Next Article in Journal
Upper Mantle beneath the Myanmar and Surrounding Tomography: New Insight into Plate Subduction and Volcanism
Next Article in Special Issue
The Use of an Airborne Laser Scanner for Rapid Identification of Invasive Tree Species Acer negundo in Riparian Forests
Previous Article in Journal
Oriented Object Detection Based on Foreground Feature Enhancement in Remote Sensing Images
Previous Article in Special Issue
Assessment of Iran’s Mangrove Forest Dynamics (1990–2020) Using Landsat Time Series
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 14
Issue 24
10.3390/rs14246227
remotesensing-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Tree Species Classification Using Plant Functional Traits and Leaf Spectral Properties along the Vertical Canopy Position

by
Yicen Zhang
1,2,
Junjie Wang
3,
Zhifeng Wu
1,
Juyu Lian
4,
Wanhui Ye
4 and
Fangyuan Yu
1,*
1
School of Geographical Sciences and Remote Sensing, Guangzhou University, Guangzhou 510006, China
2
State Key Laboratory of Vegetation and Environmental Change, Institute of Botany, The Chinese Academy of Sciences, Beijing 100093, China
3
College of Life Sciences and Oceanography, Shenzhen University, Shenzhen 518060, China
4
Key Laboratory of Vegetation Restoration and Management of Degraded Ecosystems, South China Botanical Garden, Chinese Academy of Sciences, Guangzhou 510650, China
*
Author to whom correspondence should be addressed.
Remote Sens. 2022, 14(24), 6227; https://doi.org/10.3390/rs14246227
Submission received: 2 November 2022 / Revised: 6 December 2022 / Accepted: 6 December 2022 / Published: 8 December 2022
(This article belongs to the Special Issue Applications of Remote Sensing in Forest Management and Biodiversity Conservation)
Browse Figures
Versions Notes

Abstract

:
Plant functional traits are rarely used in tree species classification, and the impact of vertical canopy positions on collecting samples for classification also remains unclear. We aim to explore the feasibility and effectiveness of leaf traits in classification, as well as to detect the effect of vertical position on classification accuracy. This work will deepen our understanding of the ecological mechanism of natural forest structure and succession from new perspectives. In this study, we collected foliar samples from three canopy layers (upper, middle and lower) and measured their spectra, as well as eight well-known leaf traits. We used a leaf hyperspectral reflectance (LHR) dataset, leaf functional traits (LFT) dataset and LFT + LHR dataset to classify six dominant tree species in a subtropical evergreen broad-leaved forest. Our results showed that the LFT + LHR dataset achieved the highest classification results (overall accuracy (OA) = 77.65% and Kappa = 0.73), followed by the LFT dataset (OA = 74.26% and Kappa = 0.69) and the LHR dataset (OA = 69.06% and Kappa = 0.63). Along the vertical canopy, the OA and Kappa increased from the lower to the upper layers, and the combination data of the three canopy layers achieved the highest accuracy. For the individual tree species, the shade-tolerant species (including Machilus chinensis, Cryptocarya chinensis and Cryptocarya concinna) produced higher accuracies than the light-demanding species (including Schima superba and Castanopsis chinensis). Our results provide an approach for enhancing tree species recognition from the plant physiology and biochemistry perspective and emphasize the importance of vertical direction in forest community research.

Graphical Abstract

1. Introduction

Tree species classification is an important foundation for forest inventory, management and conservation. It provides vital information on tree species composition [1,2] and forest structure [3,4], as well as supports species distribution mapping [5,6]. The results of tree species classification are able to be used to build species-specific allometric models for estimating volume, above-ground biomass and carbon stocks [7,8,9,10]. Moreover, tree species information also can explain biochemical patterns and assist in forest canopy biochemistry estimation [11,12,13]. Therefore, accurate identification of individual tree species is of great significance in ecological research.
Traditionally, obtaining species information requires costly, labor-intensive and time-consuming field investigations, and it is often difficult for tropical and subtropical forests due to denser forest stands and greater and irregular species mixtures. In recent years, remote-sensing-assisted classification of tree species has been increasingly adopted due to its efficient, convenient and informative characteristics [14,15,16]. In general, aerial multispectral/hyperspectral, as well as light detection and ranging (LiDAR) data are widely used for tree classification at the canopy scale [17,18]. At the leaf scale, leaf spectral reflectance is the dominant data source for individual tree species classification [19,20,21]. Obtaining higher accuracy is a common pursuit in classification, and data fusion is one of the main approaches to achieve this goal [22,23]. Clark et al. [24] combined leaf and bark hyperspectral data to recognize seven tree species in tropical wet forest. In another study, van Deventer et al. [25] used multi-season leaf hyperspectral reflectance to separate six tree species in the subtropical coastal forest. However, very few studies have considered the contribution of plant traits to the tree species classification, the combination of leaf hyperspectral and plant traits that advance or hinder species detection remains under-examined.
Species-specific information covering the aspects of structural, spectral and genetic properties is an important factor in understanding tree species classification with remote sensing [19,26,27]. Plant functional traits are considered as crucial information for reconciling dissimilarities between tree species [28]. Functional traits refer to the survival strategies of plants regulated by both abiotic and biotic factors [29,30]. Previous studies have suggested that the interspecific difference of plant traits generally occupies a substantial portion of the total variation [11,31,32], which indicates that plant traits are also the basis for differentiating individual tree species. Plant traits provide species-specific feature information in morphology, physiology and biochemistry [18]. Specifically, leaf chemical traits have been regarded as unique chemical fingerprints of species [27]; the variations of nitrogen, phosphorus and carbon could be well-explained by species [12,33,34]. Additionally, other plant traits, such as leaf mass per area [35], equivalent water thickness [36,37] and leaf dry matter content [38] also showed a marked difference among tree species.
There is a close relationship between plant functional traits and remote sensing spectral reflectance. The capabilities of remote sensing for determining plant functional traits have also been established. Ali et al. [39] quantified the leaf dry matter content and specific leaf area of a mixed mountain forest with R2 values of 0.59 and 0.85, respectively, based on airborne hyperspectral data. Wang et al. [12] estimated the leaf nitrogen concentration of a mixed forest with an R2 of 0.79 using the normalized difference nitrogen index calculated from hyperspectral data. Helfenstein et al. [40] retrieved the chlorophyll, carotenoid and water content of a temperate forest with r values of 0.896, 0.813 and 0.773, respectively, using corresponding spectral indices. However, most studies involving the interrelation between spectra and plant traits only focused on the top of the crown (a horizontal dimension) and neglected the influence of the vertical canopy positions on the results. In fact, the understory condition of forest is rather complicated. The forest canopy creates a heterogeneous environment where a gradient variation of light quality, humidity and other environmental factors is commonly observed, which facilitates the functional differentiation of individual tree species themselves in different canopy positions. Yu et al. [41] suggested that the vertical canopy has a marked impact on the spectral properties and plant functional traits of tree species. Nevertheless, the specific influence of canopy positions on tree species classification accuracy has not been tested yet. In addition, most research was conducted in simple-structure forests, such as a temperate forest [42], a dry forest [43], a coniferous forest [44] and even an urban forest [45], but tropical and subtropical forests were seldom explored due to their intricate canopy, dense forest stands and extraordinary species richness. Using the upper canopy alone to represent the entire canopy for tree species classification, especially in a forest with a complex structure and rich biodiversity, should be done with caution. Therefore, applying hyperspectral reflectance and functional traits at the leaf scale to explore the variation in classification accuracy along the vertical canopy position can develop the feature fusion method of species classification at the leaf scale and evaluate the effect of the vertical canopy position on classification accuracy. It would deepen our understanding of the underlying process of individual trees in chemical and structural variability across the vertical canopy profile, which is associated with species classification in a complex community.
In this study, we sampled the leaf material of six dominant tree species along the vertical canopy position in a subtropical evergreen broad-leaved forest and aimed to explore the variation of tree species classification by combining plant functional traits and leaf spectral properties along the vertical canopy position in a subtropical evergreen broad-leaved forest. Specifically, we aim to (1) investigate the variation of plant functional traits and leaf spectral properties along the vertical position, (2) explore the potential tree species classification by combing plant functional traits and leaf spectral properties and (3) evaluate the capability of plant functional traits for discriminating individual tree species along the vertical position.

2. Materials and Methods

2.1. Study Area

Our study was implemented with a tower crane in Dinghushan Nature Reserve (Figure 1, 23.15°–23.19°N, 112.51°–112.56°E), Guangdong Province, southern China. The sample plot covers 1.44 ha (120 m × 120 m) cantered on the tower crane, which is 60 m high, with a 60 m boom arm. It was established following the protocols from the CTFS-Forest GEO network. This region is dominated by a typical subtropical monsoon climate with an average annual temperature of 20.9 °C and average annual precipitation of 1927 mm [46]. The altitude of the tower crane plot is 48–100 m, and it is a typical subtropical monsoon evergreen broad-leaved forest with a complicated vertical canopy [47].

2.2. Field Sampling

There are about 4879 individual trees in the tower crane plot, according to the survey in November 2018 [41]. We selected six dominant tree species (Table 1), including Schima superba (Schisu), Castanopsis chinensis (Castch), Castanopsis fissa (Castfi), Machilus chinensis (Machch), Cryptocarya chinensis (Crypch) and Cryptocarya concinna (Crypco), and collected the foliar samples randomly on sunny days from 7:30 to 10:00 a.m. on 10–13 August 2019. Meanwhile, we divided the vertical canopy into a lower layer (<4 m from the ground), middle layer (4–11.3 m from the ground) and upper layer (>11.3 m from the ground) according to the standard of Gui et al. [47].
For each tree species, we chose 12–20 mature and healthy individual trees (89 trees in total) and collected 15–30 mature and healthy leaves at three layers of every single tree. The 267 fresh foliar samples were preserved in a valve bag after collection, then transported to the laboratory to measure their fresh weight, leaf area and leaf spectra within four hours.

2.3. Measurement of Leaf Spectra

We measured the leaf spectral reflectance from 350–2500 nm for each foliar sample using an ASD FieldSpec-4 portable spectroradiometer (Analytical Spectral Devices, Inc., Boulder, CO, USA) with a spectral resolution of 1.4 nm for the region of 350–1000 nm and 2 nm for the region of 1000–2500 nm. Then, we used an ASD fitted leaf clip and plant probe to measure the reflectance spectra of ten individual leaves chosen randomly. In the end, we obtained the final leaf spectrum of each sample by averaging the ten reflectance spectra. To improve the signal-to-noise ratio, we eliminated the spectral reflectance of 350–399 nm and 2201–2500 nm and obtained a final range of leaf spectral reflectance of 400–2200 nm.

2.4. Measurement of Leaf Traits

According to the standard methods of Cornelissen et al. [48], we selected and measured eight widely-used leaf traits, including leaf dry matter content (LDMC), specific leaf area (SLA), equivalent water thickness (EWT), leaf chlorophyll content (Chl), flavonoid content (Flav), leaf carbon content (LCC), leaf nitrogen content (LNC) and leaf phosphorus content (LPC). The measurement processes were as follows: First, we removed the petiole of each foliar sample using scissors and measured the fresh weight of each sample with a high-precision (0.001 g) digital scale. We also measured the Flav with a Dualex Scientific+ (Force-A, Orsay, France) meter and the leaf surface area with a LI-3000C Portable Area Meter (LI-COR, Inc., Lincoln, NE, USA). Subsequently, we placed the fresh samples into a drying oven at 65 °C for 48 h, then measured the dry weight by digital scale. Additionally, we calculated the LDMC, SLA and EWT using the basic data described above. Finally, we ground the dried samples in a mortar and stored the powder samples in paper bags for the measurement of leaf total carbon, nitrogen and phosphorus that were used to calculate the LCC, LNC and LPC.

2.5. Random Forest Classifier

Random forest is an ensemble classifier that combines multiple classification trees to successfully achieve higher classification accuracy. It has been widely used in ecological research, such as species classification and regression [49]. In this study, we divided the classification datasets into 60% and 40% ratios for training and testing, respectively. The former was used to train the algorithm for obtaining the optimal parameters; the latter was used to test the generalization ability of the specified classifier. Then, we chose the optimal mtry value by ‘10 Repeated 5-Fold Cross-Validation’ and set the ntree value as 500 for model training. We used three datasets: (1) the leaf functional traits (LFT) dataset only, (2) the leaf hyperspectral reflectance (LHR) dataset only and (3) the combined LFT + LHR dataset to classify the tree species, respectively. For each operation of classification, we set 100 iterations to obtain user accuracy, overall accuracy (OA) and Kappa from the confusion matrix, which uses true positive (TP), false positive (FP), true negative (TN) and false negative (FN) to show predicted correct or wrong in binary classification. The formulas for the evaluation parameters are as follows:
User accuracy = TP/(TP + FP)
OA = (TP + TN)/(TP + TN + FP + FN)
Kappa = (P0 − Pe)/(1 − Pe)
where P0 is calculated using the OA formula and Pe is calculated with the formula:
Pe = ((TP + FN)(TP + FP) + (FP + TN)(FN + TN))/(TP + TN + FP + FN)2
The RF classification was performed using the ‘caret’ package [50].

2.6. Statistical Analysis

We detected the impacts of species and canopy positions on the variation of the LFT dataset, LHR dataset and classification accuracies using a one-way analysis of variance (ANOVA). Then, we applied least-significant difference (LSD) post hoc tests to show the significant differences of each observed dataset among species and among canopy layers. All the statistical analyses were processed in the R software (version 4.2.1).

3. Results

3.1. Variation of Leaf Traits and Spectra

The leaf traits of the six species were significantly different (Figure 2). Specifically, LDMC of Castch, Flav of Machch, LCC of Crypch and LNC of Crypco were significantly higher than the other five species in corresponding leaf traits; Chl of Castfi and LCC of Castch were significantly lower than the remaining four species.
In terms of leaf hyperspectral reflectance (Figure 3), the variation of leaf spectra among the six tree species was observed in the spectral regions of 800–850 nm, 950–1100 nm, 1150–1350 nm, 1450–1550 nm, 1750–1850 nm and 1950–2100 nm.
Most leaf trait values showed an increasing trend from the lower to the upper layers, while the SLA value decreased as the canopy position increased (Figure 4). Specifically, LDMC of Schisu, EWT of Castch, Chl of Castch and Castfi, Flav of Schisu, Castch and Machch, and LPC of Crypch significantly increased along the vertical canopy. Several leaf traits (≥3) of Schisu and Castch varied significantly in the vertical canopy, while none of Crypco’s leaf traits changed significantly.
Additionally, the separability of leaf hyperspectral reflectance among the six tree species was gradually enhanced from the lower to the upper layers (Figure 5). The improvement of spectral separability mainly occurred in the near-infrared region of 800–1300 nm and the shortwave infrared region of 1500–1850 nm.

3.2. Comparison of Classification Accuracies

The pairwise difference in classification accuracies was significant among the three classification datasets (Table 2). The LFT + LHR dataset obtained the highest OA of 77.65% and Kappa of 0.73, while the LHR dataset achieved the lowest OA of 69.06% and Kappa of 0.63; the LFT dataset produced an OA of 74.26% and a Kappa of 0.69, which were higher than the LHR dataset.
Details of the classification accuracy for each tree species are shown in Table 3. In the three datasets of LFT, LHR and LFT + LHR, Castfi always achieved the highest accuracies, which were 84.07%, 86.21% and 89.00%, respectively. Crypco’s accuracies of 81.81%, 84.59% and 87.50%, respectively, which were generally higher than the other four species, ranked second. Meanwhile, Castch had the lowest accuracy among all species; it only recached 64.09%, 53.94% and 69.72% for the three datasets, respectively. Additionally, the classification accuracies of tree species were generally ranked in order of Castfi > Crypco > Machch > Crypch > Schisu > Castch. It is worth noting that Schisu’s accuracy fluctuated largely among different classification datasets.

3.3. Variation in Classification Accuracies along the Vertical Canopy

Overall, the combination of the upper, middle and lower (UML) layers achieved the highest OA and Kappa of all the classification datasets and outperformed the single layer significantly in three datasets (LFT, LHR and LFT + LHR, Figure 6). Moreover, the OA and Kappa were generally ranked in order of UML layer > upper layer > middle canopy > lower canopy.
For the individual tree species, the UML data generally showed the highest classification accuracy (Figure 7). Castch and Castfi’s accuracies increased significantly from the lower to the upper layers, while Crypch’s showed a significantly decreasing trend with the LFT dataset (Figure 7A). The accuracy of the LHR dataset (Figure 7B) and the LFT + LHR dataset (Figure 7C) showed a similar pattern; the accuracies of Castfi, Machch and Crypch all demonstrated a significantly increasing tendency from the lower canopy to the upper canopy.
In summary, Castfi’s accuracies maintained an increasing trend from the lower to upper layer in all classification datasets; Machch and Crypch’s accuracies generally increased with the increase in the vertical canopy.

4. Discussion

4.1. Performance Differences of Leaf Spectra and Traits in Classification

Our results showed that the combined LFT + LHR dataset achieved a higher accuracy compared to using the LFT and LHR separately. This result validated the assumption that adding characteristic plant functional traits into classification with reflectance aids the discrimination of tree species, which has a profound meaning for ecology and remote sensing research.
In recent years, the application of remote sensing spectrum in tree species classification has become increasingly enhanced, and the hyperspectral data have been widely used in tree classification because of their high resolution [17]. A number of feature selection methods, including continuous wavelet transformation [51], minimum noise fraction transformation [52] and some advanced classifiers, such as support vector machine [53] and convolutional neural network [54], were employed to improve classification efficiency. However, the phenomenon of “different objects with the same spectrum” is inevitable due to a complicated interaction between leaf and environmental background, which increases the error rate of tree species separation. Previous studies have proven that the spectral properties are largely dependent on biochemical traits, a taxonomical organization [33,55,56]. Several chemical traits have shown marked interspecific differences as well as morphological and physiological traits. Although leaf traits also have similar values among some species, the combination of leaf traits is different for any two species [57]. Asner and Martin [33] suggested that more than 90% of Amazon rainforest canopy species can be easily discriminated using 10 plant functional traits. Moreover, Shi et al. [18] improved the classification accuracy in a mixed temperate forest by adding simulated functional traits (EWT, LMA and chlorophyll) derived from leaf spectra. Our results directly showed the positive effect of filed-measured leaf traits in individual tree species classification. The defect and deficiency of spectral data can be supplemented by combining complementary leaf traits, especially in the aspect of biochemistry.
According to our results, we consider that it is possible to develop a methodology to scale up leaf traits to a larger scale, and improve the accuracy of tree species classification. Empirical and physical models are basic model inversion approaches for upscaling leaf traits to the canopy [39,58]. The former directly builds a statistical model between the leaf traits and the canopy spectra, which ignores some potential uncertainties [12,59,60]; the latter provides more robustness and transferability compared to the empirical models, but the relevant input parameters are complicated [61,62]. Therefore, it remains a challenge to estimate plant traits using remote sensing, which limits the application of the results from this study.

4.2. Impact of Vertical Canopy Position on Accuracies

Our results showed that the accuracy of tree species classification increased from the lower to upper canopy and was highest in the UML layer along the vertical canopy. In the evergreen broad-leaved forest of Dinghushan, light environment, biotic relationship and individual species characteristics are key to explaining the mechanism of the vertical canopy [47]. Light is an important factor that affects the growth status of trees in the canopy [63,64], especially in the later stage of forest succession [65]. Holmgren et al. [66] suggested that a low-light environment will reduce the interspecies competition and facilitate coexistence under conditions of sufficient moisture. Therefore, interspecific competition will intensify in the upper layer of the canopy, which is consistent with Li et al.’s [46] conclusions for Dinghushan. In the lower canopy, the weak-light, low-temperature and high-humidity facilitate most species to choose coexistence to maintain their own population, which increases the difficulty of identifying species. As the height of a tree increases, the environmental resources, such as light and heat, gradually increase, which further motivates the fierce interspecific competition. Subsequently, gradual intense competition facilitates the niche differentiation of tree species in the middle canopy, and peak at the upper canopy. Therefore, the classification accuracy presented an increasing trend with increases in the canopy. In addition, the UML data synthesized all the feature information of tree species in the canopy layer, which enabled us to generally achieve the highest accuracy for every tree species in each classification dataset.

4.3. Influence of Tree Species Growth Habits on Classification Accuracies

The six species can be divided according to their demand for sunlight intensity and for tree growth; light-demanding species include Schisu, Castch and Castfi, and shade-tolerant species include Machch, Crypch and Crypco. In our results, all three shade-tolerant species achieved relatively high accuracies, while the light-demanding species produced relatively poor results.
Tree species with different light adaptations have their own growth advantages in different successional stages of subtropical evergreen broad-leaved forests. At the beginning of forest succession, the light-demanding species participate as pioneers in the construction of the forest organization, quickly occupy the survival space and become the dominant species. With the succession of vegetation, the light-demanding species will gradually be replaced by the shade-tolerant species and lose the dominant position in the community [67]. A previous study showed that the subtropical evergreen broad-leaved forest in the Dinghushan Nature Reserve is in the later stage of succession [68], with tall and dense forests. Therefore, the primary strategy of shade-tolerant species, such as Machch, Crypch and Crypco, is to optimize resource allocation, which presented unique features and tended to be classified. Our results showed that the leaf traits of shade-tolerant species are relatively stable in the canopy, which indicated that the impact on species features from microenvironment heterogeneity was relatively slight. It supports shade-tolerant species being separated from a mixed stand and maintains a relatively high classification accuracy of them in different canopy layers. On the contrary, the light-demanding species under low-light stress must adjust their survival strategy to maintain their population. As a result, many leaf traits of Schisu and Castch varied drastically in the canopy, which led to a fluctuation in their accuracy in different canopy layers. It is worth noting that Castfi’s accuracy was the highest of all the species. According to Zhang et al. [69] and Yu et al. [70], Castfi has an intensively strong demand for sunlight in the growth process. We considered that Castfi’s accuracy was possibly correlated with its extremely poor adaptation to dense shade. As a result, Castfi had few changes in physiology and biochemistry that corresponded to shade. Although the response of Castfi is bad for its long-term survival, the persistent characters of the leaf contrarily increase its identification in the community.

5. Conclusions

We used a tower crane to collect leaf samples from three canopy layers in a subtropical evergreen broad-leaved forest and measured their leaf function traits and leaf hyperspectral reflectance as the input of classification datasets. Then, we compared the performance of the classification datasets and explored the influence of the vertical canopy on classification accuracy, drawing the following conclusions:
  • The combined LFT + LHR data achieved significantly higher accuracy compared to the LFT and LHR only datasets. Combining the leaf spectra and trait data provides a new site for the application of classification datasets at the leaf level. We proposed that leaf functional traits and other plant traits should be considered for tree species classification in the future.
  • The UML layer achieved the highest accuracy among the four canopy layers, and the accuracy increased from the lower canopy to the upper canopy. This can be attributed to the ample survival resources and the fierce interspecific competition in the upper canopy layer, which was correlated with the light gradient. This result enriches our understanding of the vertical structure in evergreen broad-leaved forests, and light gradient is key to explaining the interaction of species in a dense canopy.
  • Higher accuracies were produced by the shade-tolerant species (Machch, Crypch and Crypco) due to their stable growth strategy in the forest. By contrast, lower accuracies were yielded by the light-demanding species (Schisu and Castch) because of their low-light adaptations. We proposed that tree species discrimination from remotely sensed data should consider more ecological characteristics and information on species.

Author Contributions

Conceptualization, Y.Z., F.Y. and J.W.; methodology, Z.W., J.W., F.Y., J.L. and W.Y.; formal analysis, F.Y., Y.Z., J.W. and Z.W.; writing—original draft preparation, Y.Z.; writing—review and editing, Y.Z., F.Y. and J.W.; visualization, Y.Z.; supervision, J.L., W.Y. and Z.W.; funding acquisition, J.L., W.Y. and F.Y. All authors have read and agreed to the published version of the manuscript.

Funding

This study supported by Strategic Priority Research Program of the Chinese Academy of Sciences (XDB31030000), the National Natural Science Foundation of China (No. 41901060) and the NSFC-Guangdong Joint Foundation Key Project (U1901219).

Data Availability Statement

Not applicable.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Orka, H.O.; Dalponte, M.; Gobakken, T.; Naesset, E.; Ene, L.T. Characterizing forest species composition using multiple remote sensing data sources and inventory approaches. Scand. J. Forest Res. 2013, 28, 677–688. [Google Scholar] [CrossRef]
  2. Modzelewska, A.; Fassnacht, F.E.; Sterenczak, K. Tree species identification within an extensive forest area with diverse management regimes using airborne hyperspectral data. Int. J. Appl. Earth Obs. Geoinf. 2020, 84, 101960. [Google Scholar] [CrossRef]
  3. Orka, H.O.; Naesset, E.; Bollandsas, O.M. Classifying species of individual trees by intensity and structure features derived from airborne laser scanner data. Remote Sens. Environ. 2009, 113, 1163–1174. [Google Scholar] [CrossRef]
  4. Heinzel, J.; Koch, B. Investigating multiple data sources for tree species classification in temperate forest and use for single tree delineation. Int. J. Appl. Earth Obs. Geoinf. 2012, 18, 101–110. [Google Scholar] [CrossRef]
  5. Lucas, R.; Bunting, P.; Paterson, M.; Chisholm, L. Classification of Australian forest communities using aerial photography, CASI and HyMap data. Remote Sens. Environ. 2008, 112, 2088–2103. [Google Scholar] [CrossRef]
  6. Jones, T.G.; Coops, N.C.; Sharma, T. Assessing the utility of airborne hyperspectral and LiDAR data for species distribution mapping in the coastal Pacific Northwest, Canada. Remote Sens. Environ. 2010, 114, 2841–2852. [Google Scholar] [CrossRef]
  7. Zhao, F.; Guo, Q.H.; Kelly, M. Allometric equation choice impacts lidar-based forest biomass estimates: A case study from the Sierra National Forest, CA. Agric. For. Meteorol. 2012, 165, 64–72. [Google Scholar] [CrossRef]
  8. Henry, M.; Bombelli, A.; Trotta, C.; Alessandrini, A.; Birigazzi, L.; Sola, G.; Vieilledent, G.; Santenoise, P.; Longuetaud, F.; Valentini, R.; et al. GlobAllomeTree: International platform for tree allometric equations to support volume, biomass and carbon assessment. IForest 2013, 6, E1–E5. [Google Scholar] [CrossRef] [Green Version]
  9. Dube, T.; Gara, T.W.; Mutanga, O.; Sibanda, M.; Shoko, C.; Murwira, A.; Masocha, M.; Ndaimani, H.; Hatendi, C.M. Estimating forest standing biomass in savanna woodlands as an indicator of forest productivity using the new generation WorldView-2 sensor. Geocarto Int. 2018, 33, 178–188. [Google Scholar] [CrossRef]
  10. Yao, W.; Krzystek, P.; Heurich, M. Tree species classification and estimation of stem volume and DBH based on single tree extraction by exploiting airborne full-waveform LiDAR data. Remote Sens. Environ. 2012, 123, 368–380. [Google Scholar] [CrossRef]
  11. Asner, G.P.; Martin, R.E.; Tupayachi, R.; Anderson, C.B.; Sinca, F.; Carranza-Jimenez, L.; Martinez, P. Amazonian functional diversity from forest canopy chemical assembly. PNAS 2014, 111, 5604–5609. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  12. Wang, Z.H.; Wang, T.J.; Darvishzadeh, R.; Skidmore, A.K.; Jones, S.; Suarez, L.; Woodgate, W.; Heiden, U.; Heurich, M.; Hearne, J. Vegetation Indices for Mapping Canopy Foliar Nitrogen in a Mixed Temperate Forest. Remote Sens. 2016, 8, 491. [Google Scholar] [CrossRef] [Green Version]
  13. Gokkaya, K.; Thomas, V.; Noland, T.; McCaughey, H.; Morrison, I.; Treitz, P. Mapping continuous forest type variation by means of correlating remotely sensed metrics to canopy N:P ratio in a boreal mixedwood forest. Appl. Veg. Sci. 2015, 18, 143–157. [Google Scholar] [CrossRef]
  14. Yin, D.M.; Wang, L. How to assess the accuracy of the individual tree-based forest inventory derived from remotely sensed data: A review. Int. J. Remote Sens. 2016, 37, 4521–4553. [Google Scholar] [CrossRef]
  15. Liu, L.X.; Coops, N.C.; Aven, N.W.; Pang, Y. Mapping urban tree species using integrated airborne hyperspectral and LiDAR remote sensing data. Remote Sens. Environ. 2017, 200, 170–182. [Google Scholar] [CrossRef]
  16. Shang, X.; Chisholm, L.A. Classification of Australian Native Forest Species Using Hyperspectral Remote Sensing and Machine-Learning Classification Algorithms. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2014, 7, 2481–2489. [Google Scholar] [CrossRef]
  17. Fassnacht, F.E.; Latifi, H.; Sterenczak, K.; Modzelewska, A.; Lefsky, M.; Waser, L.T.; Straub, C.; Ghosh, A. Review of studies on tree species classification from remotely sensed data. Remote Sens. Environ. 2016, 186, 64–87. [Google Scholar] [CrossRef]
  18. Shi, Y.F.; Skidmore, A.K.; Wang, T.J.; Holzwarth, S.; Heiden, U.; Pinnel, N.; Zhu, X.; Heurich, M. Tree species classification using plant functional traits from LiDAR and hyperspectral data. Int. J. Appl. Earth Obs. Geoinf. 2018, 73, 207–219. [Google Scholar] [CrossRef]
  19. Clark, M.L.; Roberts, D.A.; Clark, D.B. Hyperspectral discrimination of tropical rain forest tree species at leaf to crown scales. Remote Sens. Environ. 2005, 96, 375–398. [Google Scholar] [CrossRef]
  20. Hesketh, M.; Sanchez-Azofeifa, G.A. The effect of seasonal spectral variation on species classification in the Panamanian tropical forest. Remote Sens. Environ. 2012, 118, 73–82. [Google Scholar] [CrossRef]
  21. Guzman, Q.J.A.; Rivard, B.; Sanchez-Azofeifa, G.A. Discrimination of liana and tree leaves from a Neotropical Dry Forest using visible-near infrared and longwave infrared reflectance spectra. Remote Sens. Environ. 2018, 219, 135–144. [Google Scholar] [CrossRef]
  22. Liao, W.Z.; Van Coillie, F.; Gao, L.R.; Li, L.W.; Zhang, B.; Chanussot, J. Deep Learning for Fusion of APEX Hyperspectral and Full-Waveform LiDAR Remote Sensing Data for Tree Species Mapping. IEEE Access 2018, 6, 68716–68729. [Google Scholar] [CrossRef]
  23. Puttonen, E.; Suomalainen, J.; Hakala, T.; Raikkonen, E.; Kaartinen, H.; Kaasalainen, S.; Litkey, P. Tree species classification from fused active hyperspectral reflectance and LIDAR measurements. For. Ecol. Manag. 2010, 260, 1843–1852. [Google Scholar] [CrossRef]
  24. Clark, M.L.; Roberts, D.A. Species-Level Differences in Hyperspectral Metrics among Tropical Rainforest Trees as Determined by a Tree-Based Classifier. Remote Sens. 2012, 4, 1820–1855. [Google Scholar] [CrossRef] [Green Version]
  25. van Deventer, H.; Cho, M.A.; Mutanga, O. Improving the classification of six evergreen subtropical tree species with multi-season data from leaf spectra simulated to WorldView-2 and RapidEye. Int. J. Remote Sens. 2017, 38, 4804–4830. [Google Scholar] [CrossRef]
  26. Cho, M.A.; Debba, P.; Mathieu, R.; Naidoo, L.; van Aardt, J.; Asner, G.P. Improving Discrimination of Savanna Tree Species Through a Multiple-Endmember Spectral Angle Mapper Approach: Canopy-Level Analysis. IEEE Trans. Geosci. Electron. 2010, 48, 4133–4142. [Google Scholar] [CrossRef]
  27. Asner, G.P.; Martin, R.E. Airborne spectranomics: Mapping canopy chemical and taxonomic diversity in tropical forests. Front. Ecol. Environ. 2009, 7, 269–276. [Google Scholar] [CrossRef] [Green Version]
  28. Lefcheck, J.S.; Bastazini, V.A.G.; Griffin, J.N. Choosing and using multiple traits in functional diversity research. Environ. Conserv. 2015, 42, 104–107. [Google Scholar] [CrossRef] [Green Version]
  29. Violle, C.; Navas, M.L.; Vile, D.; Kazakou, E.; Fortunel, C.; Hummel, I.; Garnier, E. Let the concept of trait be functional! Oikos 2007, 116, 882–892. [Google Scholar] [CrossRef]
  30. Reich, P.B. The world-wide ‘fast-slow’ plant economics spectrum: A traits manifesto. J. Ecol. 2014, 102, 275–301. [Google Scholar] [CrossRef]
  31. Albert, C.H.; Thuiller, W.; Yoccoz, N.G.; Douzet, R.; Aubert, S.; Lavorel, S. A multi-trait approach reveals the structure and the relative importance of intra- vs. interspecific variability in plant traits. Funct. Ecol. 2010, 24, 1192–1201. [Google Scholar] [CrossRef]
  32. Subedi, S.C.; Ross, M.S.; Sah, J.P.; Redwine, J.; Baraloto, C. Trait-based community assembly pattern along a forest succession gradient in a seasonally dry tropical forest. Ecosphere 2019, 10, e02719. [Google Scholar] [CrossRef] [Green Version]
  33. Asner, G.P.; Martin, R.E. Canopy phylogenetic, chemical and spectral assembly in a lowland Amazonian forest. New Phytol. 2011, 189, 999–1012. [Google Scholar] [CrossRef] [PubMed]
  34. Ewald, M.; Aerts, R.; Lenoir, J.; Fassnacht, F.E.; Nicolas, M.; Skowronek, S.; Piat, J.; Honnay, O.; Garzon-Lopez, C.X.; Feilhauer, H.; et al. LiDAR derived forest structure data improves predictions of canopy N and P concentrations from imaging spectroscopy. Remote Sens. Environ. 2018, 211, 13–25. [Google Scholar] [CrossRef]
  35. Asner, G.P.; Martin, R.E.; Tupayachi, R.; Emerson, R.; Martinez, P.; Sinca, F.; Powell, G.V.N.; Wright, S.J.; Lugo, A.E. Taxonomy and remote sensing of leaf mass per area (LMA) in humid tropical forests. Ecol. Appl. 2011, 21, 85–98. [Google Scholar] [CrossRef] [PubMed]
  36. Wang, Z.H.; Skidmore, A.K.; Darvishzadeh, R.; Heiden, U.; Heurich, M.; Wang, T.J. Leaf Nitrogen Content Indirectly Estimated by Leaf Traits Derived from the PROSPECT Model. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2015, 8, 3172–3182. [Google Scholar] [CrossRef]
  37. de Jong, S.M.; Addink, E.A.; Doelman, J.C. Detecting leaf-water content in Mediterranean trees using high-resolution spectrometry. Int. J. Appl. Earth Obs. Geoinf. 2014, 27, 128–136. [Google Scholar] [CrossRef]
  38. Auger, S.; Shipley, B. Inter-specific and intra-specific trait variation along short environmental gradients in an old-growth temperate forest. J. Veg. Sci. 2013, 24, 419–428. [Google Scholar] [CrossRef]
  39. Ali, A.M.; Skidmore, A.K.; Darvishzadeh, R.; van Duren, I.; Holzwarth, S.; Mueller, J. Retrieval of forest leaf functional traits from HySpex imagery using radiative transfer models and continuous wavelet analysis. ISPRS J. Photogramm. Remote Sens. 2016, 122, 68–80. [Google Scholar] [CrossRef] [Green Version]
  40. Helfenstein, I.S.; Schneider, F.D.; Schaepman, M.E.; Morsdorf, F. Assessing biodiversity from space: Impact of spatial and spectral resolution on trait-based functional diversity. Remote Sens. Environ. 2022, 275, 113024. [Google Scholar] [CrossRef]
  41. Yu, F.Y.; Gara, T.W.; Lian, J.Y.; Ye, W.H.; Shen, J.; Wang, T.J.; Wu, Z.F.; Wang, J.J. Understanding the Impact of Vertical Canopy Position on Leaf Spectra and Traits in an Evergreen Broadleaved Forest. Remote Sens. 2021, 13, 5057. [Google Scholar] [CrossRef]
  42. Shi, Y.F.; Wang, T.J.; Skidmore, A.K.; Heurich, M. Important LiDAR metrics for discriminating forest tree species in Central Europe. ISPRS J. Photogramm. Remote Sens. 2018, 137, 163–174. [Google Scholar] [CrossRef]
  43. Sankey, T.; Donager, J.; McVay, J.; Sankey, J.B. UAV lidar and hyperspectral fusion for forest monitoring in the southwestern USA. Remote Sens. Environ. 2017, 195, 30–43. [Google Scholar] [CrossRef]
  44. Dalponte, M.; Bruzzone, L.; Gianelle, D. Tree species classification in the Southern Alps based on the fusion of very high geometrical resolution multispectral/hyperspectral images and LiDAR data. Remote Sens. Environ. 2012, 123, 258–270. [Google Scholar] [CrossRef]
  45. Alonzo, M.; Bookhagen, B.; Roberts, D.A. Urban tree species mapping using hyperspectral and lidar data fusion. Remote Sens. Environ. 2014, 148, 70–83. [Google Scholar] [CrossRef]
  46. Li, Y.P.; Ni, Y.L.; Xu, H.; Lian, J.Y.; Ye, W.H. Relationship between variation of plant functional traits and individual growth at different vertical layers in a subtropical evergreen broad-leaved forest of Dinghushan. Biodivers. Sci. 2021, 29, 1186–1197. [Google Scholar] [CrossRef]
  47. Gui, X.J.; Lian, J.Y.; Zhang, R.Y.; Li, Y.P.; Shen, H.; Ni, Y.L.; Ye, W.H. Vertical structure and its biodiversity in a subtropical evergreen broadleaved forest at Dinghushan in Guangdong Province, China. Biodivers. Sci. 2019, 27, 619–629. [Google Scholar]
  48. Cornelissen, J.H.C.; Lavorel, S.; Garnier, E.; Diaz, S.; Buchmann, N.; Gurvich, D.E.; Reich, P.B.; ter Steege, H.; Morgan, H.D.; van der Heijden, M.G.A.; et al. A handbook of protocols for standardised and easy measurement of plant functional traits worldwide. Aust. J. Bot. 2003, 51, 335–380. [Google Scholar] [CrossRef] [Green Version]
  49. Cutler, D.R.; Edwards, T.C.; Beard, K.H.; Cutler, A.; Hess, K.T. Random forests for classification in ecology. Ecology 2007, 88, 2783–2792. [Google Scholar] [CrossRef]
  50. Kuhn, M. Building Predictive Models in R Using the caret Package. J. Stat. Softw. 2008, 28, 1–26. [Google Scholar] [CrossRef] [Green Version]
  51. Harrison, D.; Rivard, B.; Sanchez-Azofeifa, A. Classification of tree species based on longwave hyperspectral data from leaves, a case study for a tropical dry forest. Int. J. Appl. Earth Obs. Geoinf. 2018, 66, 93–105. [Google Scholar] [CrossRef]
  52. Fassnacht, F.E.; Neumann, C.; Forster, M.; Buddenbaum, H.; Ghosh, A.; Clasen, A.; Joshi, P.K.; Koch, B. Comparison of Feature Reduction Algorithms for Classifying Tree Species with Hyperspectral Data on Three Central European Test Sites. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2014, 7, 2547–2561. [Google Scholar] [CrossRef]
  53. Wietecha, M.; Jelowicki, L.; Mitelsztedt, K.; Miscicki, S.; Sterenczak, K. The capability of species-related forest stand characteristics determination with the use of hyperspectral data. Remote Sens. Environ. 2019, 231, 111232. [Google Scholar] [CrossRef]
  54. Fricker, G.A.; Ventura, J.D.; Wolf, J.A.; North, M.P.; Davis, F.W.; Franklin, J. A Convolutional Neural Network Classifier Identifies Tree Species in Mixed-Conifer Forest from Hyperspectral Imagery. Remote Sens. 2019, 11, 2326. [Google Scholar] [CrossRef] [Green Version]
  55. Feret, J.B.; Asner, G.P. Spectroscopic classification of tropical forest species using radiative transfer modeling. Remote Sens. Environ. 2011, 115, 2415–2422. [Google Scholar] [CrossRef]
  56. Castro-Esau, K.L.; Sanchez-Azofeifa, G.A.; Rivard, B.; Wright, S.J.; Quesada, M. Variability in leaf optical properties of Mesoamerican trees and the potential for species classification. Am. J. Bot. 2006, 93, 517–530. [Google Scholar] [CrossRef]
  57. Zhao, Y.J.; Zeng, Y.; Zhao, D.; Wu, B.F.; Zhao, Q.J. The Optimal Leaf Biochemical Selection for Mapping Species Diversity Based on Imaging Spectroscopy. Remote Sens. 2016, 8, 216. [Google Scholar] [CrossRef] [Green Version]
  58. Gara, T.W.; Skidmore, A.K.; Darvishzadeh, R.; Wang, T.J. Leaf to canopy upscaling approach affects the estimation of canopy traits. GIsci. Remote Sens. 2019, 56, 554–575. [Google Scholar] [CrossRef]
  59. Knyazikhin, Y.; Schull, M.A.; Stenberg, P.; Mottus, M.; Rautiainen, M.; Yang, Y.; Marshak, A.; Carmona, P.L.; Kaufmann, R.K.; Lewis, P.; et al. Hyperspectral remote sensing of foliar nitrogen content. PNAS 2013, 110, E185–E192. [Google Scholar] [CrossRef] [Green Version]
  60. Wang, Z.; Skidmore, A.K.; Darvishzadeh, R.; Wang, T. Mapping forest canopy nitrogen content by inversion of coupled leaf-canopy radiative transfer models from airborne hyperspectral imagery. Agric. For. Meteorol. 2018, 253, 247–260. [Google Scholar] [CrossRef]
  61. Malenovsky, Z.; Homolova, L.; Lukes, P.; Buddenbaum, H.; Verrelst, J.; Alonso, L.; Schaepman, M.E.; Lauret, N.; Gastellu-Etchegorry, J.P. Variability and Uncertainty Challenges in Scaling Imaging Spectroscopy Retrievals and Validations from Leaves Up to Vegetation Canopies. Surv. Geophys. 2019, 40, 631–656. [Google Scholar] [CrossRef] [PubMed]
  62. Morsdorf, F.; Nichol, C.; Malthus, T.; Woodhouse, I.H. Assessing forest structural and physiological information content of multi-spectral LiDAR waveforms by radiative transfer modelling. Remote Sens. Environ. 2009, 113, 2152–2163. [Google Scholar] [CrossRef] [Green Version]
  63. D’Andrea, R.; Guittar, J.; O’Dwyer, J.P.; Figueroa, H.; Wright, S.J.; Condit, R.; Ostling, A. Counting niches: Abundance-by-trait patterns reveal niche partitioning in a Neotropical forest. Ecology 2020, 101, e03019. [Google Scholar] [CrossRef] [PubMed]
  64. Weerasinghe, L.K.; Creek, D.; Crous, K.Y.; Xiang, S.; Liddell, M.J.; Turnbull, M.H.; Atkin, O.K. Canopy position affects the relationships between leaf respiration and associated traits in a tropical rainforest in Far North Queensland. Tree Physiol. 2014, 34, 564–584. [Google Scholar] [CrossRef] [Green Version]
  65. Lebrija-Trejos, E.; Perez-Garcia, E.A.; Meave, J.A.; Bongers, F.; Poorter, L. Functional traits and environmental filtering drive community assembly in a species-rich tropical system. Ecology 2010, 91, 386–398. [Google Scholar] [CrossRef] [Green Version]
  66. Holmgren, M.; Scheffer, M.; Huston, M.A. the interplay of facilitation and competition in plant communities. Ecology 1997, 78, 1966–1975. [Google Scholar] [CrossRef]
  67. Liu, W.D.; Su, J.R. Successional leaf traits of monsoon evergreen broad-leaved forest, Southwest China. IForest 2017, 10, 391–396. [Google Scholar] [CrossRef]
  68. Zou, S.; Zhou, G.Y.; Zhang, Q.M.; Xu, S.; Xiong, X.; Xia, Y.J.; Liu, S.Z.; Meng, Z.; Chu, G.W. Long-term (1992–2015) dynamics of community composition and structure in a monsoon evergreen broad-leaved forest in Dinghushan Biosphere Reserve. Chin. J. Plant Ecol. 2018, 42, 442–452. [Google Scholar]
  69. Zhang, J.Z.; Lin, G.Z.; Lin, Z.F.; Peng, C.L. Response of Photosynthesis to Growth Light Intensity in Some South Subtropical Woody Plants. J. Trop. Subtrop. Bot. 2005, 13, 413–418. [Google Scholar]
  70. Yu, Z.C.; Liu, X.T.; Zheng, X.T.; Huang, X.D.; Peng, C.L. Relationship between the Accumulation Ability of Photoprotective Substances and the Photosynthetic Capacity in Leaves of Four Woody Plants Grown under Two Light Intensities. J. South China Normal Univ. (Nat. Sci. Ed.). 2019, 51, 62–69. [Google Scholar]
Remotesensing 14 06227 g001 550
Figure 1. Location of the tower crane plot in Dinghushan Nature Reserve, China.
Figure 1. Location of the tower crane plot in Dinghushan Nature Reserve, China.
Remotesensing 14 06227 g001
Remotesensing 14 06227 g002 550
Figure 2. Interspecific variation in eight leaf functional traits (LDMC, SLA, EWT, Chl, Flav, LCC, LNC and LPC). Different letters (a, b, c, etc.) indicate a significant difference (p < 0.05) between tree species in leaf trait values.
Figure 2. Interspecific variation in eight leaf functional traits (LDMC, SLA, EWT, Chl, Flav, LCC, LNC and LPC). Different letters (a, b, c, etc.) indicate a significant difference (p < 0.05) between tree species in leaf trait values.
Remotesensing 14 06227 g002
Remotesensing 14 06227 g003 550
Figure 3. Interspecific variation in leaf hyperspectral reflectance.
Figure 3. Interspecific variation in leaf hyperspectral reflectance.
Remotesensing 14 06227 g003
Remotesensing 14 06227 g004 550
Figure 4. Variation of eight leaf traits in each tree species across three canopy layers (lower, middle and upper). The symbols ****, ***, **, and * represent significance at the 0.0001, 0.001, 0.01 and 0.05 levels, respectively. The symbol ns indicates no significant difference (p > 0.05).
Figure 4. Variation of eight leaf traits in each tree species across three canopy layers (lower, middle and upper). The symbols ****, ***, **, and * represent significance at the 0.0001, 0.001, 0.01 and 0.05 levels, respectively. The symbol ns indicates no significant difference (p > 0.05).
Remotesensing 14 06227 g004
Remotesensing 14 06227 g005 550
Figure 5. Variation of leaf hyperspectral reflectance in each tree species across three canopy layers.
Figure 5. Variation of leaf hyperspectral reflectance in each tree species across three canopy layers.
Remotesensing 14 06227 g005
Remotesensing 14 06227 g006 550
Figure 6. Variation of OA and Kappa in different canopy layers based on (A) the LFT dataset, (B) LHR dataset and (C) LFT + LHR dataset. The UML is the pooled data of the upper, middle and lower layers.
Figure 6. Variation of OA and Kappa in different canopy layers based on (A) the LFT dataset, (B) LHR dataset and (C) LFT + LHR dataset. The UML is the pooled data of the upper, middle and lower layers.
Remotesensing 14 06227 g006
Remotesensing 14 06227 g007 550
Figure 7. Variation of classification accuracies in different canopy layers of individual trees based on (A) the LFT dataset, (B) LHR dataset and (C) LFT + LHR dataset. The UML is the pooled dataset of the upper, middle and lower layers. The symbols **** and *** represent significance at the 0.0001 and 0.001 levels, respectively. The symbol ns indicates no significant difference (p > 0.05).
Figure 7. Variation of classification accuracies in different canopy layers of individual trees based on (A) the LFT dataset, (B) LHR dataset and (C) LFT + LHR dataset. The UML is the pooled dataset of the upper, middle and lower layers. The symbols **** and *** represent significance at the 0.0001 and 0.001 levels, respectively. The symbol ns indicates no significant difference (p > 0.05).
Remotesensing 14 06227 g007
Table
Table 1. Details of the six dominant tree species in the study.
Table 1. Details of the six dominant tree species in the study.
SpeciesAbbreviationSample Size
Schima superbaSchisu16
Castanopsis chinensisCastch12
Castanopsis fissaCastfi15
Machilus chinensisMachch20
Cryptocarya chinensisCrypch13
Cryptocarya concinnaCrypco13
Table
Table 2. OA (%) and Kappa based on different classification datasets.
Table 2. OA (%) and Kappa based on different classification datasets.
Classification DatasetsOA (%)Kappa
LFT dataset74.26 b0.69 b
LHR dataset69.06 c0.63 c
LFT + LHR dataset77.65 a0.73 a
Different letters (a–c) represent a significant difference (p < 0.05) between the classification datasets.
Table
Table 3. The classification accuracies (%) of tree species based on different datasets.
Table 3. The classification accuracies (%) of tree species based on different datasets.
Classification DatasetsSchisuCastchCastfiMachchCrypchCrypco
LFT82.65 a64.09 c84.07 a72.80 b71.87 b81.81 a
LHR60.73 d53.94 e86.21 a69.74 b66.28 c84.59 a
LFT + LHR73.95 d69.72 e89.00 a78.85 b76.26 c87.50 a
Different letters (a–e) represent a significant difference (p < 0.05) between the classification datasets.
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Zhang, Y.; Wang, J.; Wu, Z.; Lian, J.; Ye, W.; Yu, F. Tree Species Classification Using Plant Functional Traits and Leaf Spectral Properties along the Vertical Canopy Position. Remote Sens. 2022, 14, 6227. https://doi.org/10.3390/rs14246227

AMA Style

Zhang Y, Wang J, Wu Z, Lian J, Ye W, Yu F. Tree Species Classification Using Plant Functional Traits and Leaf Spectral Properties along the Vertical Canopy Position. Remote Sensing. 2022; 14(24):6227. https://doi.org/10.3390/rs14246227

Chicago/Turabian Style

Zhang, Yicen, Junjie Wang, Zhifeng Wu, Juyu Lian, Wanhui Ye, and Fangyuan Yu. 2022. "Tree Species Classification Using Plant Functional Traits and Leaf Spectral Properties along the Vertical Canopy Position" Remote Sensing 14, no. 24: 6227. https://doi.org/10.3390/rs14246227

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop