Precision Estimation of Rice Nitrogen Fertilizer Topdressing According to the Nitrogen Nutrition Index Using UAV Multi-Spectral Remote Sensing: A Case Study in Southwest China
The precision estimation of N fertilizer application according to the nitrogen nutrition index (NNI) using unmanned aerial vehicle (UAV) multi-spectral measurements remains to be tested in different rice cultivars and planting areas. Therefore, two field experiments were conducted using varied N rat...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-04-01
|
| Series: | Plants |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2223-7747/14/8/1195 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The precision estimation of N fertilizer application according to the nitrogen nutrition index (NNI) using unmanned aerial vehicle (UAV) multi-spectral measurements remains to be tested in different rice cultivars and planting areas. Therefore, two field experiments were conducted using varied N rates (0, 60, 120, 160, and 200 kg N ha<sup>−1</sup>) on two rice cultivars, Yunjing37 (YJ-37, <i>Oryza sativa subsp. Japonica Kato.</i>, the Institute of Food Crops at the Yunnan Academy of Agricultural Sciences, Kunming, China) and Jiyou6135 (JY-6135, <i>Oryza sativa subsp. indica Kato.</i>, Hunan Longping Gaoke Nongping seed industry Co., Ltd., Changsha, China), in southwest China. The rice canopy spectral images were measured by the UAV’s multi-spectral remote sensing at three growing stages. The NNI was calculated based on the critical N (Nc) dilution curve. A random forest model integrating multi-vegetation indices established the NNI inversion, facilitating precise N topdressing through a linear platform of NNI-Relative Yield and the remote sensing NNI-based N balance approaches. The Nc dilution curve calibrated with aboveground dry matter demonstrated the highest accuracy (R<sup>2</sup> = 0.93, 0.97 for shoot components in cultivars YJ-37 and JY-6135), outperforming stem (R<sup>2</sup> = 0.70, 0.76) and leaf (R<sup>2</sup> = 0.80, 0.89) based models. The RF combined with six vegetation index combinations was found to be the best predictor of NNI at each growing period (YJ-37: R<sup>2</sup> is 0.70–0.97, RMSE is 0.02~0.04; JY-6135: R<sup>2</sup> is 0.71–0.92, RMSE is 0.04~0.05). The RF surpassed BPNN/PLSR by 6.14–10.10% in R<sup>2</sup> and 13.71–33.65% in error reduction across the critical rice growth stages. The topdressing amounts of YJ-37 and JY-6135 were 111–124 kg ha<sup>−1</sup> and 80–133 kg ha<sup>−1</sup>, with low errors of 2.50~8.73 kg ha<sup>−1</sup> for YJ-37 and 2.52~5.53 kg ha<sup>−1</sup> for JY-6135 in the jointing (JT) and heading (HD) stages. These results are promising for the precise topdressing of rice using a remote sensing NNI-based N balance method. The combination of UAV multi-spectral imaging with the NNI-nitrogen balance method was tested for the first time in southwest China, demonstrating its feasibility and offering a regional approach for precise rice topdressing. |
|---|---|
| ISSN: | 2223-7747 |