of the slower, quad-copter missions.
2. The SCCCI and FENDVI VIs most often were highly related by SEq to early corn and cotton canopy N status. In general, compound red-edge VIs outperformed simple slope-based VIs in most cases.
3. The VariRite^™ technology was successfully implemented in a producer’s field using VI calibrated imagery captured on an sUAS. Deployment of a VRN prescription across corn and cotton provided evidence for reduction of fertilizer N inputs towards the mutual goals of N redistribution and sustained yield.