Institute of Forest Resource Information Techniques

The biodiversity-productivity relationship (BPR) is foundational to our understanding of the global extinction crisis and its impacts on ecosystem functioning. Understanding BPR is critical for the accurate valuation and effective conservation of biodiversity. Using ground-sourced data from 777,126 permanent plots, spanning 44 countries and most terrestrial biomes, we reveal a globally consistent positive concave-down BPR, showing that continued biodiversity loss would result in an accelerating decline in forest productivity worldwide. The value of biodiversity in maintaining commercial forest productivity alone-US$166 billion to 490 billion per year according to our estimation-is more than twice what it would cost to implement effective global conservation. This highlights the need for a worldwide reassessment of biodiversity values, forest management strategies, and conservation priorities.

The Watershed Allied Telemetry Experimental Research (WATER) is a simultaneous airborne, satellite‐borne, and ground‐based remote sensing experiment aiming to improve the observability, understanding, and predictability of hydrological and related ecological processes at a catchment scale. WATER consists of the cold region, forest, and arid region hydrological experiments as well as a hydrometeorology experiment and took place in the Heihe River Basin, a typical inland river basin in the northwest of China. The field campaigns have been completed, with an intensive observation period lasting from 7 March to 12 April, from 15 May to 22 July, and from 23 August to 5 September 2008: in total, 120 days. Twenty‐five airborne missions were flown. Airborne sensors including microwave radiometers at L, K, and Ka bands, imaging spectrometer, thermal imager, CCD, and lidar were used. Various satellite data were collected. Ground measurements were carried out at four scales, that is, key experimental area, foci experimental area, experiment site, and elementary sampling plot, using ground‐based remote sensing instruments, densified network of automatic meteorological stations, flux towers, and hydrological stations. On the basis of these measurements, the remote sensing retrieval models and algorithms of water cycle variables are to be developed or improved, and a catchment‐scale land/hydrological data assimilation system is being developed. This paper reviews the background, scientific objectives, experiment design, filed campaign implementation, and current status of WATER. The analysis of the data will continue over the next 2 years, and limited revisits to the field are anticipated.

The development of forestry as a scientific and management discipline over the last two centuries has mainly emphasized intensive management operations focused on increased commodity production, mostly wood. This “conventional” forest management approach has typically favored production of even-aged, single-species stands. While alternative management regimes have generally received less attention, this has been changing over the last three decades, especially in countries with developed economies. Reasons for this change include a combination of new information and concerns about the ecological consequences of intensive forestry practices and a willingness on the part of many forest owners and society to embrace a wider set of management objectives. Alternative silvicultural approaches are characterized by a set of fundamental principles, including avoidance of clearcutting, an emphasis on structural diversity and small-scale variability, deployment of mixed species with natural regeneration, and avoidance of intensive site-preparation methods. Our compilation of the authors’ experiences and perspectives from various parts of the world aims to initiate a larger discussion concerning the constraints to and the potential of adopting alternative silvicultural practices. The results suggest that a wider adoption of alternative silvicultural practices is currently hindered by a suite of ecological, economic, logistical, informational, cultural, and historical constraints. Individual contexts display their own unique combinations and relative significance of these constraints, and accordingly, targeted efforts, such as regulations and incentives, may help to overcome specific challenges. In a broader context, we propose that less emphases on strict applications of principles and on stand structures might provide additional flexibility and facilitate the adoption of alternative silvicultural regimes in a broader set of circumstances. At the same time, the acceptance of alternative silvicultural systems as the “preferred or default mode of management” will necessitate and benefit from the continued development of the scientific basis and valuation of a variety of ecosystem goods and services. This publication is aimed to further the discussion in this context.

General-purpose foundation models have led to recent breakthroughs in artificial intelligence. In remote sensing, self-supervised learning (SSL) and Masked Image Modeling (MIM) have been adopted to build foundation models. However, these models primarily learn low-level features and require annotated data for fine-tuning. Moreover, they are inapplicable for retrieval and zero-shot applications due to the lack of language understanding. To address these limitations, we propose RemoteCLIP, the first vision-language foundation model for remote sensing that aims to learn robust visual features with rich semantics and aligned text embeddings for seamless downstream application. To address the scarcity of pre-training data, we leverage data scaling which converts heterogeneous annotations into a unified image-caption data format based on Box-to-Caption (B2C) and Mask-to-Box (M2B) conversion. By further incorporating UAV imagery, we produce a 12 × larger pretraining dataset than the combination of all available datasets. RemoteCLIP can be applied to a variety of downstream tasks, including zero-shot image classification, linear probing, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">k</i> -NN classification, few-shot classification, image-text retrieval, and object counting in remote sensing images. Evaluation on 16 datasets, including a newly introduced RemoteCount benchmark to test the object counting ability, shows that RemoteCLIP consistently outperforms baseline foundation models across different model scales. Impressively, RemoteCLIP beats the state-of-the-art method by 9.14% mean recall on the RSITMD dataset and 8.92% on the RSICD dataset. For zero-shot classification, our RemoteCLIP outperforms the CLIP baseline by up to 6.39% average accuracy on 12 downstream datasets.

Abstract Viewed within a historical context, Asia has experienced dramatic land transformations, and currently more than 50% of Asian land area is under agriculture. The consequences of this transformation are manifold. Southeast Asia has the highest deforestation rate of any major tropical region. Many of the world's large rivers and lakes in Asia have been heavily degraded. About 11 of 19 world megacities with more than 10 million inhabitants are in Asia. These land use activities have resulted in substantial negative ecological consequences, including increased anthropogenic CO 2 emissions, deteriorated air and water quality, alteration of regional climate, an increase of disease and a loss of biodiversity. Although land use occurs at the local level, it has the potential to cause ecological impact across local, regional and global scales. Reducing the negative environmental impacts of land use change while maintaining economic viability and social acceptability is an major challenge for most developing countries in Asia.

Abstract Forest productivity may be determined not only by biodiversity but also by environmental factors and stand structure attributes. However, the relative importance of these factors in determining productivity is still controversial for subtropical forests. Based on a large dataset from 600 permanent forest inventory plots across subtropical China, we examined the relationship between biodiversity and forest productivity and tested whether stand structural attributes (stand density in terms of trees per ha, age and tree size) and environmental factors (climate and site conditions) had larger effects on productivity. Furthermore, we quantified the relative importance of environmental factors, stand structure and diversity in determining forest productivity. Diversity, together with stand structure and site conditions, regulated the variability in forest productivity. The relationship between diversity and forest productivity did not vary along environmental gradients. Stand density and age were more important modulators of forest productivity than diversity. Synthesis . Diversity had significant and positive effects on productivity in species‐rich subtropical forests, but the effects of stand density and age were also important. Our work highlights that while biodiversity conservation is often important, the regulation of stand structure can be even more important to maintain high productivity in subtropical forests.

Land cover type is a crucial parameter that is required for various land surface models that simulate water and carbon cycles, ecosystem dynamics, and climate change. Many land use/land cover maps used in recent years have been derived from field investigations and remote-sensing observations. However, no land cover map that is derived from a single source (such as satellite observation) properly meets the needs of land surface simulation in China. This article presents a decision-fuse method to produce a higher-accuracy land cover map by combining multi-source local data based on the Dempster–Shafer (D–S) evidence theory. A practical evidence generation scheme was used to integrate multi-source land cover classification information. The basic probability values of the input data were obtained from literature reviews and expert knowledge. A Multi-source Integrated Chinese Land Cover (MICLCover) map was generated by combining multi-source land cover/land use classification maps including a 1:1,000,000 vegetation map, a 1:100,000 land use map for the year 2000, a 1:1,000,000 swamp-wetland map, a glacier map, and a Moderate-Resolution Imaging Spectroradiometer land cover map for China in 2001 (MODIS2001). The merit of this new map is that it uses a common classification system (the International Geosphere-Biosphere Programme (IGBP) land cover classification system), and it has a unified 1 km resolution. The accuracy of the new map was validated by a hybrid procedure. The validation results show great improvement in accuracy for the MICLCover map. The local-scale visual comparison validations for three regions show that the MICLCover map provides more spatial details on land cover at the local scale compared with other popular land cover products. The improvement in accuracy is true for all classes but particularly for cropland, urban, glacier, wetland, and water body classes. Validation by comparison with the China Forestry Scientific Data Center (CFSDC)–Forest Inventory Data (FID) data shows that overall forest accuracies in five provinces increased to between 42.19% and 88.65% for our MICLCover map, while those of the MODIS2001 map increased between 27.77% and 77.89%. The validation all over China shows that the overall accuracy of the MICLCover map is 71%, which is higher than the accuracies of other land cover maps. This map therefore can be used as an important input for land surface models of China. It has the potential to improve the modeling accuracy of land surface processes as well as to support other aspects of scientific land surface investigations in China.

Relationships between stand growth and structural diversity were examined in spruce-dominated forests in New Brunswick, Canada. Net growth, survivor growth, mortality, and recruitment represented stand growth, and tree species, size, and height diversity indices were used to describe structural diversity. Mixed-effects second-order polynomial regressions were employed for statistical analysis. Results showed stand structural diversity had a significant positive effect on net growth and survivor growth by volume but not on mortality and recruitment. Among the tested diversity indices, the integrated diversity of tree species and height contributed most to stand net growth and survivor growth. Structural diversity showed increasing trends throughout the developmental stages from young, immature, mature, and overmature forest stands. This relationship between stand growth and structural diversity may be due to stands featuring high structural diversity that enhances niche complementarities of resource use because trees exist within different horizontal and vertical layers, and strong competition resulted from size differences among trees. It is recommended to include effects of species and structural diversity in forest growth modeling initiatives. Moreover, uneven-aged stand management in conjunction with selective or partial cutting to maintain high structural diversity is also recommended to maintain biodiversity and rapid growth in spruce-dominated forests.

Multiview learning (MVL), which enhances the learners' performance by coordinating complementarity and consistency among different views, has attracted much attention. The multiview generalized eigenvalue proximal support vector machine (MvGSVM) is a recently proposed effective binary classification method, which introduces the concept of MVL into the classical generalized eigenvalue proximal support vector machine (GEPSVM). However, this approach cannot guarantee good classification performance and robustness yet. In this article, we develop multiview robust double-sided twin SVM (MvRDTSVM) with SVM-type problems, which introduces a set of double-sided constraints into the proposed model to promote classification performance. To improve the robustness of MvRDTSVM against outliers, we take L1-norm as the distance metric. Also, a fast version of MvRDTSVM (called MvFRDTSVM) is further presented. The reformulated problems are complex, and solving them are very challenging. As one of the main contributions of this article, we design two effective iterative algorithms to optimize the proposed nonconvex problems and then conduct theoretical analysis on the algorithms. The experimental results verify the effectiveness of our proposed methods.

The global availability of high spatial resolution images makes mapping tree species distribution possible for better management of forest resources. Previous research mainly focused on mapping single tree species, but information about the spatial distribution of all kinds of trees, especially plantations, is often required. This research aims to identify suitable variables and algorithms for classifying land cover, forest, and tree species. Bi-temporal ZiYuan-3 multispectral and stereo images were used. Spectral responses and textures from multispectral imagery, canopy height features from bi-temporal stereo imagery, and slope and elevation from the stereo-derived digital surface model data were examined through comparative analysis of six classification algorithms including maximum likelihood classifier (MLC), k-nearest neighbor (kNN), decision tree (DT), random forest (RF), artificial neural network (ANN), and support vector machine (SVM). The results showed that use of multiple source data—spectral bands, vegetation indices, textures, and topographic factors—considerably improved land-cover and forest classification accuracies compared to spectral bands alone, which the highest overall accuracy of 84.5% for land cover classes was from the SVM, and, of 89.2% for forest classes, was from the MLC. The combination of leaf-on and leaf-off seasonal images further improved classification accuracies by 7.8% to 15.0% for land cover classes and by 6.0% to 11.8% for forest classes compared to single season spectral image. The combination of multiple source data also improved land cover classification by 3.7% to 15.5% and forest classification by 1.0% to 12.7% compared to the spectral image alone. MLC provided better land-cover and forest classification accuracies than machine learning algorithms when spectral data alone were used. However, some machine learning approaches such as RF and SVM provided better performance than MLC when multiple data sources were used. Further addition of canopy height features into multiple source data had no or limited effects in improving land-cover or forest classification, but improved classification accuracies of some tree species such as birch and Mongolia scotch pine. Considering tree species classification, Chinese pine, Mongolia scotch pine, red pine, aspen and elm, and other broadleaf trees as having classification accuracies of over 92%, and larch and birch have relatively low accuracies of 87.3% and 84.5%. However, these high classification accuracies are from different data sources and classification algorithms, and no one classification algorithm provided the best accuracy for all tree species classes. This research implies the same data source and the classification algorithm cannot provide the best classification results for different land cover classes. It is necessary to develop a comprehensive classification procedure using an expert-based approach or hierarchical-based classification approach that can employ specific data variables and algorithm for each tree species class.

Of late, there are many studies on the robust discriminant analysis, which adopt L <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -norm as the distance metric, but their results are not robust enough to gain universal acceptance. To overcome this problem, the authors of this article present a nonpeaked discriminant analysis (NPDA) technique, in which cutting L <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -norm is adopted as the distance metric. As this kind of norm can better eliminate heavy outliers in learning models, the proposed algorithm is expected to be stronger in performing feature extraction tasks for data representation than the existing robust discriminant analysis techniques, which are based on the L <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -norm distance metric. The authors also present a comprehensive analysis to show that cutting L <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -norm distance can be computed equally well, using the difference between two special convex functions. Against this background, an efficient iterative algorithm is designed for the optimization of the proposed objective. Theoretical proofs on the convergence of the algorithm are also presented. Theoretical insights and effectiveness of the proposed method are validated by experimental tests on several real data sets.

-norm as the distance metric. However, both of their robustness and discriminant power are limited. In this article, we present a new robust discriminant subspace (RDS) learning method for feature extraction, with an objective function formulated in a different form. To guarantee the subspace to be robust and discriminative, we measure the within-class distances based on [Formula: see text]-norm and use [Formula: see text]-norm to measure the between-class distances. This also makes our method include rotational invariance. Since the proposed model involves both [Formula: see text]-norm maximization and [Formula: see text]-norm minimization, it is very challenging to solve. To address this problem, we present an efficient nongreedy iterative algorithm. Besides, motivated by trace ratio criterion, a mechanism of automatically balancing the contributions of different terms in our objective is found. RDS is very flexible, as it can be extended to other existing feature extraction techniques. An in-depth theoretical analysis of the algorithm's convergence is presented in this article. Experiments are conducted on several typical databases for image classification, and the promising results indicate the effectiveness of RDS.

Forest carbon sequestration capacity in China remains uncertain due to underrepresented tree demographic dynamics and overlooked of harvest impacts. In this study, we employ a process-based biogeochemical model to make projections by using national forest inventories, covering approximately 415,000 permanent plots, revealing an expansion in biomass carbon stock by 13.6 ± 1.5 Pg C from 2020 to 2100, with additional sink through augmentation of wood product pool (0.6-2.0 Pg C) and spatiotemporal optimization of forest management (2.3 ± 0.03 Pg C). We find that statistical model might cause large bias in long-term projection due to underrepresentation or neglect of wood harvest and forest demographic changes. Remarkably, disregarding the repercussions of harvesting on forest age can result in a premature shift in the timing of the carbon sink peak by 1-3 decades. Our findings emphasize the pressing necessity for the swift implementation of optimal forest management strategies for carbon sequestration enhancement.

We describe the design, implementation and performance of a novel airborne system, which integrates commercial waveform LiDAR, CCD (Charge-Coupled Device) camera and hyperspectral sensors into a common platform system. CAF’s (The Chinese Academy of Forestry) LiCHy (LiDAR, CCD and Hyperspectral) Airborne Observation System is a unique system that permits simultaneous measurements of vegetation vertical structure, horizontal pattern, and foliar spectra from different view angles at very high spatial resolution (~1 m) on a wide range of airborne platforms. The horizontal geo-location accuracy of LiDAR and CCD is about 0.5 m, with LiDAR vertical resolution and accuracy 0.15 m and 0.3 m, respectively. The geo-location accuracy of hyperspectral image is within 2 pixels for nadir view observations and 5–7 pixels for large off-nadir observations of 55° with multi-angle modular when comparing to LiDAR product. The complementary nature of LiCHy’s sensors makes it an effective and comprehensive system for forest inventory, change detection, biodiversity monitoring, carbon accounting and ecosystem service evaluation. The LiCHy system has acquired more than 8000 km2 of data over typical forests across China. These data are being used to investigate potential LiDAR and optical remote sensing applications in forest management, forest carbon accounting, biodiversity evaluation, and to aid in the development of similar satellite configurations. This paper describes the integration of the LiCHy system, the instrument performance and data processing workflow. We also demonstrate LiCHy’s data characteristics, current coverage, and potential vegetation applications.

Deriving individual tree crown (ITC) information from light detection and ranging (LiDAR) data is of great significance to forest resource assessment and smart management. After proof-of-concept studies, advanced deep learning methods have been shown to have high efficiency and accuracy in remote sensing data analysis and geoscience problem solving. This study proposes a novel concept for synergetic use of the YOLO-v4 deep learning network based on heightmaps directly generated from airborne LiDAR data for ITC segmentation and a computer graphics algorithm for refinement of the segmentation results involving overlapping tree crowns. This concept overcomes the limitations experienced by existing ITC segmentation methods that use aerial photographs to obtain texture and crown appearance information and commonly encounter interference due to heterogeneous solar illumination intensities or interlacing branches and leaves. Three generative adversarial networks (WGAN, CycleGAN, and SinGAN) were employed to generate synthetic images. These images were coupled with manually labeled training samples to train the network. Three forest plots, namely, a tree nursery, forest landscape and mixed tree plantation, were used to verify the effectiveness of our approach. The results showed that the overall recall of our method for detecting ITCs in the three forest plot types reached 83.6%, with an overall precision of 81.4%. Compared with reference field measurement data, the coefficient of determination R 2 was ≥ 79.93% for tree crown width estimation, and the accuracy of our deep learning method was not influenced by the values of key parameters, yielding 3.9% greater accuracy than the traditional watershed method. The results demonstrate an enhancement of tree crown segmentation in the form of a heightmap for different forest plot types using the concept of deep learning, and our method bypasses the visual complications arising from aerial images featuring diverse textures and unordered scanned points with irregular geometrical properties.

Water conservation is the core wetland ecological function. The pattern of water conservation in geography often exhibits local spatial heterogeneity. In this paper, the Dongting Lake (DT) and Poyang Lake (PY) wetlands were taken as research objects, and the water yield (WY) module of the Integrated Valuation of Ecosystem Services and Trade-offs (InVEST) model was used to spatially quantify water conservation, with the aim of revealing the differences in water conservation in homogeneous wetlands. We observed spatio-temporal changes in the WY over six periods from 1998 to 2018 and used multivariate analysis of variance (MANOVA) and the system cluster method to explore the patterns and driving factors of water conservation in the DT and PY wetlands. The research results were as follows: (1) the WY in the two wetlands showed declining trends, but the trend in the DT wetland was more obvious. (2) Water conservation in the PY wetland exhibited a more sensitive response to climate than water conservation in the DT wetland. (3) From the perspective of the WY level, the PY wetland was generally high, while the DT wetland was generally low. (4) Climate change was the main contributor to and direct driving factor of water production, while wetland changes and socio-economic factors were indirect driving factors. (5) The effect of wetland cover changes on the WY showed significant differences among wetland cover types, indicating that the drivers of the WY changes had a clear relationship with landscape changes and transfers in wetland cover types. (6) The InVEST model was applicable for evaluating lake wetland water conservation. In summary, paying more attention to the differences in the changes in heterogeneous and homogenous regions can help us understand the factors that impact water conservation. Additionally, rational adjustment of indirect driving forces is conducive to ecological safety management.

Abstract Applying allometric equations in combination with forest inventory data is an effective approach to use when qualifying forest biomass and carbon storage on a regional scale. The objectives of this study were to (1) develop general allometric tree component biomass equations and (2) investigate tree biomass allocation patterns for Pinus massoniana , a principal tree species native to southern China, by applying 197 samples across 20 site locations. The additive allometric equations utilized to compute stem, branch, needle, root, aboveground, and total tree biomass were developed by nonlinear seemingly unrelated regression. Results show that the relative proportion of stem biomass to tree biomass increased while the contribution of canopy biomass to tree biomass decreased as trees continued to grow through time. Total root biomass was a large biomass pool in itself, and its relative proportion to tree biomass exhibited a slight increase with tree growth. Although equations employing stem diameter at breast height (dbh) alone as a predictor could accurately predict stem, aboveground, root, and total tree biomass, they were poorly fitted to predict the canopy biomass component. The inclusion of the tree height ( H ) variable either slightly improved or did not in any way increase model fitness. Validation results demonstrate that these equations are suitable to estimate stem, aboveground, and total tree biomass across a broad range of P . massoniana stands on a regional scale.

Weibull distribution was used to fit tree diameter data collected from 86 sample plots located in Chinese pine stand in Beijing. To estimate the Weibull distribution parameters, three methods [namely maximum likelihood estimation method (MLE), method of moment (MOM) and least-squares regression method (LSM)] were compared and evaluated on the basis of the mean square error (MSE) and sample size. For these sample plots, the moment method was superior for estimating the parameters of Weibull distribution for tree diameter distribution.

This study aims to estimate forest above-ground biomass and biomass components in a stand of Picea crassifolia (a coniferous tree) located on Qilian Mountain, western China via low density small-footprint airborne LiDAR data. LiDAR points were first classified into ground points and vegetation points. After, vegetation statistics, including height quantiles, mean height, and fractional cover were calculated. Stepwise multiple regression models were used to develop equations that relate the vegetation statistics from field inventory data with field-based estimates of biomass for each sample plot. The results showed that stem, branch, and above-ground biomass may be estimated with relatively higher accuracies; estimates have adjusted R2 values of 0.748, 0.749, and 0.727, respectively, root mean squared error (RMSE) values of 9.876, 1.520, and 15.237 Mg·ha−1, respectively, and relative RMSE values of 12.783%, 12.423%, and 14.163%, respectively. Moreover, fruit and crown biomass may be estimated with relatively high accuracies; estimates have adjusted R2 values of 0.578 and 0.648, respectively, RMSE values of 1.022 and 5.963 Mg·ha−1, respectively, and relative RMSE values of 23.273% and 19.665%, respectively. In contrast, foliage biomass estimates have relatively low accuracies; they had an adjusted R2 value of 0.356, an RMSE of 3.691 Mg·ha−1, and a relative RMSE of 26.953%. Finally, above-ground biomass and biomass component spatial maps were established using stepwise multiple regression equations. These maps are very useful for updating and modifying forest base maps and registries.

SUMMARY Forest inventory in China provides valuable information on forestry policy formulation, forest management planning, and international forestry related reports. It includes the national forest inventory (Level I), the forest management planning inventory (Level II), and the forest operation design inventory (Level III) as the primary components. This paper describes the evolution of the inventory system focusing on forest inventories at levels I and II in terms of sampling strategy, plot design, output, and the application of information techniques. The current forest inventory system faces challenges in its transition from a forestry development strategy catered towards timber production in China to one where ecological-oriented management is the primary concern. At the same time, information needs to be expanded from local to national and international scales. Potential improvements are discussed that include sampling design, plot design, potential use of new remote sensing techniques as well as ...