381

March 18, 2018 | Author: rrdpereira | Category: Wheat, Oat, Biomass, Agriculture, Foods


Comments



Description

ESTIMATING CROP BIOMASS AND NITROGEN UPTAKE USINGCROPSPECTM, A NEWLY DEVELOPED ACTIVE CROP-CANOPY REFLECTANCE SENSOR S. Reusch, J. Jasper, and A. Link Research Centre Hanninghof Yara International ASA Duelmen, Germany ABSTRACT In-season variable rate nitrogen fertilizer application needs efficient determination of the nitrogen nutrition status of crops with high spatial and temporal resolution. A suitable approach to get this information fast and at low cost is proximal sensing of the light that is reflected from the crop canopy. CropSpecTM is an active vehicle mounted crop canopy sensor. Using pulsed laser diodes as light source, the sensor is designed to look at the crop at an oblique view from a height of up to 4 meters in order to enable a large footprint for accurate measurements in various crops. The objective of the research presented in this paper was to assess the agronomic performance of this newly developed sensor, i.e. the accuracy of biomass and N uptake estimations that can be achieved when using the instrument in different arable crops. Field experiments with increasing nitrogen rates have been conducted during the 2009 growing season in winter wheat, winter barley, spring barley, oats, oilseed rape and corn. Crop canopy reflectance was measured in frequent intervals throughout the main growing season (2 to 9 measurement dates per crop). Destructive plant biomass samples were taken at the measurement dates in order to determine the actual biomass growth, N concentration and N uptake of the crop. Sensor readings were related to these crop parameters, crop specific calibration functions were derived, and the capacity to predict the actual crop biomass and N uptake was investigated. The impact of different varieties and seed densities on the prediction quality was investigated in the winter barley and corn trials. Those were set up as multi-factorial experiments, with three varieties and two seed densities in winter barley and three seed densities in corn. Keywords: CropSpec, canopy reflectance, N uptake 2004). The objective of the research presented in this paper was to assess the agronomic performance of this device. the accuracy of biomass and N uptake estimations that can be achieved when using the instrument in different arable crops. 2008.. Consequently.e. Reusch. which may lead to more reliable measurements. Holland et al. 2004). Furthermore. 2005). . Previous research gave evidence that information on the in-field variability of nitrogen uptake is most valuable for the derivation of site-specific N fertilizer recommendations (Link et al. 2004. Vegetation indices combining reflectance values at wavelengths in the infrared and the red edge have proven to be superior to those that combine an infrared wavelength with wavelengths taken from the visible range (Mutanga and Skidmore. especially in early growth stages (Mistele et al. i.. 2004).g.INTRODUCTION Proximal sensing of the light reflectance properties of crop canopies is a feasible and cost efficient approach to determine the nitrogen nutrition status of crops at a spatial and temporal resolution that is required for variable rate nitrogen fertilization. Sensors that are scanning the crop at an oblique view ‘see’ more crop biomass and less soil surface. measurements at oblique view result in a larger footprint that improves the stability of the signal in crops with wide row-to-row distances. Solie et al. 2009).. there are differences in terms of the technical set-up that can have an impact on their performance and suitability for specific agronomic applications.. Jasper et al.. The footprint size can be enlarged further if the system allows for a long sensor-to-crop distance. Schmidhalter et al. 2001. 2003. 2005.. Investigations on optimal wavebands and spectral indices with utmost sensitivity to the N status of the crop have shown that the appropriate choice of a vegetation index is essential for good N uptake prediction (Reusch. various crop canopy sensors have been developed in recent years (e.. Sensors differ in the light sources that are used (with implications for the choice of electro-magnetic wavebands that are measured and vegetation indices that are calculated) and in their viewing geometry and sensor-to-crop distance (affecting the size of the footprint that is scanned and the amount of crop biomass that is seen in the field of view of the system).. Lammel et al.. 2002. 2003. Mistele et al. Although the basic measurement principles are similar for all canopy reflectance sensors. With the CropSpecTM a new crop canopy sensor has been developed that fulfils the technical requirements for site-specific measurements of the nitrogen nutrition status of crops as described above. Mueller et al. Figure 1. For oilseed rape. winter wheat. USA. developed by Topcon Positioning Systems. Trial layout Field trials with winter oilseed rape.1 m when mounted at a height of 3 m (Figure 1). Crop Oilseed rape Winter wheat Spring barley Oats Plot size m2 30 x 15 25 x 15 20 x 15 20 x 15 No. and oats one-factorial trials without replications were conducted. Table 1. 51°50ƍ26ƍƍ1ƒ5ƍ28ƍƍ E). spring barley. Due to its oblique view with a viewing angle of 50° the sensor is scanning the crop with a footprint width of about 3. The system comprises a transmitter with two spectral channels (735 nm and 808 nm) using pulsed laser diodes (1 W. CA. Trial setup of the one-factorial trials. winter wheat.. Duelmen. of N rates 10 10 10 10 Range of N rates kg N/ha 0-270 0-315 0-180 0-180 In winter barley a three-factorial trial was set up as a split-plot design with three contrasting barley varieties as main plots. The sensor can be integrated into or mounted on top of agricultural vehicles at a mounting height between 2 and 4 meters. winter barley and corn were conducted on sandy soils in the northwest of Germany in 2009 (Research Centre Hanninghof. two seed densities as first order sub-plots. Livermore.MATERIAL AND METHODS CropSpecTM CropSpecTM is an active vehicle mounted crop canopy sensor. 10kHz) as modulated light sources and one receiver unit. and 10 different N rates (0 . spring barley.270 kg N/ha) as second order sub-plots . CropSpecTM – viewing geometry and footprint. Inc. oats. with 10 plots of increasing N rates (at equal increments) located along one or two tramlines of the field (Table 1). The barley varieties used in the trials were chosen from a list of varieties commonly grown in farm practice.. Nitrogen was applied in three dressings at growth stages BBCH 20. Contrasting barley varieties. SD2) and 10 N rates (N0-N9). BBCH 31 and BBCH 37 (Lancashire et al. 2 seed densities (SD1. Trial layout of the winter barley trial. Highlight Lomerit Naomie Dark green leaves Tall plants Low crop density Very light green leaves Medium plant height Medium crop density Medium green leaves Medium plant height Medium crop density Figure 3. . The strips were fertilized with increasing N rates from 0 to 160 kg N/ha (9 different N rates) applied in one dressing at planting (Figure 4). 1991).(Figure 2). resulting in contrasting canopy architectures that may cause modified reflectance properties. Corn was planted on two trial sites in strip trials without replication at three seed densities (7. They are characterized by distinct differences in phenological properties like leaf colour. Highlight Tramline SD 2 Lomerit SD 1 SD 2 Naomie SD 1 SD 2 SD 1 N5 N6 N7 N8 N9 N5 N6 N7 N8 N9 N5 N6 N7 N8 N9 N5 N6 N7 N8 N9 N5 N6 N7 N8 N9 N5 N6 N7 N8 N9 29 28 27 26 25 19 18 17 16 15 29 28 27 26 25 13 14 20 21 22 23 24 10 11 12 13 14 20 21 22 23 24 14 15 12 13 16 11 15 17 10 12 18 24 16 19 23 17 25 22 11 26 21 10 27 20 18 28 19 29 7. 3 varieties. 10 and 13 seeds/m2). stem length and leaf orientation (Figure 3).5 m N4 N3 N2 N1 N0 N4 N3 N2 N1 N0 N4 N3 N2 N1 N0 N4 N3 N2 N1 N0 N4 N3 N2 N1 N0 N4 N3 N2 N1 N0 3m SD 1: 150 m-2 SD 2: 350 m-2 Figure 2. Trial layout of the corn trials. Measurement dates and growth stages (GS). Table 2. Table 2). including all growth stages that are relevant for nitrogen fertilizer application (2 to 9 measurement dates per crop. Reflectance measurements and plant sampling Crop canopy reflectance was measured in frequent intervals throughout the main growing season of the crops.Tramline N0 SD1 SD2 SD3 N1 N2 N3 N4 N5 N6 N7 N8 10 11 12 13 14 15 16 17 18 20 21 22 23 24 25 26 27 28 30 31 32 33 34 35 36 37 38 15 m ~25 m Figure 4. Oilseed rape Winter wheat Spring barley Oats Winter barley Corn Date GS Date GS Date GS Date GS Date GS Date GS mm/dd BBCH mm/dd BBCH mm/dd BBCH mm/dd BBCH mm/dd BBCH mm/dd BBCH 04/01 30 03/18 21/25 04/28 15 04/28 15 04/02 28 06/02 18 04/09 50 04/09 29 05/12 31 05/08 31 04/15 30 06/12 19 04/20 31 05/22 37 05/22 39 04/24 33 06/18 33 04/30 32/33 05/29 49 05/29 49 04/30 37 06/30 39 05/12 35 06/05 57 06/05 60 05/08 44 07/10 51 05/19 39 05/18 55 05/28 55 05/28 69 06/08 59/61 06/17 69 Sensor readings were calculated as simple ratios of the reflectance values obtained from the two spectral channels: R (1) Sensor reading 808  1 . 3 seed densities (SD1-SD3) and 9 N rates (N0-N8) . The samples were weighed and . x 100 R735 with R735 and R808 being the relative reflectance at 735 and 808 nm.96 m2) in cereals and two subplots in oilseed rape (1m2 each) and corn (3m2 each). destructive plant samples of the above ground biomass were taken by cutting one subplot (0. At the measurement dates. respectively. Therefore. oilseed rape is reacting on differences in N supply with an immediate variation in biomass growth. For oilseed rape the sensor values have been related to the amount of aboveground biomass (t FM/ha).dried. determined by destructive plant sampling.. and the capacity of CropSpec measurements to predict the actual crop biomass (oilseed rape) and N uptake (cereals and corn) was investigated. crop. Sensor readings of N treatments at different growth stages.88 and an RMSE of 2. fresh matter production is seen as the most suitable indicator for the nutritional status of oilseed rape at early growth stages. the sensor based estimate of the crop biomass. 2005). Figure 6 shows the relationship between sensor based fresh matter predictions and fresh matter biomass determined by destructive plant sampling of the N treatments at BBCH 30 and BBCH 50. Increasing sensor readings from date to date at a given N rate and more distinct differences between N rates at the later growth stage reflect the growth and N uptake pattern of the crop. and the N uptake calculated. is fairly accurate. especially leaf size.and growth stage-specific calibration functions were derived. aboveground dry matter production was determined. rape. RESULTS AND DISCUSSION Oilseed rape Figure 5 shows the sensor readings obtained in the nitrogen treatments of the oilseed rape trial at growth stages BBCH 30 and BBCH 50. Sensor readings of the respective plots were related to these crop parameters. its total N content analyzed (Kjeldahl method). where the most sensitive crop parameter for estimating its N status is the actual N uptake (Link et al. 45 40 Sensor reading 35 30 GS 30 25 20 GS 50 15 10 5 0 N0 N1 N2 N3 N4 N5 N6 N7 N8 N9 Treatment Figure 5. . with an r2 value of 0. Considering the unavoidable trial error when taking a subsample from a heterogeneous oilseed rape crop.53 t/ha. In contrast to cereals. Winter wheat The plots of the winter wheat trial have been measured nine times during the vegetation period. From growth stage BBCH 29 (end of tillering) onwards. the sensor readings showed a clear N response. rape. RMSE = 9.88 25 GS 30 RMSE = 2.40 Fresh matter [t/ha] 35 30 y = 1. Sensor values for a given N rate increased until BBCH 39.53 t/ha 20 GS 50 15 GS 30-50 10 5 0 10 0 30 20 40 Fresh matter (predicted) [t/ha] Figure 6. .5 kg N/ha. Fresh matter prediction from sensor readings.02x r² = 0. CropSpec measurements enabled the prediction of the current nitrogen uptake of winter wheat with high accuracy (r2 = 0. from tillering (BBCH 21/25) to end of flowering (BBCH 69). wheat. 100 Sensor reading 90 GS 21/25 80 GS 29 70 GS 31 60 GS 32/33 GS 35 50 GS 39 40 GS 55 30 GS 59/61 20 GS 69 10 0 N0 N1 N2 N3 N4 N5 N6 N7 N8 N9 Treatment Figure 7. Sensor readings of N treatments at different growth stages. Following ear emergence (BBCH 55) the sensor values declined due to the beginning senescence of older leaves (Figure 7).96. Spring barley and oats In spring barley and spring sown oats most of the nitrogen fertilizer is commonly applied at.Figure 8). Only at ear emergence (BBCH 55) there seems to be a tendency to overestimate N uptake. If those late growth stages are excluded from the calibration. Topdressed N will rarely be applied later than at the beginning of stem elongation (BBCH 31/32).96 RMSE = 9. Whereas for spring barley good sensor based N uptake predictions have been achieved until end of ear emergence (r2 = 0.7 kg N/ha. RMSE = 18.1 kg N/ha). Nevertheless. 200 GS 21/25 GS 29 N-Uptake [kg/ha] 160 GS 31 GS 32/33 120 GS 35 y = 1x r² = 0. wheat. the feasibility of CropSpec based N uptake estimation has been tested also for later growth stages until flowering.93. . N uptake prediction from sensor readings. Figure 9). RMSE = 8. the prediction quality in oats is almost as good as in spring barley (r2 = 0. panicle emergence in oats caused a significant underestimation of the N uptake (Figure 10).5 kg N/ha 80 GS 39 GS 55 GS 69 40 GS 21-69 0 0 40 80 120 160 200 N-Uptake (predicted) [kg/ha] Figure 8. planting.92. or immediately after. 120 140 GS 49 GS 60 GS 15-49 .00x r² = 0.140 N-Uptake [kg/ha] 120 GS 15 100 GS 31 80 GS 37 y = 1.92 RMSE = 18. N uptake prediction from sensor readings.00x r2 = 0. oats. spring barley.1 kg N/ha 60 40 20 0 0 20 40 60 80 100 N-Uptake (predicted) [kg/ha] Figure 10.93 RMSE = 8. 140 N-Uptake [kg/ha] 120 100 GS 15 GS 31 80 GS 39 y = 1.7 kg N/ha 60 GS 49 GS 57 40 GS 15-57 Linear (GS 15-57) 20 0 0 20 40 60 80 100 120 140 N-Uptake (predicted) [kg/ha] Figure 9. N uptake prediction from sensor readings. Since the plant samples were taken on small sub-plots whereas the non-destructive sensor measurements covered the whole plot this caused some noise in the dataset.86 RMSE = 13. .00x r² = 0. 2 seed densities. winter barley. The dataset includes measurements for the three varieties tested.3 kg N/ha 60 40 GS 55 GS 69 GS 29-69 20 0 0 20 40 60 80 100 120 140 160 180 N-Uptake (predicted) [kg/ha] Figure 11. both seeding rates and all dates. In fact. the barley trial was characterized by uneven crop stands in the plots caused by a severe drought in spring in combination with an uneven slurry application in the previous year.86 and a root mean square error (RMSE) of 13. was found to be slightly lower. N uptake prediction from sensor readings.3 kg N/ha. Although the data gave evidence of a small impact of the crop variety. Compared to the results obtained in winter wheat the accuracy of the sensor based N uptake prediction. with the N uptake of the light green cultivar Lomerit being slightly underestimated (Figure 12). 3 varieties. 180 160 GS 28 N-Uptake [kg/ha] 140 GS 30 120 GS 33 100 GS 37 80 GS 44 y = 1. as compared to the winter wheat trial. As there is also no significant influence of the two seed densities on the N uptake prediction (Figure 13). the lower accuracy of the N uptake prediction in winter barley is most likely due to a bigger trial error. this cannot sufficiently explain the weaker correlation between estimated and real N uptake.Winter barley The predictive power of the derivation of N uptake estimations from sensor readings in winter barley is shown in Figure 11. with a coefficient of determination (r2) of 0. winter barley. winter barley.89 100 Highlight Lomerit 80 Naomie 60 40 20 0 0 20 40 60 80 100 120 140 160 180 N-Uptake (predicted) [kg/ha] Figure 12.82 160 y = 0.96x r² = 0.92 N-Uptake [kg/ha] 140 120 y = 0. N uptake prediction from sensor readings as affected by seed density.85 120 100 SD 1 SD 2 80 60 40 20 0 0 20 40 60 80 100 120 140 160 180 N-Uptake (predicted) [kg/ha] Figure 13. 180 y = 1.78 160 N-Uptake [kg/ha] 140 y = 0.10x r² = 0.02x r² = 0. N uptake prediction from sensor readings as affected by variety. .98x r² = 0.180 y = 1. 2 seed densities.95x r² = 0. 3 varieties. As those later growth stages are of no relevance with regard to fertilizer topdressings. and decreased considerably at the end of stem elongation and the beginning of tassel emergence (BBCH 51).96 RMSE = 3. corn. until BBCH 37 (seven nodes detectable).8 kg N/ha 100 80 Site 1 GS 39 Site 1 GS 51 Site 2 GS 17 60 Site 2 GS 32 Site 2 GS 37 40 GS 17-37 20 0 0 20 40 60 80 100 120 140 160 N-Uptake (predicted) [kg/ha] Figure 14. As this is crucial information for site-specific crop management decisions. when used in conjunction with scientifically sound decision rules. the new instrument is. the calculation of the correlation does only include growth stages from BBCH 17 (seven leaves unfolded) to BBCH 37.e.0x r² = 0. a feasible tool for variable rate application of nitrogen fertilizer and other farm inputs. CONCLUSION The results presented in this paper show that the newly developed CropSpecTM crop canopy reflectance sensor is capable to provide accurate nondestructive measurements of the above-ground biomass and N uptake of various crops with high spatial resolution. i. as shown in Figure 14. . 3 seed densities. covering N uptake values in a range between 0 and 80 kg N/ha. N uptake prediction from sensor readings.Corn The accuracy of CropSpec based N uptake predictions in corn was high for early growth stages. 160 140 Site 1 GS 18 N-Uptake [kg/ha] 120 Site 1 GS 19 Site 1 GS 33 y = 1. 2 sites. and Schmidhalter. K. P. Uppsala. van den Boom. The Netherlands. (eds. In: Van Henten. and Horst. Annals of Applied Biology 119. J. 3999-4014. July 25-28. F. 2009. Wageningen Academic Publishers. Germany.). Langelüddecke. R.. Gutser. Proceedings of the 7th European Conference on Precision Agriculture. et al. 2005. USA. 2004. Precision agriculture ’09. Reusch. E. G. S. Jasper. J. Minneapolis. K. Suitability of different crop parameters for the determination of site-specific nitrogen fertilizer demand. A.. Stauss. H. P. Wollring. International Journal of Remote Sensing 25(19). and Reusch. June 9-12.. A. 172-182.. Link. USA. Weber. In: Stafford.). The Netherlands. Dordrecht. P. Plant canopy sensor with modulated polychromatic light source. D. Validation of field-scaled spectral measurements of the nitrogen status in winter wheat. July 25-28. In: Stafford. Jasper. (ed. 2008. 2004. Bleiholder. Schepers. In: Proceedings of the 7th International Conference on Precision Agriculture. Biosystems Engineering 101. and Witzenberger.). R. MN. J.. 2004. A uniform decimal code for growth stages of crops and weeds. Wageningen Academic Publishers. O. Mueller. Meyer-Schatz. Sweden. L. C.. and Link. A. Precision Agriculture. Wageningen.. J. Boettcher. U. Mutanga. 561-601. and Lockhorst. Wageningen. 297-302. 2001. variety and growth stage. In: Proceedings of the 7th International Conference on Precision Agriculture. Shanahan.). W. Kluwer Academic Publishers. Active sensing of the N status of wheat using optimized wavelength combination – impact of seed rate. K. Analysis of vegetation indices derived from hyperspectral reflection measurements for estimating crop canopy parameters of oilseed rape (Brassica napus L. F.. Tractor based remote sensing for variable nitrogen fertilizer application. Wageningen. H. and Reusch. H. S. S. E. B. Proceedings of the 5th European Conference on Precision Agriculture. July 5-8. Minneapolis. 1991. Optimisation of oblique-view remote measurement of crop Nuptake under changing irradiance conditions.. Precision Agriculture ’05. J. J.REFERENCES Holland. and Werner. 2003. J.. P.. V. June 15-19. (eds. The Netherlands. Mistele. .J. (eds. T. Proceedings of the 4th European Conference on Precision Agriculture. Reusch. Lammel. Lancashire. J.. J. 573-578.). E.. U. and Kage. P. Plant nutrition – Food security and sustainability of agro-ecosystems. Berlin. Narrow band vegetation indices overcome the saturation problem in biomass estimation.. S. 694-695. and Skidmore. S. 23-30. MN. P. Goense. A. In: Horst. The Netherlands. D. Proceedings of the 6th International Conference on Precision Agriculture. J. Raun. The Netherlands. July 14-17. June 15-19. Berlin. S. V. R.. (eds. Proceedings of the 5th European Conference on Precision Agriculture. 261-266.. Minneapolis. J. P. C. Schmidhalter. Wageningen Academic Publishers. Jungert. Stone.. The Netherlands.. R.. 2005. Wageningen Academic Publishers. Sweden. N. K. USA. (ed.. S. et al. L. B. D. E. Manhart.. Wageningen. Freeman. Bredemeier. R. W. Needham. E. In: Stafford. 2005.Reusch. Precision Agriculture ’05. 615619. C. Mistele. and Washmon.. R. U. B. Field-scale validation of a tractor based multispectral crop scanner to determine biomass and nitrogen uptake of winter wheat. S. J. Germany. Optimum waveband selection for determining the nitrogen uptake in winter wheat by active remote sensing. and Werner. P. . 2002. In: Stafford. In: Robert. Reed. G. and Gerl. June 9-12. G. (eds. P. Mullen. Johnson.). 2003.. Precision Agriculture.. MN. V. Wageningen. Proceedings of the 4th European Conference on Precision Agriculture. Solie. Gutser. M. Real-time sensing and N fertilization with a field scale GreenSeekerTM applicator...).). Uppsala.
Copyright © 2024 DOKUMEN.SITE Inc.