Spectral diversity is a biodiversity dimension based on electromagnetic radiation spectrum reflected by plant, showing the variation of spectral reflective ratio in different bands among interspecific and intraspecific plant individuals. Spectral diversity has become an important technique for plant diversity monitoring and assessment since the differences of spectral reflectance can comprehensively indicate the differences of biochemical components and morphological and structural characteristics among plants. Here we introduce the concept of spectral diversity and its ecological significance, compare the technical advantages and disadvantages among multiple sources and platforms producing spectral data, summarize the monitoring and evaluation methodologies of plant diversity based on the applications of spectral diversity, and discuss the ability of spectral diversity to integrate different biodiversity dimensions and the prospect of the application of spectral diversity in biodiversity research. Spectral diversity will serve the monitoring and assessment of plant diversity at multiple spatial scales, especially combined with near-ground remote sensing based on unmanned aerial vehicle technology, can achieve fine-scale monitoring and assessment of plant diversity, and thus has broad application prospects in biodiversity conservation and management.

Plant functional traits are the measurable characteristics that indicates plant adaptation to the environment, and understanding the patterns of certain characteristics, and their drivers is an essential component of plant ecology and earth system modeling research. Traditional field-based approaches for characterizing plant functional traits are time-consuming, labor-intensive and expensive, and usually focus on the traits of peak growing season and dominant species, making the scaling extension and spatiotemporal coverage of plant functional traits a great challenge. In contrast, newly emerging multi-scale hyperspectral remote sensing techniques potentially provide new avenues to easily identify and characterize functional traits. Here we first overview the principles and brief history of hyperspectral remote sensing technology for plant functional traits monitoring. Then, we detailed the principal methods for modelling the spectral-trait relationships, including empirical and semi-empirical statistical methods and inversion methods relying on physical-based modelling, among which the statistical partial least squares regression is widely used. We then used case studies to demonstrate the application while illustrating the remaining problems of plant functional traits monitoring using the hyperspectral remote sensing techniques respectively at leaf, community and landscape scales. Finally, we highlight four important future directions to advance hyperspectral remote sensing of plant functional traits, including: 1) exploring the generalizability and underlying mechanisms of spectral-trait modelling; 2) developing novel, transparent methodology that scales the spectral-trait relationships from leaf, canopy to satellite levels; 3) elucidating the pattern and drivers of remotely sensed plant functional traits and diversity across various spatiotemporal scales; and 4) investigating the linkage among environment, plant functional diversity, biodiversity and ecosystem functioning.

Recent advances in solar-induced chlorophyll fluorescence (SIF), which is a complement to optical remote sensing based on greenness observation, have made it possible to monitor the photosynthesis of plants in terrestrial ecosystems using state-of-the-art technologies. With the rapid development of tower-based, unmanned aerial vehicle (UAV), airborne and space-borne SIF observation technology and improving understanding of SIF mechanism, SIF is providing essential data support and mechanism understanding for the estimation of biological traits and gross primary production of terrestrial ecosystem, early detection of abiotic stress, extraction of photosynthetic phenology and monitoring of transpiration. In this review, we first introduce the fundamental theory, the observation systems and technologies and the retrieval method of SIF. Then, we review the applications of SIF in terrestrial ecosystem monitoring. Finally, we propose a roadmap of activities to facilitate future directions and discuss critical emerging applications of SIF in terrestrial ecosystem monitoring that can benefit from cross-disciplinary expertise.

Solar radiation is fundamental to the maintenance and development of forest ecosystem functions and services. Therefore, modeling the radiation transfer process in forest is of great significance for understanding forest ecosystem processes. In recent years, the rapid development of three-dimensional radiative transfer models makes it possible to accurately simulate the distribution and dynamics of radiation within forest canopies. In order to better understand three-dimensional radiative transfer models and make them better serve forest ecosystem research, we review the principles, applications and future prospects of these models. Firstly, common principles of three-dimensional radiative transfer models such as radiosity and ray tracing are briefly introduced, and then the applications of three-dimensional radiative transfer models in forest ecosystem research are summarized. Finally, future opportunities of integrating multiple datasets and models to better facilitate forest ecosystem research, such as model coupling and making various models easier to use, are discussed. With the accumulation of ecological big data and improvement of ecosystem progress models, three-dimensional radiative transfer models will play a more important role in theoretical research and practices of forest ecology in the future.

The protected areas are crucial to the maintenance of national ecological security and improvement of biodiversity conservation. Application of real-time, high-frequency and multi-scale ecological monitoring in the protected areas serves an effective means for identifying the dynamics of ecosystem, which is key to the implement of ecosystem health management in the protected areas. However, due to the lack of a unified ecosystem monitoring and research network and the corresponding indicator system of the protected areas in China, the composition and dynamics of ecosystem in many protected areas remains unclear, which can dimmish the ability to cope with emerging issues of biodiversity conservation. Lack of the data obtained from the ecological monitoring network can also hamper the evaluation of ecosystem health status and conservation effectiveness of the protected areas on a national scale. As such, it is necessary to construct a national scale monitoring and research network for the composition and dynamics of ecosystem in the protected areas, as well as a scientific, systematic and normative indicator system for this monitoring network. By addressing the aims and objectives of biodiversity and ecosystem monitoring in the protected areas and with reference to the indicator systems of existing ecological monitoring networks both in China and abroad, this study summarized the basic principles of establishing the indicator system and the selection of indicators. Accordingly, an indicator system for the ecosystem monitoring network of the protected areas was established and applied to 6 national nature reserves for demonstration. The established indicator system consists of 30 indicators to comprehensively monitor changes in the 6 key elements that compose an ecosystem, which is habitat, biota, meteorology, soil, atmospheric and water environment, and landscape. The indicator system was effectively applied to monitor the long-term and dynamic changes in the status and evolution of ecosystem components and structures in different ecosystem types of protected areas including forest, grassland, wetland and desert. The normalized and standardized data achieved from the established monitoring network can further be used for the evaluation of conservation effectiveness and healthy management of the protected areas.

Aims The relationship between biodiversity and ecosystem function is an important ecological issue that is increasingly receiving global attention. Plant functional diversity, as one of the most important components of biodiversity, is directly linked to ecosystem functions. Traditional in-situ monitoring of grassland plant functional diversity is not only time-consuming and laborious, but also difficult to expand to large-scale research due to the limitations of time and space. The development of remote sensing technology provides an economical and effective means for assessing the grassland functional diversity over large areas. We estimated functional diversity and aboveground biomass based on Sentinel-2 satellite images and field data across the meadow steppe in the Ulgai Management Area of Xilin Gol League in Nei Mongol.

Methods We selected 46 spectral feature variables from the Sentinel-2 satellite imagery in the study area. Next, three methods, including stepwise regression, partial least squares regression (PLSR), and random forest regression (RFR) were applied to retrieve the grassland functional richness (FRic), functional evenness (FEve) and functional divergence (FDiv). Finally, the grassland aboveground biomass was also estimated using PLSR method, and the relationships between remotely sensed grassland functional diversity and grassland aboveground biomass were analyzed.

Important findings Our results showed that: (1) Band 11, optimized soil adjusted vegetation index (OSAVI), water band index (WBI) were the most important predictor of FRic; Band 6, Band 10, Band 12, carotenoid reflectance index 1 (CRI1), double-peak optical index (D), normalized difference index 45 (NDI45) were significantly related to FEve; and Band 5, Band 9, Band10, Band11, weighted difference vegetation index (WDVI), convex hull area played a critical role in predicting FDiv. (2) Based on 10-fold cross-validation, the retrieval accuracies of FRic and FEve estimated by stepwise regression were much higher than that of the other two regression methods, with R² of 0.52 and 0.44, respectively. However, the FDiv was best estimated by PLSR (R² = 0.61). (3) Grassland aboveground biomass was estimated with an accuracy of R² = 0.61, and FRic was the best indicator of aboveground biomass (R² = 0.40), followed by FDiv (R² = 0.28) and FEve (R² = 0.27). Our findings indicated the ability of Sentinel-2 satellite images to estimate grassland plant functional diversity, providing reference and basis for grassland plant functional diversity estimation at a large regional scale.

Aims Plant diversity monitoring is the basis of biodiversity assessment and developing conservation policy. Traditional forest plant diversity monitoring is mainly based on field surveys, which is difficult to quickly obtain the spatial distribution and dynamic change of forest plant diversity. The development of remote sensing technology provides an important tool for assessing forest plant diversity at the regional scale. In this study, we explored two methods of forest plant diversity estimation based on Sentinel-2A satellite images and field data in three selected national nature reserves (Liangshui, Fenglin, and Hunchun).

Methods We used two methods to estimate forest plant diversity: (1) Direct estimation based on spectral diversity at the pixel and cluster scales, respectively; (2) Indirect estimation based on random forest regression. The spectral diversity was calculated based on the coefficient of variation and convex hull area at the pixel scale, respectively. K-means clustering method was used for cluster analysis to calculate the spectral diversity between clusters. For the indirect estimation, we used 10-fold cross validation to select characteristic variables for later diversity calculation.

Important findings Our results showed that: (1) At the pixel scale, the estimation accuracy of Shannon-Wiener diversity index based on convex hull area (R²= 0.74) was better than that of coefficient of variation (R²= 0.60); (2) The pixel-based estimation accuracy of Shannon-Wiener diversity index outperformed clustering basis (R²= 0.59); (3) Based on six feature variables, the Shannon-Wiener diversity index was best estimated using the random forest regression algorithm (R²= 0.79); (4) Both the Simpson diversity index and species richness could not be accurately estimated by the above methods. Our findings indicate the capability of Sentinel-2A satellite images to estimate the Shannon-Wiener diversity index, providing reference and basis for forest plant diversity estimation at a large scale.

Aims With increasing anthropogenic activities in coastal areas, human disturbances have been identified as major causes of the decline of coastal mangroves and undemine the sustainable development. Monitoring the spatial-temporal dynamics of typical human activities in mangrove ecosystems and adjacent areas is critical in conservation and restoration of local mangrove ecosystems.

Methods We proposed an object-oriented machine learning method based on seasonal water fluctuations, using Landsat satellite imagery on Google Earth Engine platform. Inundation frequency was incorporated as a classification feature to obtain the spatial pattern of aquaculture ponds, which is concerned as the key driver of degradation and losses of mangroves. We revealed the dynamics of aquaculture ponds at a 30 m-resolution between 1990 and 2020 in China’s coastal regions with mangrove community detected, including Guangdong, Fujian, Zhejiang, Taiwan, Guangxi, and Hainan.

Important findings The total area of coastal aquaculture ponds in 1990 was about 2 963 km², which increased to 5 200 km² in 2000 and 5 377 km² in 2010, and then decreased to 4 805 km² in 2020. The maximum appeared between 2010 and 2020, but there was a significant regional variation in the changing pattern and peaking time of coastal aquaculture ponds. Coastal aquaculture ponds were mainly concentrated in the region of 21°-24° N (Guangdong and Guangxi). The spatial pattern of mangroves was shown as a staggered arrangement to that of aquaculture ponds. Our results also indicate a symbiotic relationship between aquaculture ponds and mangroves at latitude 21°-22° N, where a large number of mangroves grow along the edges of aquaculture ponds. This special distribution of mangroves and aquaculture ponds leads to a high level of interconnections between these two ecosystems, which can be recognized as the typical areas in exploring the impacts of human activities on mangrove ecosystems. The conversion of mangroves to aquaculture ponds was the primary cause of mangrove loss, which led to the extreme fragmentation and aggregation of mangrove patches in different areas. Our research on the spatial-temporal pattern of coastal aquaculture ponds provides an accurate dataset to assess the impacts of increasing human activities on mangrove ecosystems, and may contribute to the identification of priority restoration area.

Aims Aboveground biomass (AGB) is one of the most important factors affecting grassland ecosystem function and is commonly measured in grassland research. AGB is often measured using the harvest method, which can cause great disturbance to plant communities, especially for those long-term monitoring plots. A non-destructive method for AGB estimation is thus needed.

Methods Here, we conducted field measurements at a land-use manipulation experiment in a typical steppe in Nei Mongol, China. We obtained the fractional vegetation cover (FVC) using digital photographs. We also measured leaf area index (LAI), vegetation height, and plant species richness. Three different models were used to estimate AGB: univariate regression model, stepwise regression model, and random forest model.

Important findings We found that FVC, LAI, mean vegetation height, maximum vegetation height and richness were highly correlated with AGB variation. AGB can be accurately predicted by a stepwise regression model developed based on the local plant community. The determination coefficient (R²) and root-mean-square error (RMSE) of the stepwise regression model can reach 0.91 and 35.60 g·m^-2, respectively. Overall, our study provides a rapid and non-destructive method for AGB measurement that can be used as an alternative to the traditional harvest method.

Aims Based on the solar-induced chlorophyll fluorescence (SIF), this study was conducted to reveal the benefit of vegetation productivity in the revegetation region with a significant increase in land surface greenness under the large-scale implementation of ecological programs in the Loess Plateau.

Methods By interpreting satellite-observed terrestrial greenness changes and land use/cover dynamics, we first identified the spatial distribution of revegetation and existing vegetation in the Loess Plateau in the last 20 years. Then, using SIF and meteorological data, the gross primary productivity (GPP) of the revegetation and existing vegetation was calculated according to the revised mechanistic light response (rMLR) model. Finally, we adopted the comparative analysis approach to compare the differences in GPP of the revegetation based on the SIF observations.

Important findings Our results indicated that the ecological programs have made a widespread increase in land surface greenness in the Loess Plateau. In the period 2001 to 2020, the area of revegetated forest was 35 000 km², accounting for 7.42% of the total area, whereas revegetated grassland area was 110 000 km², accounting for 25.25% of the total area. Overall, the photosynthetic capacity and vegetation productivity of the revegetated forests were lower than that of existing forests in the Loess Plateau, while revegetated grassland was higher. GPP of the revegetated forest was equivalent to 83.86% of the existing forest, and GPP of the revegetated grassland was equivalent to 121.10% of that of the existing grassland. At the same leaf area index (LAI) level, GPP of revegetation and existing vegetation showed differences that GPP gap increased as LAI became higher. Revegetation transformed from bare land showed the lowest vegetation productivity, whereas forest growth and grassland restoration from cropland were the optimal land use/cover transition pattern for the revegetated forest and revegetated grassland, respectively. LAI increasing rate and restoration time also affected the productivity of revegetation, revegetated areas with higher LAI increasing rate showed more extensive productivity benefits. Vegetation productivity of revegetated forest increased with standage, while revegetated grassland with shorter restoration periods showed higher productivity. Overall, although ecological programs have widely increased vegetation cover and biomass in the Loess Plateau, however, there exists a certain gap in GPP between the revegetation and existing vegetation areas (especially in forests), thereby affecting the ecological benefits of the ecological programs.

Aims Accurately obtaining forest structural attributes is important for forest ecosystem research and protection. As a key data source, satellite remote sensing data are used to derive various regional and global products of forest structure and conditions, which are widely used in forest condition evaluation, forest biomass estimation, and forest disturbance and biodiversity monitoring. However, these products derived from satellite remote sensing data lack verification for Chinaʼs forested areas, and their accuracy and uncertainty under different forest structure and terrain conditions is not clear. Light detection and ranging (LiDAR) has the advantage of acquiring high-precision three-dimensional information. It has been widely used in monitoring forest ecosystems and validating various datasets of forest structure derived from remote sensing data. This study focused on evaluating the accuracy of Global Land Surface Satellite Products System-Leaf Area Index (GLASS LAI), Global Land Cover Facility-Tree Canopy Cover (GLCF TCC), and Global Forest Canopy Height (GFCH) products in China based on massive unmanned aerial vehicle (UAV) LiDAR data.

Methods We collected nationwide LiDAR point cloud data at 114 sites in China’s forested areas to build the benchmark validation dataset including canopy cover, canopy height and LAI. The corresponding pixel values of the above three products were extracted using the geolocation from UAV LiDAR data. The coefficient of determination (R²) and root mean square error (RMSE) were used to evaluate the accuracy and uncertainty of the three products. The uncertainty under different forest types, canopy cover and terrain conditions were also analyzed.

Important findings The results indicate that compared to the LAI, canopy cover and canopy height derived from UAV LiDAR data, GLASS LAI (R² = 0.29, RMSE= 2.1 m²·m^-2), GLCF TCC (R²= 0.47, RMSE= 31%), GFCH (R²= 0.37, RMSE = 5 m) all exhibit large uncertainties and suffer from saturation problems in China’s forested areas, and their accuracy varies significantly across forest types, canopy cover and terrain conditions. In general, the GLASS LAI and GLCF TCC are mainly influenced by forest types and canopy cover, respectively. In contrast, both slope and canopy cover have large influences on the accuracy of GFCH.