A multi-scale remote sensing monitoring method for nitrogen and phosphorus nutrient limitation in evergreen broad-leaved forests

By constructing a multi-scale nitrogen-phosphorus ratio inversion model based on the hyperspectral data of arbor samples, the problem of insufficient universality of the method in remote sensing monitoring of nitrogen-phosphorus nutrients in evergreen broadleaf forests is solved, and more accurate large-area nitrogen-phosphorus ratio estimation is achieved and cross-region model versatility, providing fast and accurate nitrogen-phosphorus nutrient restriction discrimination.

CN118706762BActive Publication Date: 2024-11-08GUANGZHOU INST OF GEOGRAPHY GUANGDONG ACAD OF SCI
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411194537.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-29
Publication Date
2024-11-08
Estimated Expiration
2044-08-29

AI Technical Summary

Technical Problem

The existing technology has insufficient method universality in the remote sensing monitoring of nitrogen and phosphorus nutrients in evergreen broadleaf forests, lack of large-area accurate estimation, low spatial resolution, failure to accurately distinguish nitrogen and phosphorus limitations and low universality of cross-geographical regional models. .

Method used

Using multi-scale remote sensing monitoring method, the nitrogen-phosphorus ratio inversion model is constructed by obtaining leaf hyperspectral data and canopy hyperspectral images of typical tropical and subtropical forest ecosystems. , combined with cross-validation method, determine the optimal parameters, form a multi-scale inversion model, and perform cross-site area and year verification to generate a remote sensing space mapping of nitrogen and phosphorus ratio.

Benefits of technology

A more accurate nitrogen-phosphorus ratio estimation is achieved, which avoids uncertainty in separate estimation, and provides a fast and accurate inversion algorithm for large-area nitrogen-phosphorus nutrient limitations, improving the universality and accuracy of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118706762B_ABST
    Figure CN118706762B_ABST
Patent Text Reader

Abstract

The present invention discloses a multi-scale remote sensing monitoring method for nitrogen and phosphorus nutrient limitation in evergreen broad-leaved forests. By utilizing leaf hyperspectral data and canopy hyperspectral images of tree samples and based on linear and nonlinear nonparametric regression methods, a multi-scale inversion model for nitrogen and phosphorus nutrient limitation in evergreen broad-leaved forests is constructed, thereby obtaining more accurate estimation results and avoiding the uncertainty generated in the process of estimating nitrogen and phosphorus elements separately.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of remote sensing monitoring, and in particular to a multi-scale remote sensing monitoring method for nitrogen and phosphorus nutrient limitation in evergreen broad-leaved forests. Background Art

[0002] The supply of nitrogen and phosphorus can affect the growth of plants in most ecosystems, and the nitrogen-phosphorus ratio is often used to characterize the supply and limitation of the two elements. It is generally believed that a nitrogen-phosphorus ratio less than 14 is considered nitrogen-limited, while a nitrogen-phosphorus ratio greater than 16 is considered phosphorus-limited. Both nitrogen-limited and phosphorus-limited conditions will have a negative impact on plant growth. Most areas near the Tropic of Cancer in the world are deserts or semi-deserts. Only South my country has a large area of ​​subtropical evergreen forests, which account for about 25% of my country's total forest area and have a high species diversity. It plays an important role in carbon sinks, total primary productivity and biodiversity. However, subtropical and tropical forests are relatively deficient in phosphorus due to strong leaching. In recent decades, due to the intensification of atmospheric nitrogen deposition, the nitrogen supply of plants has increased, resulting in an imbalance in the nitrogen-phosphorus nutrient ratio, affecting the performance and function of plants. Plant growth is mostly phosphorus-limited. At present, there is a lack of quantitative characterization and detailed understanding of the nitrogen and phosphorus limitations of evergreen broad-leaved forest ecosystems.

[0003] Nitrogen and phosphorus are usually measured by field destructive sampling, where leaves are dried and ground before adding chemical reagents, and then measured using a spectrophotometer. This traditional method is time-consuming, labor-intensive, and environmentally unfriendly. Hyperspectral remote sensing can provide detailed narrow-band information to depict the physiological and biochemical components inside leaves, providing a new technical means for rapid, efficient, and large-scale estimation of nitrogen and phosphorus. Nitrogen content estimation based on hyperspectral remote sensing usually uses spectral features in the visible and short-wave infrared regions, on the one hand, using the strong correlation between leaf nitrogen and chlorophyll, and on the other hand, using the absorption characteristics of dry matter and protein. Remote sensing quantitative estimation of phosphorus content is more challenging, mainly because phosphorus lacks obvious spectral absorption characteristics, and its remote sensing inversion mostly uses indirect relationships, such as using the correlation between phosphorus and nitrogen. Since nitrogen and phosphorus account for a small proportion of leaves, their remote sensing estimates have large uncertainties, which further affects the estimation of nitrogen-phosphorus ratios, limiting the accurate identification of plant nitrogen and phosphorus nutrient limitations.

[0004] At present, there are the following problems with plant nitrogen and phosphorus nutrient limitation: first, most studies are limited to single-point level, lacking accurate estimation over a large area; second, the spatial resolution of discrimination is low, mostly staying at the kilometer grid level; third, most studies only focus on nitrogen limitation or phosphorus limitation, lacking comprehensive monitoring of the two nutrient elements; fourth, the remote sensing-based discrimination method is not accurate enough, especially when crossing geographical regions, years, and growing seasons, the model universality and inversion accuracy are very low; fifth, most of them are limited to leaf-level measurements, and there is a lack of research on inversion estimation that integrates multi-scale remote sensing data. Due to the above problems, most studies on plant nitrogen and phosphorus nutrient limitation remain at the single-point level, and there is still a lack of inversion algorithms that can accurately, quickly, and discriminate plant nitrogen and phosphorus nutrient limitation over a large area. Summary of the invention

[0005] In view of the above problems, the present invention proposes a multi-scale remote sensing monitoring method for nitrogen and phosphorus nutrient limitation in evergreen broad-leaved forests, which mainly solves the problem of insufficient versatility of existing remote sensing monitoring methods for nitrogen and phosphorus nutrients in plants.

[0006] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0007] A multi-scale remote sensing monitoring method for nitrogen and phosphorus nutrient limitation in evergreen broad-leaved forests, comprising the following steps:

[0008] Obtain tree samples of typical tropical and subtropical forest ecosystems at different phases, measure leaf hyperspectral data of the tree samples, and determine the sample nitrogen-phosphorus ratio of the tree samples;

[0009] Obtain canopy hyperspectral images of sampled tree crowns at different phases, as well as the corresponding canopy spectral reflectance;

[0010] The canopy hyperspectral images are stitched into sampling area images, and non-vegetation and shadow areas in the sampling area images are removed according to the canopy spectral reflectance to obtain the average spectrum corresponding to each of the sampled tree crowns;

[0011] Based on the sample nitrogen-phosphorus ratio and the average spectrum, a first nitrogen-phosphorus ratio inversion model and a second nitrogen-phosphorus ratio inversion model are constructed respectively by using a linear nonparametric regression method and a nonlinear nonparametric regression method, and the optimal parameters of the first nitrogen-phosphorus ratio inversion model and the second nitrogen-phosphorus ratio inversion model are obtained by using a cross-validation method, and the optimal parameters are fused to form a multi-scale inversion model;

[0012] The inversion results of the multi-scale inversion model are verified across site areas and years, one site area or year data of the inversion results is retained as a verification sample, the other remaining data of the inversion results are used as training samples of the multi-scale inversion model, and the accuracy and universality of the trained multi-scale inversion model are evaluated using the verification samples;

[0013] The multi-scale inversion model that meets the output conditions is applied to the target area, and non-vegetation in the target area is removed to obtain a nitrogen-phosphorus ratio remote sensing spatial map. Finally, the nitrogen-phosphorus ratio remote sensing spatial map is divided into several intervals according to preset plant nitrogen and phosphorus nutrient limitation threshold conditions.

[0014] In some embodiments, the tree samples include dry leaves and fresh leaves.

[0015] In some embodiments, the nonlinear nonparametric regression method is a random forest regression method.

[0016] In some embodiments, the number of decision trees of the random forest regression method is set to at least 500, wherein the number of variables used for segmentation at each node of the decision tree is set to 5 to 70, and the interval is set to 5.

[0017] In some embodiments, the evaluation index of the multi-scale inversion model includes at least one of the coefficient of determination between the measured value and the predicted value, the root mean square error, the deviation and the relative root mean square error.

[0018] In some embodiments, the plant nitrogen and phosphorus nutrient limitation threshold conditions include: nitrogen to phosphorus ratio <14, 14<nitrogen to phosphorus ratio <16, 16<nitrogen to phosphorus ratio <20, 20<nitrogen to phosphorus ratio <30, and nitrogen to phosphorus ratio >30.

[0019] In some embodiments, the nitrogen-phosphorus ratio remote sensing spatial mapping is divided into a nitrogen-limited interval, an unrestricted interval, a low-phosphorus-limited interval, a moderate-phosphorus-limited interval, and a severe-phosphorus-limited interval, and each interval corresponds to the plant nitrogen and phosphorus nutrient limitation threshold conditions in turn.

[0020] The beneficial effects of the present invention are as follows: by utilizing leaf hyperspectral data and canopy hyperspectral images of tree samples, based on linear and nonlinear nonparametric regression methods, a multi-scale inversion model of nitrogen and phosphorus nutrient limitation in evergreen broad-leaved forests is constructed, so that more accurate estimation results can be obtained, avoiding the uncertainty generated in the process of estimating nitrogen and phosphorus elements separately. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A schematic diagram of a process flow of a multi-scale remote sensing monitoring method for nitrogen and phosphorus nutrient limitation in evergreen broad-leaved forests disclosed in an embodiment of the present invention;

[0022] Figure 2 The canopy hyperspectral image collected by drone in the target area and the schematic diagram of the canopy positioning of field sampling;

[0023] Figure 3 This is a schematic diagram of the nitrogen-to-phosphorus ratio variation range for typical stations in the subtropical and tropical regions of South China;

[0024] Figure 4 Hyperspectral data of dry leaves from typical stations in subtropical and tropical regions of South China;

[0025] Figure 5 Hyperspectral data of fresh leaves from typical stations in subtropical and tropical regions of South China;

[0026] Figure 6 Hyperspectral images of the canopy collected by drones;

[0027] Figure 7 Plant nitrogen-to-phosphorus ratios predicted for the multiscale inversion model;

[0028] Figure 8 Schematic diagram of the root mean square error of the plant nitrogen-phosphorus ratio predicted by the model and measured by the leaf hyperspectral data of dry and fresh leaves across sites;

[0029] Fig. 9 Schematic diagram of the importance of variable features in the multi-scale plant nitrogen-phosphorus ratio model of dry leaves, fresh leaves and canopy for partial least squares regression and random forest regression;

[0030] Fig.10 The spatial distribution map and density diagram of plant nitrogen-phosphorus ratio obtained by inversion of partial least squares regression and random forest regression models;

[0031] Fig.11 Schematic diagram of remote sensing monitoring results of nitrogen and phosphorus nutrient limitation of plants in the target area. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the content of the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. It is understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for the convenience of description, only the parts related to the present invention are shown in the accompanying drawings, rather than all the contents.

[0033] This embodiment proposes a multi-scale remote sensing monitoring method for nitrogen and phosphorus nutrient limitation in evergreen broad-leaved forests. Figure 1 As shown, the following steps are included:

[0034] S1, obtain tree samples in typical tropical and subtropical forest ecosystems at different phases, measure leaf hyperspectral data of tree samples, and determine the sample nitrogen-phosphorus ratio of tree samples.

[0035] The tree samples obtained in this embodiment include dry leaves and fresh leaves, taking the Heshan, Dapeng and Liuzhou sites in the subtropical region and the Wanning site in the tropical region as examples. Figure 2As shown. A field survey was conducted in 2020-2021 to obtain the leaves of the top sun leaves of the canopy at different phases, including 97 species, including 196 samples in Heshan, 24 sites in Dapeng, 39 samples in Liuzhou, and 101 samples in Wanning. The leaf samples collected were used to obtain the leaf hyperspectral data of fresh leaves using a Malvern ground feature spectrometer and a contact leaf clamp. The fresh leaves were then dried and the leaf hyperspectral data of the dry leaves were measured. The nitrogen and phosphorus contents of the dry leaves were measured using the Kjeldahl method and a spectrophotometer, and the nitrogen-phosphorus ratio was calculated. At this point, the leaf hyperspectral data of the tree samples (including hyperspectral images of dry leaves and fresh leaves) and the nitrogen-phosphorus ratio were obtained for subsequent calculations.

[0036] like Figure 3 As shown, Heshan had the highest nitrogen-phosphorus ratio in 2020, and Wanning had the lowest nitrogen-phosphorus ratio. The difference in nitrogen-phosphorus ratios between Heshan in 2020 and 2021 was large, and the difference in nitrogen-phosphorus ratios of plants at Dapeng and Liuzhou stations was small. The average nitrogen-phosphorus ratio of each station exceeded 20, with a minimum value of 7.34 and a maximum value of 49.90.

[0037] like Figure 4 As shown, the dry leaf spectrum varies the most in the visible light 550-650nm, near infrared 1100-1200nm and short-wave infrared 2100-2300nm bands. Dapeng has the highest dry leaf spectrum in the visible light region, Wanning has the highest short-wave infrared reflectivity, and the differences between the various sites in the 1400-1800nm ​​band are relatively small. For the fresh leaf spectrum, the variations of the various sites in the entire spectral range are relatively large. Figure 5 As shown, except for the near infrared 800-1300nm, Wanning has the highest average fresh leaf spectrum. In the visible light region, the fresh leaf spectra of Wanning, Heshan 2021 and Dapeng are higher than those of Heshan 2020 and Liuzhou. In the near infrared band, Dapeng has the highest average spectrum, followed by Wanning, and Liuzhou has the lowest. In the short-wave infrared band, Wanning has the highest fresh leaf spectrum, followed by Liuzhou, while Heshan 2020 and Dapeng have lower spectra.

[0038] S2, obtain canopy hyperspectral images of sampled tree crowns at different phases and the corresponding canopy spectral reflectance.

[0039] In step S2, for the subtropical evergreen broad-leaved forest in South China described in S1, an unmanned aerial vehicle equipped with an imaging spectrometer can be used. The unmanned aerial vehicle flies at an altitude of about 85 meters and obtains hyperspectral remote sensing images around noon in May and October when the weather is clear and cloudless with little wind, such as Figure 2 As shown in the figure, the red circle represents the sampled tree crown. Furthermore, the visible light camera carried by the UAV was used to obtain RGB images with a flight altitude of 135m and a spatial resolution of 0.03m to assist in identifying the sampled tree crown in the hyperspectral remote sensing image.

[0040] See also Figure 6For the canopy spectra collected by drones, Heshan 2020 and 2021 are very similar in the visible light band, but there are large differences in the red edge and near-infrared spectral range. Due to multiple scattering caused by a higher leaf area index, the near-infrared spectrum of Heshan in 2020 is significantly higher than that of Heshan in 2021.

[0041] While the UAV is acquiring canopy hyperspectral images, it can also synchronously measure solar radiation energy based on the ASD ground object spectrometer, and obtain the canopy reflectance through preprocessing such as radiation calibration and atmospheric correction.

[0042] S3, stitching the canopy hyperspectral images into sampling area images, removing non-vegetation and shadow areas in the sampling area images according to the canopy spectral reflectance, and obtaining the average spectrum corresponding to each sampled tree crown.

[0043] In step S3, the hyperspectral images obtained in S2 were spliced ​​using Agisoft Metashape software to obtain the image of the entire sampling area. The pixels of each tree crown were obtained based on the GPS information position of the sampled tree crowns and the high-resolution visible light images of the drone. The non-vegetation and shadow areas were removed using the threshold condition of near-infrared reflectivity less than 0.5 to obtain the average spectrum of each sampled tree crown.

[0044] S4, based on the sample nitrogen-phosphorus ratio and the average spectrum, the first nitrogen-phosphorus ratio inversion model and the second nitrogen-phosphorus ratio inversion model were constructed respectively by using the linear nonparametric regression method and the nonlinear nonparametric regression method. The optimal parameters of the first nitrogen-phosphorus ratio inversion model and the second nitrogen-phosphorus ratio inversion model were obtained by using the cross-validation method, and the optimal parameters were fused to form a multi-scale inversion model.

[0045] In one example, two nitrogen-phosphorus ratio inversion models were constructed using the partial least squares regression method and the random forest regression method for the leaf hyperspectral data of fresh and dry leaves obtained by S1 and the hyperspectral remote sensing images obtained by the drone. Among them, partial least squares regression is a linear non-parametric model that can convert raw data into a small number of potential factors and consider the covariance of the response variable and the predictor variable. In order to reduce the overfitting of the simulation, the cross-validation method is used to determine the model parameters, that is, the number of potential factors. Random forest regression is a nonlinear non-parametric model that uses a self-sampling technique to construct a certain number of decision trees for inversion. The number of decision trees in the random forest regression method is set to at least 500, wherein the number of variables used for segmentation at the node of each decision tree is set to 5 to 70, and the interval is set to 5, such as 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65 or 70, and its value is determined by the k-fold cross-validation method.

[0046] S5, verify the inversion results of the multi-scale inversion model across site areas and years, retain one site area or year data of the inversion results as a verification sample, and use the other remaining data of the inversion results as training samples of the multi-scale inversion model. Use the verification samples to evaluate the accuracy and versatility of the trained multi-scale inversion model.

[0047] There are two ways to verify the nitrogen-phosphorus ratio inversion model, the main difference is the sample segmentation. The first way is to randomly divide all the data into a training set (2 / 3) and a validation set (1 / 3). The second way is to use the data from one station or one period as the training set, and the data collected from the remaining stations or another period as the validation set. The second way can better test the universality and generalizability of the model. The evaluation indicators of the multi-scale inversion model include the determination coefficient R between the measured value and the predicted value 2 , root mean square error RMSE, bias BIAS and relative root mean square error NRMSE.

[0048] like Figure 7 As shown in Table 1, at the dry leaf level, partial least squares regression has a good prediction accuracy for nitrogen-phosphorus ratio, with an R2 of 0.45. Fresh leaf spectra have higher prediction accuracy than dry leaf spectra, with an R2 of more than 0.50 and a root mean square error of 4.04. When pushed up to the canopy spectrum, the random forest regression model can more accurately predict the nitrogen-phosphorus ratio, and its accuracy is generally higher than partial least squares regression, with an R2 of 0.61 and a root mean square error of 3.79.

[0049] Table 1. Multi-scale remote sensing inversion accuracy of plant nitrogen-phosphorus ratio (randomized cross-validation)

[0050]

[0051] like Figure 8 As shown in Table 2, when the nitrogen-phosphorus ratio models of various scales are used for cross-site and cross-year predictions, the relative errors are large. Among them, the root mean square error of the dry leaf model is about 5, and the error is highest for Heshan 2020. The error of the fresh leaf model is slightly higher than that of the dry leaf model when predicting across sites, and the error is highest for the Dapeng site. For the cross-year prediction of the canopy model, both the partial least squares regression and the random forest regression models produce large errors. The main reason is that there are large differences in the nitrogen-phosphorus ratio range and canopy spectrum collected in the two phases, which makes it difficult to extrapolate the inversion model. Therefore, in future field sampling, the variability of nitrogen-phosphorus ratio and canopy reflectance should be fully considered, so as to scientifically, reasonably and more effectively sample, and then establish a multi-scale remote sensing inversion model of nitrogen-phosphorus ratio with higher versatility and extrapolation.

[0052] Table 2. Accuracy of multi-scale remote sensing inversion of plant nitrogen-phosphorus ratio (validated across sites and years)

[0053]

[0054] like Fig. 9 As shown in the figure, by analyzing the importance of variables in the partial least squares and random forest regression models, it was found that the two were highly correlated, with the Pearson correlation coefficient between 0.57 and 0.84, and the highest correlation was found for the fresh leaf model. The nitrogen-phosphorus ratio dry leaf prediction model was most important in the shortwave infrared 1900-2000nm and 2250-2300nm bands. The fresh leaf model was more important in the visible light region of 550nm and the red edge region, followed by the importance around 1400nm and 1900nm. For the canopy model, the more important bands included 550-570nm, 680-700nm, and 730-750nm.

[0055] like Fig.10 As shown in the figure, the partial least squares regression and random forest regression models are applied to the UAV hyperspectral remote sensing images to obtain the spatial distribution map and probability density map of the plant nitrogen-phosphorus ratio. As can be seen from the figure, the nitrogen-phosphorus ratio inversion results obtained by the two models are relatively consistent, which is consistent with the field survey results. However, for the two different phases of 2020 and 2021, there are obvious differences in the nitrogen-phosphorus ratio. 2020 is significantly higher than 2021, and the two models show similar differences.

[0056] S6, applying the multi-scale inversion model that meets the output conditions to the target area, removing non-vegetation in the target area, obtaining the nitrogen-phosphorus ratio remote sensing spatial mapping, and finally dividing the nitrogen-phosphorus ratio remote sensing spatial mapping into several intervals according to the preset plant nitrogen and phosphorus nutrient limitation threshold conditions.

[0057] In one example, plant nitrogen and phosphorus nutrient limitation threshold conditions include: nitrogen-phosphorus ratio <14, 14<nitrogen-phosphorus ratio <16, 16<nitrogen-phosphorus ratio <20, 20<nitrogen-phosphorus ratio <30, and nitrogen-phosphorus ratio >30. According to the plant nitrogen and phosphorus nutrient limitation threshold conditions, the nitrogen-phosphorus ratio remote sensing spatial mapping is divided into a nitrogen limitation interval, an unrestricted interval, a low phosphorus limitation interval, a moderate phosphorus limitation interval, and a severe phosphorus limitation interval.

[0058] like Fig.11 As shown in the figure, in 2020, the nitrogen-phosphorus ratio of plants in most areas of the study area was greater than 16, ranging from 20-30, mainly showing moderate phosphorus limitation, and there was a certain degree of severe phosphorus limitation in the southeastern area in the lower right corner. In 2021, it was quite different, with a wider distribution range of nitrogen-phosphorus ratio, and plant growth was subject to both nitrogen and phosphorus limitations. It can be seen from the distribution map that there are more nitrogen-limited growth in the eastern, central and southern regions of the study area, mainly native broad-leaved tree species and acacia species, and there are more phosphorus growth areas in the north and west. Compared with 2020, some forests in the central and southern parts of 2021 were also in a better nitrogen-phosphorus-free state.

[0059] The above results show that the hyperspectral data of dry leaves, fresh leaves and canopy provide a fast, accurate and large-scale remote sensing method for estimating the nitrogen and phosphorus growth limitations of plants. By directly constructing a quantitative inversion model for the nitrogen-phosphorus ratio, more accurate estimation results can be obtained, avoiding the uncertainty generated by estimating nitrogen and phosphorus separately. The study can accurately characterize the nitrogen-phosphorus ratio of plants in typical subtropical and tropical regions in southern my country, and evaluate the generalizability of the model through cross-site and cross-year cross-validation. The results help to understand the growth of evergreen broad-leaved forest ecosystems in the region and their response to global change, and also provide an important scientific basis for efficient and accurate forest cultivation and management.

[0060] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable ordinary technicians in the field to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made based on the essence of the content of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-scale remote sensing monitoring method for nitrogen and phosphorus nutrient limitation in evergreen broad-leaved forests, characterized in that: The following steps are involved: Obtain tree samples of typical tropical and subtropical forest ecosystems at different phases, measure leaf hyperspectral data of the tree samples, and determine the sample nitrogen-phosphorus ratio of the tree samples; Obtain canopy hyperspectral images of sampled tree crowns at different phases, as well as the corresponding canopy spectral reflectance; The canopy hyperspectral images are stitched into sampling area images, and non-vegetation and shadow areas in the sampling area images are removed according to the canopy spectral reflectance to obtain the average spectrum corresponding to each of the sampled tree crowns; Based on the leaf hyperspectral data of the tree sample, the sample nitrogen-phosphorus ratio and the average spectrum corresponding to the sampled tree crown, a first nitrogen-phosphorus ratio inversion model and a second nitrogen-phosphorus ratio inversion model are constructed respectively by using a linear nonparametric regression method and a nonlinear nonparametric regression method, and the optimal parameters of the first nitrogen-phosphorus ratio inversion model and the second nitrogen-phosphorus ratio inversion model are obtained by using a cross-validation method, and the optimal parameters are fused to form a multi-scale inversion model; The inversion results of the multi-scale inversion model are verified across site areas and years, one site area or year data of the inversion results is retained as a verification sample, and the other remaining site areas or remaining year data of the inversion results are used as training samples of the multi-scale inversion model, and the accuracy and universality of the trained multi-scale inversion model are evaluated using the verification samples; Applying the multi-scale inversion model that meets the output conditions to the target area, removing non-vegetation in the target area, obtaining a nitrogen-phosphorus ratio remote sensing spatial mapping, and finally dividing the nitrogen-phosphorus ratio remote sensing spatial mapping into a number of intervals according to a preset plant nitrogen and phosphorus nutrient limitation threshold condition; The tree samples include dry leaves and fresh leaves; the linear nonparametric regression method adopts partial least squares regression method; The nonlinear nonparametric regression method is a random forest regression method; The number of decision trees of the random forest regression method is set to at least 500, wherein the number of variables used for segmentation at the node of each decision tree is set to 5 to 70, and the interval is set to 5.

2. The multi-scale remote sensing monitoring method for nitrogen and phosphorus nutrient limitation in evergreen broad-leaved forests according to claim 1, characterized in that: The evaluation index of the multi-scale inversion model includes at least one of a determination coefficient between a measured value and a predicted value, a root mean square error, a deviation and a relative root mean square error.

3. The multi-scale remote sensing monitoring method for nitrogen and phosphorus nutrient limitation in evergreen broad-leaved forests according to claim 1, characterized in that: The plant nitrogen and phosphorus nutrient limitation threshold conditions include: nitrogen-phosphorus ratio <14, 14<nitrogen-phosphorus ratio <16, 16<nitrogen-phosphorus ratio <20, 20<nitrogen-phosphorus ratio <30, and nitrogen-phosphorus ratio >30.

4. The multi-scale remote sensing monitoring method for nitrogen and phosphorus nutrient limitation in evergreen broad-leaved forests according to claim 3, characterized in that: The nitrogen-phosphorus ratio remote sensing spatial mapping is divided into a nitrogen-limited interval, an unrestricted interval, a low-phosphorus-limited interval, a moderate-phosphorus-limited interval and a severe-phosphorus-limited interval, and each interval corresponds to the plant nitrogen and phosphorus nutrient limitation threshold conditions in turn.

Citation Information

Patent Citations

  • Inversion model method for chlorophyll content in japonica rice canopy based on particle swarm optimization extreme learning machine (PSI-ELM)

    CN109030378A

  • Rape nitrogen-phosphorus-potassium nutrient diagnosis method based on near-earth hyperspectral data and integrated model

    CN111257240A

  • Fruit tree canopy nitrogen content remote sensing inversion method and system based on image shadow removal

    CN112179853A

  • Method and device for monitoring comprehensive growth of potted lettuce

    US20210056685A1