• Aucun résultat trouvé

Map-based variable-rate nitrogen application based on proximal and remote sensing techniques

N/A
N/A
Protected

Academic year: 2022

Partager "Map-based variable-rate nitrogen application based on proximal and remote sensing techniques"

Copied!
2
0
0

Texte intégral

(1)

MAP-BASED VARIABLE-RATE NITROGEN APPLICATION BASED ON PROXIMAL AND REMOTE SENSING TECHNIQUES

Angela Guerrero, Stefaan De Neve, Abdul M. Mouazen*

Department of Environment, Faculty of Bioscience Engineering, Ghent University, Coupure Links 653, 9000- Gent, Belgium

*E-mail of corresponding author: Abdul.Mouazen@UGent.be

This study reports the implementation results of map-based variable rate nitrogen fertilization (VRNF), using fusion of multi-layer soil data collected with an online sensor and normalized difference vegetation index (NDVI) from remote sensing in comparison with traditional uniform rate nitrogen fertilisation (URNF). Online soil measurement was carried out in two commercial fields located in Flanders in Belgium using a multisensor platform consisting of a fiber type visible and near infrared spectrometer. Partial least square regression (PLSR) models were developed to predict pH, organic carbon, moisture content, calcium, sodium, magnesium, cation exchange capacity, phosphorous and potassium. The predicted soil properties were fused with NDVI and yield data to delineate management zone maps with different fertility classes. Three nitrogen (N) recommendations treatments were implemented and compared including URNF, VRNF1, aimed to apply 50% more of recommended rate (RR) of fertilizer to the high fertile zones, RR to the medium fertile zones and 50% less to the least fertile zones and VRNF2 aimed to apply the opposite VRNF1 scenario. A cost-benefit analysis was performed using the cost of fertilizer applied as input and the price of barley and wheat in Belgium in 2019 as output. Results comparing the treatments present a promising potential for VRNF. It was observed that VRNF2 was more profitable compared to both VRNF1 (11.55 - 19.52 EUR per ha) and URNF (16.40 - 148.78 EUR per ha). VRNF2 led not only to economic profit but also to reduce the negative environmental impact, by reducing the amount of N fertilizer used, compared to both the URNF and VRNF1.

This result confirms that VRNF2 is the ideal approach for site-specific N fertilization in barley and wheat, as it maintains or increase crop yield by up to 10.4% and reduces N fertilizer application by up to 19.44%, compared to URNF.

Key words: Variable rate nitrogen fertilization, proximal soil sensing, data fusion, profitability, management zone.

(2)

Cost-benefit and nitrogen (N) fertilizer use analyses in two fields (e.g., Kouter and Grootland). Data is shown for uniform rate nitrogen fertilisation (URNF), variable rate1 high fertility (VRNF1-H), variable rate1 medium high fertility (VRNF1-MH), variable rate1 medium fertility (VRNF1-M), variable rate1 medium low fertility (VRNF1-ML), variable rate1 low fertility (VRNF1-L), variable rate2 high fertility (VRNF2-H), variable rate2 medium high fertility (VRNF2-MH), variable rate2 medium fertility (VRNF2-M), variable rate2 medium low fertility (VRNF2-ML) and variable rate2 low fertility (VRNF2-L) plots.

KOUTER FIELD GROOTLAND FIELD

Treatment Area (ha)

Yield (t ha-1)

Cost of fertilizer (EUR ha-1)

Yield (EUR ha-1)

Profit (EUR ha-1)

Comparison between treatment (EUR ha-1)

N units ha-1

Comparison per treatment (N

units)

Treatment Area (ha)

Yield (t ha-1)

Cost of fertilizer (EUR ha-1)

Yield (EUR ha-1)

Profit (EUR ha-1)

Comparison between treatment (EUR ha-1)

N units ha-1 Comparison per treatment (N

units)

UR 4.98 8.04 81.18 1222.08 1140.91 ---- 125 ---- UR 6.7 10.46 87.35 1673.6 1586.25 ---- 138 ----

VRNF1_H 2.37 8.84 126.44 180 VRNF1_H 0.65 9.8 122.3 193.2

VRNF1_M 0.47 10.51 81.18 125 VRNF1_MH 1.58 10.59 104.82 165.6

VRNF1_L 1.45 7.55 41.04 UR-VRNF1 80 UR-VRNF1 VRNF1_ML 1.79 10.64 69.88 110.4

Total VRNF1

4.29 8.97 92.66 1362.83 1270.17 129.27 140.22 -15.22 VRNF1_L 0.72 10.9 52.41 UR-VRNF1 82.8 UR-VRNF1

VRNF2_H 2.11 8.79 41.04 80 Total VRNF1 4.74 10.48 86.02 1677.12 1591.1 4.85 135.9 2.1

VRNF2_M 1.43 8.99 81.18 125 VRNF2_H 2.12 10.14 52.41 82.8

VRNF2_L 1.07 9.13 126.44 UR - VRNF2 180 UR - VRNF2 VRNF2_MH 2.13 10.65 69.88 110.4

Total VRNF2

4.61 8.97 73.35 1363.03 1289.68 148.78 117.21 7.79 VRNF2_ML 1.14 10.58 104.82 165.6

TOTAL AREA

13.89 VRNF2_L* 0 0 0 UR - VRNF2 0 UR - VRNF2

Total VRNF2 5.39 10.46 70.37 1673.01 1602.64 16.4 111.17 26.83

TOTAL AREA 16.83

* The subplot of VRNF2_L does not present any data because the subplot was eliminated from the application map due to the buffer in the field.

Références

Documents relatifs

Regarding the variable selection method, the performances of the neural network trained with 90 variables shows a marginal degradation of the performance as compared to the

On the basis of the results obtained when applying this and the previously proposed methods to experimental data from a typical industrial robot, it can be concluded that the

The first part will be used to calibrate the model according to the two techniques: ordinary least squares (OLS) and Kalman filter (KF). The second part will be

Vincent Vandewalle, Thierry Mottet, Matthieu Marbac. Model-based variable clustering. CMStatis- tics/ERCIM 2017 - 10th International Conference of the ERCIM WG on Computational

Models in competition are composed of the relevant clustering variables S, the subset R of S required to explain the irrelevant variables according to a linear regression, in

Keywords: Variable annuities, indifference pricing, stochastic control, utility maximization, backward stochastic differential equation.. MSC2000 subject classification: 60H99,

Examples include magnetic resonance imaging (MRI), radio-interferometry, mobile-robot sampling, ... In this paper, we propose to generate continuous sampling trajectories by drawing

On utilise beaucoup Geogebra pour l’étude de la statistique à deux variables, alors que cette application nous permet de le faire pour une variable.. Puisque l’on utilise la