New research published in Chemometrics Journal presented a new deep learning framework designed to dramatically improve the accuracy and interpretability of nitrogen (N) and chlorophyll (Chl) estimation in corn canopies. This study, led by researcher Li Tian of Heilongjiang Bayi Agricultural University in China, introduces a hybrid neural network architecture that has the potential to advance precision agriculture by providing rapid and non-destructive biochemical assessments at field scale (1). The results of this study demonstrate the potential of deep learning and explainable artificial intelligence (AI) in nitrogen and chlorophyll estimation.
What role does canopy nitrogen and chlorophyll content play in optimizing fertilizer application?
It is common to use fertilizers to grow crops. Fertilizer provides farmers with the nutrients their crops need to grow and get the best yield. Regarding canopy nitrogen and chlorophyll content, these two variables are essential to enable accurate fertilization through monitoring (2). Analyzing nitrogen and chlorophyll content can also help make agricultural practices more sustainable and ensure less environmental impact (2).
Spectroscopy also plays a role in this process. For example, near-infrared (NIR) spectroscopy has long been used to analyze the biochemical components of plants. However, researchers have noted that traditional machine learning (ML) approaches often struggle to model the nonlinear behavior of spectral data (1). These limitations pose challenges when applying NIR spectroscopy to real-time nitrogen management, where both accuracy and model interpretability are important.
What did the researchers do in their study?
In their research, the research team developed a hybrid model that combines a convolutional neural network (CNN) and a gated recurrent unit (GRU), further enhanced with a convolutional block attention module (CBAM). This architecture allows the system to extract spatial and temporal patterns embedded in hyperspectral measurements while assigning dynamic weights to the most informative features (1). The addition of explainable artificial intelligence (AI) tools allows researchers and agronomists to understand how specific wavelengths contribute to model predictions (1).
As part of the experimental procedure, the research team collected hyperspectral images from 200 corn canopy samples and applied a rigorous preprocessing pipeline. Sequential Savitzky-Golay smoothing (SG), standard normal variate (SNV), and SG transformation improved the average test set R² increases by 0.016 units, improving the overall clarity of the spectral signal (1). The researchers then tried to remove the redundant data that had been collected. To do this, they employed two dimensionality reduction techniques: sequential projection algorithm (SPA) and competitive adaptive reweighting sampling (CARS). These two methods helped reduce the spectral feature set from 176 bands to 10 and 22 major wavelengths (1).
When we compared the CNN-GRU-CBAM model with most traditional machine learning (ML) and deep learning models, the CNN-GRU-CBAM model outperformed them. On the test set, the CNN-GRU-CBAM model is RNitrogen is 0.934 and chlorophyll is 0.788², giving low RMSE values of 1.940 and 0.216, respectively (1). To further validate the model, SHapley Additive exPlanations (SHAP) analysis identified the spectral regions most responsible for prediction accuracy and confirmed the biological relevance of the model output.
What are the main takeaways from this study?
The main takeaway from this research is that integrating deep learning and explainable AI has the potential to improve agricultural monitoring practices. As the researchers demonstrated in their study, the framework enables reversal of biochemical content across multiple crop species (1). By integrating deep learning and explainable AI, this approach provides more transparent, accurate, and scalable tools for modern precision agriculture.
References
- Kong, H. Tian, L. Yi, S. Others. Estimation of nitrogen and chlorophyll content in corn canopy using CNN-GRU-CBAM and hyperspectral imagery. J. Chemom. 2025, 39 (12), e70093. Doi:
10.1002/cem.70093 - Chao, J. Gao, A. Wang, B. et al. Modeling and visualization of nitrogen and chlorophyll in greenhouses Solanum lycopersicum L. leaves by hyperspectral imaging for nitrogen stress diagnosis. Plants (Basel) 2025, 14 (21), 3276. DOI:
10.3390/plant14213276
