Accurate assessment of grass nitrogen status based on multispectral data from two optical sensors and the critical nitrogen dilution curve

TY - ABST

T1 - Accurate assessment of grass nitrogen status based on multispectral data from two optical sensors and the critical nitrogen dilution curve

AU - Zhang, Shaohui

AU - Lærke, Poul Erik

AU - Andersen, Mathias Neumann

AU - Manevski, Kiril

PY - 2022/11

Y1 - 2022/11

N2 - The critical nitrogen dilution curve (CNDC) is efficient for diagnosing plant nitrogen (N) deficiency. For perennial plants, including grasses (family Poaceae), deriving CNDC is not straightforward due to in-season modulations by cuts and fertilization. Gislum et al. (2009) developed CNDC for grass species for seed production in Denmark and suggested improvements since it was not possible to calculate a % critical N concentration at biomass lower than 2 Mg ha-1. Optical sensors deployed on the field can partly answer the question of how much and when to fertilize, e.g., the Yara’s N tester (Yara International ASA, Oslo, Norway) that allows assessing plant N requirement and real-time variable-rate spread of fertilizer. Moreover, multispectral data from unmanned aerial vehicle (UAV) with their ultra-high spatial and temporal resolution can also be used to estimate crop N status (Peng et al., 2021), though high accuracy is difficult to achieve due to the dynamics and the transformation of N. Accurate N management requires establishment of a solid ‘link’ between the remote sensing data and the crop N status. Parametric regression involving models of linear or nonlinear nature has often been used for describing this link (Peng et al., 2021). Peng et al. (2021) investigated how well potato N status can be described by RS data obtained from ground, air- and spaceborne sensors using parametric and non-parametric, i.e., machine learning regression. Few previous studies have used RS to calculate the N status of grasses and predict their N requirements. The objective of this study is to estimate precisely the amount of N fertilizer needed for optimal plant growth according to N requirement and the CNDC by two optical sensors (Yara N sensor and UAV-mounted) based on machine learning method (random forest and alike). Two-year field experiment has been initiated with grass (Lolium perenne) in 2022 and 2023 (established in September 2021) on a sandy loam soil in Denmark, with four nitrogen rates (N0: 0 kg ha-1, N1: 75 kg ha-1, N2: 300 kg ha-1 and N3: 450 kg ha-1). We present initial results of the canopy multispectral reflectance and CNDC developed for the grass and provide novel insight for improving N diagnosis and management of grass in Denmark and Europe.

AB - The critical nitrogen dilution curve (CNDC) is efficient for diagnosing plant nitrogen (N) deficiency. For perennial plants, including grasses (family Poaceae), deriving CNDC is not straightforward due to in-season modulations by cuts and fertilization. Gislum et al. (2009) developed CNDC for grass species for seed production in Denmark and suggested improvements since it was not possible to calculate a % critical N concentration at biomass lower than 2 Mg ha-1. Optical sensors deployed on the field can partly answer the question of how much and when to fertilize, e.g., the Yara’s N tester (Yara International ASA, Oslo, Norway) that allows assessing plant N requirement and real-time variable-rate spread of fertilizer. Moreover, multispectral data from unmanned aerial vehicle (UAV) with their ultra-high spatial and temporal resolution can also be used to estimate crop N status (Peng et al., 2021), though high accuracy is difficult to achieve due to the dynamics and the transformation of N. Accurate N management requires establishment of a solid ‘link’ between the remote sensing data and the crop N status. Parametric regression involving models of linear or nonlinear nature has often been used for describing this link (Peng et al., 2021). Peng et al. (2021) investigated how well potato N status can be described by RS data obtained from ground, air- and spaceborne sensors using parametric and non-parametric, i.e., machine learning regression. Few previous studies have used RS to calculate the N status of grasses and predict their N requirements. The objective of this study is to estimate precisely the amount of N fertilizer needed for optimal plant growth according to N requirement and the CNDC by two optical sensors (Yara N sensor and UAV-mounted) based on machine learning method (random forest and alike). Two-year field experiment has been initiated with grass (Lolium perenne) in 2022 and 2023 (established in September 2021) on a sandy loam soil in Denmark, with four nitrogen rates (N0: 0 kg ha-1, N1: 75 kg ha-1, N2: 300 kg ha-1 and N3: 450 kg ha-1). We present initial results of the canopy multispectral reflectance and CNDC developed for the grass and provide novel insight for improving N diagnosis and management of grass in Denmark and Europe.

M3 - Conference abstract in proceedings

SP - 201

BT - Proceedings of the XXI International Nitrogen Workshop. Halving nitrogen waste by 2030

A2 - Lassaletta, L.

A2 - Sanz-Cobeña, A.

A2 - Pinsard, C.

A2 - Garde, S.

PB - CEPADE-Universidad Politecnica de Madrid

T2 - The XXI International Nitrogen Workshop. Halving nitrogen waste by 2030.

Y2 - 24 October 2022 through 28 October 2022

ER -