Abstract
Accurate estimation of aboveground dry matter accumulation and plant nitrogen content in summer maize is essential for optimizing both yield and nitrogen-use efficiency. Exclusive reliance on two-dimensional multispectral imagery results in data saturation and elevated estimation errors. This study proposes an integrated approach utilizing UAV-based multispectral data, SPAD index, and plant height index, employing deep learning algorithms to develop a precise model for inferring aboveground dry matter accumulation and plant nitrogen content. A field experiment incorporating five nitrogen application levels (N0: 0 kg·ha−1; N1: 120 kg·ha−1; N2: 240 kg·ha−1; N3: 300 kg·ha−1; N4: 360 kg·ha−1) and four summer maize varieties was conducted in the Huanghuaihai region. The results demonstrated that the aboveground dry matter accumulation and plant nitrogen content of the four maize varieties consistently followed a critical nitrogen dilution curve (CNDC) pattern (R2 ≥ 0.88), yielding a unified CNDC model (Nc = 34.92 ± 0.64DM−0.35±0.01, R2 = 0.94). The random forest (RF) model demonstrated exceptional precision in predicting aboveground dry matter accumulation (R2 = 0.94, RMSE = 1.56 t ha−1) and plant nitrogen content (R2 = 0.92, RMSE = 1.98 g/kg). This method exhibits higher accuracy compared to using vegetation index alone for predicting aboveground dry matter (R2 = 0.92, RMSE = 1.56) and plant nitrogen concentration (R2 = 0.91, RMSE = 2.49). Its performance significantly surpassed that of the support vector machine (SVM) and partial least squares regression (PLSR) models. This study indicates that the incorporation of SPADi and plant height index enhances the accuracy of drone multispectral-based random forest inversion models for nitrogen concentration and aboveground dry matter accumulation in summer maize throughout its growth period. Furthermore, when combined with key nitrogen dilution curves, this approach enables non-destructive and precise detection of nitrogen status in summer maize, thereby providing a scientific basis for nitrogen management and yield prediction.
Abstract from the paper’s OpenAlex record; open copy licensed cc-by. Read at the source ↗
The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.
Claims 8
Claim · model-written summary
Four maize varieties followed a critical nitrogen dilution curve with R2 at least 0.88.
Zea maysOrganisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“the aboveground dry matter accumulation and plant nitrogen content of the four maize varieties consistently followed a critical nitrogen dilution curve (CNDC) pattern (R2 ≥ 0.88)”
Claim · model-written summary
The unified CNDC model had R2 of 0.94.
Not reportedTaxon
Verbatim
e0“a unified CNDC model (Nc = 34.92 ± 0.64DM−0.35±0.01, R2 = 0.94)”
Claim · model-written summary
The RF model predicted aboveground dry matter with R2 0.94 and RMSE 1.56 t ha−1.
Not reportedTaxon
Verbatim
e0“predicting aboveground dry matter accumulation (R2 = 0.94, RMSE = 1.56 t ha−1)”
Claim · model-written summary
The RF model predicted plant nitrogen content with R2 0.92 and RMSE 1.98 g/kg.
Not reportedTaxon
Verbatim
e0“plant nitrogen content (R2 = 0.92, RMSE = 1.98 g/kg)”
Claim · model-written summary
The integrated method had higher accuracy than vegetation index alone for aboveground dry matter prediction.
- Outcome
- predicting aboveground dry mattere0
- Direction
- higher accuracye0
- Value
- R2 = 0.92e0
- Comparator
- vegetation index alonee0
Not reportedTaxon
Verbatim
e0“higher accuracy compared to using vegetation index alone for predicting aboveground dry matter (R2 = 0.92, RMSE = 1.56)”
Claim · model-written summary
The integrated method had higher accuracy than vegetation index alone for plant nitrogen concentration prediction.
- Outcome
- plant nitrogen concentratione0
- Direction
- higher accuracye0
- Value
- R2 = 0.91e0
- Comparator
- vegetation index alonee0
Not reportedTaxon
Verbatim
e0“higher accuracy compared to using vegetation index alone for predicting aboveground dry matter (R2 = 0.92, RMSE = 1.56) and plant nitrogen concentration (R2 = 0.91, RMSE = 2.49)”
Claim · model-written summary
The integrated model significantly outperformed SVM and PLSR models.
- Outcome
- performancee0
- Direction
- surpassede0
- Comparator
- support vector machine (SVM) and partial least squares regression (PLSR) modelse0
Not reportedTaxonValue
Verbatim
e0“Its performance significantly surpassed that of the support vector machine (SVM) and partial least squares regression (PLSR) models”
Claim · model-written summary
Adding SPADi and plant height index enhanced model accuracy for maize nitrogen concentration and aboveground dry matter.
Not reportedComparatorValue
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“incorporation of SPADi and plant height index enhances the accuracy of drone multispectral-based random forest inversion models for nitrogen concentration and aboveground dry matter accumulation in summer maize”