The influence of Unmanned Agricultural Aircraft System design on spray drift

Der Einfluss der Bauart unbemannter Luftfahrzeuge auf die Abdrift der Spritzflüssigkeit

Journal für Kulturpflanzen, 72 (1). S. 1–11, 2020, ISSN 1867-0911, DOI: 10.5073/JfK.2020.01.01, Verlag Eugen Ulmer KG, Stuttgart

Correspondence

Dr. Andreas Herbst, Julius-Kühn-Institut (JKI), Federal Research Centre for Cultivated Plants, Institute for Application Techniques in Plant Protection, Messeweg 11/12, 38104 Braunschweig, Germany, E-Mail: andreas.herbst@julius-kuehn.de

Accepted

18 October 2019

This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/deed.en).
Dies ist ein Open-Access-Artikel, der unter den Bedingungen der Creative Commons Namensnennung 4.0 International Lizenz (CC BY 4.0) zur Verfügung gestellt wird (https://creativecommons.org/licenses/by/4.0/deed.de).

Symbols and abbreviations

Fig. 1. UAASs involved in the study (a – 1 rotor, b – 6 rotor, c – 8 rotor (TTA) and d – 8 rotor (DJI)

All UAASs were equipped with individual spraying systems but in accordance to other test programs in Germany the same nozzles were used: Lechler TR 80–0067 (hollow cone) and Lechler IDK 120–015 (air induction flat fan). The intention was to calibrate all spraying systems to obtain application rates of 75 l ha^–1 for the IDK nozzle and 40 l ha^–1 for the TR nozzle, in combination with an appropriate application speed and the relevant working width. This was not exactly possible for several reasons: The flight speed had to be set to a constant value of 2 m s^–1 and it was not feasible to re-adjust the UAAS pump capacity for each test. The actual application rate values obtained for each UAAS and nozzle type resulting from nozzle flow rate and assumed working width (flight path spacing) are listed in Table 2. The “1 rotor” UAAS was equipped with a spray boom providing a large nozzle spacing. This resulted in a lower number of nozzles per working width compared to the other UAASs. So this UAAS produced a lower application rate as it was intended to keep the nozzle flow rate constant to achieve uniform droplet size characteristics.

The spray drift ground sediment was determined in field tests conducted following ISO 22866 (ISO, 2005). The test site was a flat field without vegetation apart from a few weeds (see Fig. 2). There were no significant obstructions around the field to influence the air flow.

Fig. 2. Test site with experi­mental arrangement for spray drift measure­ment. Bars holding de­posit plastic cards in front.

The treated area for all tests was approximately 100 m long and as wide as three working widths of the relevant UAAS. The flight direction was perpendicular to the nominal wind direction. So the edge of the treated field (EOF) was a line half of the working width downwind from the first flight line. The nominal distance from the nozzles to the ground was always 1.5 m. For practical reasons, the first swath of each treatment was sprayed under manual flight control; the next two swaths were treated flying automatically parallel to this line.

The spray deposit distribution on the treated field was determined using horizontal plastic cards clamped to 10 m long metal bars with a spacing of 0.5 m. Two of these bars, each of them holding 21 cards, were placed on the ground, approximately 3 m apart in the centre of the flight lines so that the first card was located on the EOF. The size of the cards was 5 cm × 10 cm or 5.4 cm × 8.5 cm, respectively.

The ground sediment of spray drift was determined at lines 3 m, 5 m, 10 m and 20 m downwind from EOF using 10 petri dishes of 145 mm diameter at each downwind distance. Three metal plates 1 m long and 0.6 m wide were placed on the ground within a distance of approximately 10 m, each supporting 3 or 4 petri dishes at each distance. Hence, the petri dishes were not evenly distributed along each line.

Grids of monofilament line collectors were used to sample airborne drift 1 m downwind from EOF. Three system racks held the lines (see Fig. 3). The racks were 2 m wide. Fold-back clips were mounted on the upwind side from 0.3 m to 1.8 m height from the ground with a spacing of 0.3 m to facilitate the handling of the line collectors.

Fig. 3. Frame holding the line collectors for airborne drift, metal plates sup­porting the petri dishes in the background

All tests were conducted at least 3 times. The number of actual replicates are listed in table 2. In addition, one blind test was made with the 1 rotor and one with the 6 rotor UAAS without spraying liquid to determine a potential contamination of the collectors from tracer deposited on the site from former tests.

Weather data, such as wind speed and direction as well as air temperature, were recorded with as ample rate of 1 s^–1. A three-dimensional ultrasonic anemometer (Gill Instruments Ltd., WindMaster 3D) was used, mounted 2 m above the ground.

The spray liquid was water with Brilliant Sulfoflavine (BSF, Chroma, lot no. 381928) as tracer dye with a concentration of 2 g l^–1. All samplers for drift and deposit were collected within less than 10 min after each test and stored in a box protected from light exposure in order to minimise degradation. The line collectors were coiled into 170 mm × 120 mm ziplock bags turned inside out before and back after. The plastic cards were put into 150 mm × 100 mm ziplock bags in a way to avoid any cross contamination.

The samplers were stored in a dark, cool room and analysed within 14 days after the tests. For analyses, the tracer was extracted from the samplers using de-ionised water. The petri dishes were filled with 40 ml and put on a shaker for 20 min. Frequency and amplitude were chosen in order to make sure that the inner walls of the dishes were washed completely. The bags with the plastic cards and line collector were filled with 160 ml or 40 ml, respectively, of de-ionised water and shaken thoroughly by hand.

Analyses of samplers were conducted using a fluori­meter Perkin Elmer LS 55 set at 405 nm wavelength for fluorescence excitation and 500 nm for emission detection. Samples of the spray liquid taken from the sprayers after each treatment were diluted in de-ionised water to a concentration of 250 μl l^–1 as the calibration liquid for calculation of the volume of spray liquid found on each sampler. The volume of the spray liquid V_c on each collector was calculated as:

(1)with

From these values, the deposit d_c on each deposit and drift collector was calculated as percentage of application rate:

(2)with

The sampler area of a line collector was considered as the product of its length l (1.8 m) and its diameter d (2 mm).

The spray deposit measured in the treated area were used to characterise the evenness of spray distribution. The coefficient of variation CV was calculated for each test as the average from the two lines of plastic cards considering two working widths from the first line of flight. Then the average CV was calculated from these values. Three of the tests were excluded from this evaluation as they showed very low deposit. This was probably caused by mistakes during fluorimetry as the drift values from the same tests were inconspicuous.

The median values for each distance from EOF were calculated from the drift ground deposit data from all repetitions of each set-up (UAAS-nozzle combination). The average relative drift values for all set-ups were correlated with the airborne drift index ADX calculated from the deposits on the line collectors as:

(3)

with the total relative airborne drift

, (4)

the relative height of the centre of drift plume

(5)

and

A statistical evaluation was conducted to prove the null hypothesis of no differences in spray drift from different UAAS designs and nozzles at a level of significance α = 5%. For this purpose, the median sediment curves were calculated for each test repetition. Afterwards these curves were condensed to one value each, the average median AM_r. Since for some of the repetitions the values for 3 m distance from field edge was missing due to the loss of petri dishes, the medians were averaged for 5 m to 20 m distance. An analysis of variance and a multiple comparison with a Tukey test was conducted using the SAS procedure GLM for the AM_r as well as for the ADX_r values for each repetition.

The median sediment curve over all repetitions is normally used for the ultimate evaluation of spray drift. As this should be done alternatively using the ADX para­meter, a correlation analyses was performed using the linear model:

(6)with

This analyses using the SAS procedure GLM gave the coefficients of the model as well as the confidence curves for the prediction of average median sediment AM from ADX for α = 5%.

Fig. 4. Average wind conditi­ons for each test

Since the deviations were small in most cases, all measurements were included in the evaluation. The average air conditions were quite uniform so that the air temperature ranged from 21°C to 28°C, the relative air humidity from 14% to 45%.

One of the most important criteria for chemical application is the evenness of spray deposit distribution. The results of the spray deposit measurements are shown in Fig. 5. It indicates the coefficient of variation CV as average and range from all repetitions. There are no significant differences between UAASs or nozzles and the variability between repetitions can be very large.

Fig. 5. Average (columns) of the coefficient of va­riance CV of spray de­posit distribution with range (indicators) of values for all repetiti­ons.

As Fig. 6 shows, the measured drift sediment downwind from the treated area was significantly influenced by the nozzle type. The TR hollow cone nozzle produces much more drift than the IDK air induction flat fan nozzle. There is obviously no consistent influence of the UAAS design on the drift sediment. Only for the DJI UAAS with 8 rotors the difference between the nozzles is lower. A problem occurred during the tests, as many of the petri dishes from 1 m and with the 1 rotor UAAS also from 3 m were blown off. This is why for these distances less or even none measuring values were available for evaluation.

Fig. 6. Median drift sediment for the different nozzles at each UAAS

The same principal results can be seen from Fig. 7 showing the differences in airborne drift profiles. Also with this method a higher drift potential from the TR nozzle was found with a lower difference to the IDK nozzle for 8 rotor UAAS from DJI. As the sediment curves show a decrease with distance, the shape of the airborne drift profiles show higher deposits on the lower lines. In general, there is no important influence on the shape or magnitude from UAAS design.

Fig. 7. Average vertical profile of airborne drift potenti­al for the different noz­zles at each UAAS

While Figures 5 and 6 mainly demonstrate the influence of the nozzle type on spray drift, Fig. 8 illustrates more clearly that the UAAS design does hardly affect the drift. The only deviation can be seen from the TTA 8 rotor UAAS for the lower part of the vertical airborne drift profile. But this is not supported by the equivalent sediment curve. Comparing the sediment curves to the standard drift curves used by German authorities for drift risk assessments for boom sprayers (dotted line), it can be shown that spray drift from all UAASs is higher than from boom sprayers, especially at lower distances from EOF.

Fig. 8. Spray drift from diffe­rent UAASs for the two nozzles (left: median se­diment curves compa­red to standard drift values for boom spray­ers, right: average verti­cal profiles)

As the evaluation of curves and profiles can be difficult, it was attempted to condense both into single parameters. So the average median drift sediment AM and the airborne drift index ADX were calculated (Fig. 9).

Fig. 9. Average (columns) median drift sediment AM (left) and airborne drift index ADX (right) from different UAASs for the two nozzles with standard deviation (indicators) of repetitions. Different letters mark si­gnificant differences (α = 5%).

On one hand, the analysis of variance showed whether there was significant differences in the average test results caused by the nozzle type, the UAAS design or the interaction (nozzle × UAAS). The main result of these analyses is that there was a significant influence of nozzle type on drift sediment as well as on drift potential. But the UAAS design as well as the interaction (nozzle × UAAS) had a significant effect on the ADX only (Table 3).

On the other hand, the analyses of variance provided the basis for a multiple comparison of the parameters using the Tukey test. As Fig. 9 shows, the standard deviation of the test repetitions, especially for the TR 80–0067 nozzle, was quite high. So in most cases, it was not possible to detect differences in drift sediment or drift potential between the UAAS designs. Only the ADX characterising the drift potential from the TR 80–0067 nozzle for the 1 rotor UAAS was proved different from the 6 rotor and the DJI 8 rotor design. But in general, no distinct influence of the UAAS design is evident.

As the test site space was limited and the arrangement of all collectors was time consuming, it was not possible to change location for each test. Therefore it was not possible to avoid the contamination of the test site due to repeated application of test liquid. Blind tests should show the influence of this contamination on the test results. The results are shown in Fig. 10 in comparison to the spray drift measured for the same UAAS equipped with the IDK 120–015 nozzle providing a relatively low drift potential.

Fig. 10. Collector contamination after two blind tests (left: median sediment curves, right: average vertical profi­les) compared to spray drift measured for the same UAAS with IDK 120–015

During the blind test a contamination of the petri dish collectors with tracer dye could be found although no liquid was sprayed. This “background noise” could reach the order of magnitude of spray drift from the IDK nozzle. No contamination was found at the same test on the line collectors for airborne drift potential. In contrast to petri dishes these samplers obviously collect liquid droplets only.

Despite of these findings, an attempt was made to analyse the correlation between the results of sediment and airborne drift measurements as in general the same main trends were found with both methods (Fig. 11). The result of the regression analysis clearly showed a significant correlation between the drift potential and the drift sediment. The coefficients of the model according to equation (6) with their 95% confidence intervals are a = 0.0443 (0.0342 … 0.0523) and b = 0.2375 (0.0404 … 0.4346). The coefficient of determination obtained of R² = 0.958 suggests that more than 95% of the variation of the average median drift sediment in the test is caused by the drift potential index ADX.

Fig. 11. Correlation of the air­borne drift potential with drift sediment (R² = 0.958). The dot­ted lines mark the lower and upper bounds of the 95% confidence inter­vals for the mean pre­dicted values.

The application quality observed in terms of the coefficient of variation of the spray liquid cross distribution on the treated area was independent from the UAAS spraying system configuration and nozzle type in a range of (40 ± 10)% and was in good agreement to former studies and in the range of distribution quality from knapsack sprayers (Qin et al., 2018; Wang C et al., 2018). These results were obtained although the optimisation of spray distribution was not the aim of this study. The flight patterns were chosen mainly to obtain the target application rates independent from the design of the spraying systems. Obviously the air flow or other factors play a more significant role on spray distribution than the design of the spraying system.

The measurement of spray drift ground sediment using Petri dishes was difficult, although they had proven advantageous for different application techniques in many studies before. In general, they are good to handle and to protect from cross contamination as they are protected by a lid during transport. In case of drift measurements from UAAS application, considerable background contamination can occur due to movements of tracer dye particles generated by the downwash airflow from former applications on the same field. This would require moving to a blank area for every test. But this kind of contamination could also emerge from former flight lines during one test. Another problem was that Petri dishes closer than 5 m from the EOF must be fixed to a proper support as otherwise they easily can be blown off.

Background contamination was no problem for the line collectors used for measurement of spray drift potential. Experiences from JKI wind tunnel tests, where these collectors are also used, show that the lines do not collect dry particles.

Drift potential measured as airborne drift close to the edge of the field does very well correlate with ground sediment but regression coefficients are valid only for the tests of this study. Results suggest though that it should be possible to use this correlation to convert the ADX values into average median sediments in the future. This could be very helpful for the relative evaluation of application techniques but a lot more tests are necessary to establish a reliable conversion model valid for a range of conditions. This is also interesting for drift tests with other kinds of application techniques.

The project was supported by grants of the German Research Foundation (DFG, grant number HE 8061/1–1) under the Joint Sino-German Research program, of the Project 31761133019 supported by the National Natural Science Foundation of China and of the China Agriculture Research System CARS-28–20.