Development and Test of Rice Nitrogen Nutrition Index Estimation Model Based on Airborne Multi-Spectrum

Lijuan Song, Wanjun Ye, Shu Wang, Shubing Liu, M. C. Wang, Ge Su, Hongwen BiBangladesh Journal of Botany 2023 · DOI

Abstract

Both excessive and deficient nitrogen (N) concentration in the growing media can affect the growth, development, yield, and quality of rice. Traditional methods of N determination require destructive sampling, which is time-consuming and laborious. This study established a non-destructive N diagnosis model for cold-region rice (varieties Wuyoudao4 and Songjing9) in Heilongjiang Province, China, using UAV-mounted multispectral cameras. Canopy NDVI was used to estimate the Nitrogen Nutrition Index (NNI). Results showed that the interphase N diagnosis models (WYD-4: NNI=0.3916e^1.0809NDVI; SJ-9: NNI=0.3325e^1.2705NDVI) effectively estimate nitrogen status with nRMSE values between 10.36% and 12.43%.

Introduction

Nitrogen (N) is the primary driver of rice growth, but its management requires a delicate balance. Insufficient nitrogen leads to poor yields, while excessive application results in economic waste and environmental pollution. Traditionally, nitrogen status is determined through destructive sampling—harvesting, drying, and chemical analysis—which is labor-intensive and difficult to scale across large farms.

While satellite and ground-based sensors offer alternatives, they are often limited by cloud cover or slow operational speeds. This research addresses the need for a rapid, accurate, and non-destructive "middle ground" using Unmanned Aerial Vehicles (UAVs) to diagnose nitrogen needs in real-time, specifically tailored for the unique growth patterns of cold-region rice.

Methods and Materials

Field experiments were conducted over two years (2016–2017) in Wuchang, Heilongjiang Province, focusing on two rice cultivars: Wuyoudao4 (WYD-4) and Songjing9 (SJ-9).

Data Acquisition

An eight-rotor UAV equipped with a 6-band multispectral camera (470nm to 810nm) was used to capture canopy reflectance at an altitude of 100m. Synchronized ground truth data included Leaf Nitrogen Concentration (LNC), Leaf Dry Matter (LDM), and SPAD values.

Nitrogen Nutrition Index (NNI) Modeling

The study focused on the Nitrogen Nutrition Index (NNI), defined as the ratio of actual nitrogen concentration to the critical nitrogen concentration required for maximum growth. This approach accounts for the "dilution effect," where nitrogen concentration naturally decreases as biomass increases.

Temporal dynamics of canopy NDVI (WYD-4) and SPAD readings (SJ-9) across different nitrogen treatments.

The Normalized Difference Vegetation Index (NDVI) was extracted from the multispectral data and used to develop estimation models across key growth periods.

Results and Discussion

The researchers successfully established exponential models to estimate NNI based on canopy NDVI. The models performed best during the "interphase" between stem elongation and booting, as earlier stages were prone to spectral interference from soil and water backgrounds.

Model Validation

The variety-specific models were validated using an independent dataset from 2016:

  • Wuyoudao4 (WYD-4): $NNI = 0.3916e^{1.0809 \cdot NDVI}$ (RMSE=0.12, nRMSE=12.43%)
  • Songjing9 (SJ-9): $NNI = 0.3325e^{1.2705 \cdot NDVI}$ (RMSE=0.10, nRMSE=10.36%)
Validation of the NNI estimation models showing predicted vs. observed values for WYD-4 and SJ-9.

The study found that cultivar architecture influences model performance, with Songjing9 showing a stronger correlation ($R^2 = 0.502$) than Wuyoudao4 ($R^2 = 0.376$). However, the low nRMSE values across both varieties indicate high model stability for precision agriculture applications.

Conclusion

UAV-based multispectral remote sensing offers a robust, non-destructive method for real-time nitrogen diagnosis in cold-region rice. By incorporating biomass through the NNI, these models provide a more accurate assessment of crop needs than simple leaf nitrogen content. Future work will focus on developing background-masking algorithms for early growth stages and creating more universal models across diverse rice varieties.

Reference: Bangladesh J. Bot. 52(2): 529-538, 2023 (June) Special

Summary updated

Abstract

Both excessive and deficient nitrogen (N) concentration in the growing media can affect the growth, development, yield, and quality of rice. Traditional methods of N determination of plant require destructive and rigorous sampling, which is also time-consuming and laborious. However, rapid and non-destructive nitrogen diagnosis has become an important area of research in precision agriculture. Heilongjiang Province, China is a cold climate rice growing area, where the growth and fertilization of rice follows a definite pattern. In the present study, two varieties of rice (Wuyoudao4 and Songjing9) and as location Heilongjiang Province were selected. Nitrogen diagnosis of rice was carried out based on airborne multi-spectrum methodology. Canopy spectral data of rice at key growth periods were obtained by using a UAV equipped with a multi-spectral camera, and agronomic parameters such as leaf N content and dry matter weight were obtained synchronically. An airborne multispectral canopy normalized vegetation index (NDVI) model for N diagnosis based on its critical concentration curve was established as a nondestructive N diagnosis of rice for cold region. Results showed that canopy NDVI can estimate rice nitrogen nutrition index (NNI) properly over the growth period. The coefficient of determination R2, root mean square error (RMSE) and standard root mean square error (nRMSE) were compared to determine the best effect of the index model. The interphase nitrogen diagnosis model of WYD-4 based on NDVI for cold region was as follows: NNI=0.3916e1.0809*NDVI (RMSE=0.12, nRMSE=12.43%), SJ-9: NNI=0.3325e1.2705*NDVI (RMSE=0.10, nRMSE =10.36%), indicating that the established model can better estimate the nitrogen status of rice. Bangladesh J. Bot. 52(2): 529-538, 2023 (June) Special
BibTeX
@article{song2023development,
  author = {Lijuan Song and Wanjun Ye and Shu Wang and Shubing Liu and M. C. Wang and Ge Su and Hongwen Bi},
  title = {{Development and Test of Rice Nitrogen Nutrition Index Estimation Model Based on Airborne Multi-Spectrum}},
  journal = {Bangladesh Journal of Botany},
  year = {2023},
  doi = {10.3329/bjb.v52i2.68217},
}
Figure