Zhiqiang Xiao

Use of General Regression Neural Networks for Generating the GLASS Leaf Area Index Product From Time-Series MODIS Surface Reflectance

Leaf area index (LAI) products at regional and global scales are being routinely generated from individual instrument data acquired at a specific time. As a result of cloud contamination and other factors, these LAI products are spatially and temporally discontinuous and are also inaccurate for some vegetation types in many areas. A better strategy is to use multi-temporal data. In this paper, a method was developed to estimate LAI from time-series remote sensing data using general regression neural networks (GRNNs). A database was generated from Moderate-Resolution Imaging Spectroradiometer (MODIS) and CYCLOPES LAI products as well as MODIS reflectance products of the BELMANIP sites during the period from 2001-2003. The effective CYCLOPES LAI was first converted to true LAI, which was then combined with the MODIS LAI according to their uncertainties determined from the ground-measured true LAI. The MODIS reflectance was reprocessed to remove remaining effects. GRNNs were then trained over the fused LAI and reprocessed MODIS reflectance for each biome type to retrieve LAI from time-series remote sensing data. The reprocessed MODIS reflectance data from an entire year were inputted into the GRNNs to estimate the 1-year LAI profiles. Extensive validations for all biome types were carried out, and it was demonstrated that the method is able to estimate temporally continuous LAI profiles with much improved accuracy compared with that of the current MODIS and CYCLOPES LAI products. This new method is being used to produce the Global Land Surface Satellite LAI products in China.

A long-term Global LAnd Surface Satellite (GLASS) data-set for environmental studies

Recently, five Global LAnd Surface Satellite (GLASS) products have been released: leaf area index (LAI), shortwave broadband albedo, longwave broadband emissivity, incident short radiation, and photosynthetically active radiation (PAR). The first three products cover the years 1982–2012 (LAI) and 1981–2010 (albedo and emissivity) at 1–5 km and 8-day resolutions, and the last two radiation products span the period 2008–2010 at 5 km and 3-h resolutions. These products have been evaluated and validated, and the preliminary results indicate that they are of higher quality and accuracy than the existing products. In particular, the first three products have much longer time series, and are therefore highly suitable for various environmental studies. This paper outlines the algorithms, product characteristics, preliminary validation results, potential applications and some examples of initial analysis of these products.

Detection and attribution of vegetation greening trend in China over the last 30 years

The reliable detection and attribution of changes in vegetation growth is a prerequisite for the development of strategies for the sustainable management of ecosystems. This is an extraordinary challenge. To our knowledge, this study is the first to comprehensively detect and attribute a greening trend in China over the last three decades. We use three different satellite-derived Leaf Area Index (LAI) datasets for detection as well as five different process-based ecosystem models for attribution. Rising atmospheric CO2 concentration and nitrogen deposition are identified as the most likely causes of the greening trend in China, explaining 85% and 41% of the average growing-season LAI trend (LAIGS) estimated by satellite datasets (average trend of 0.0070 yr(-1), ranging from 0.0035 yr(-1) to 0.0127 yr(-1)), respectively. The contribution of nitrogen deposition is more clearly seen in southern China than in the north of the country. Models disagree about the contribution of climate change alone to the trend in LAIGS at the country scale (one model shows a significant increasing trend, whereas two others show significant decreasing trends). However, the models generally agree on the negative impacts of climate change in north China and Inner Mongolia and the positive impact in the Qinghai-Xizang plateau. Provincial forest area change tends to be significantly correlated with the trend of LAIGS (P < 0.05), and marginally significantly (P = 0.07) correlated with the residual of LAIGS trend, calculated as the trend observed by satellite minus that estimated by models through considering the effects of climate change, rising CO2 concentration and nitrogen deposition, across different provinces. This result highlights the important role of Chinas afforestation program in explaining the spatial patterns of trend in vegetation growth.

A Temporally Integrated Inversion Method for Estimating Leaf Area Index From MODIS Data

Multiple leaf area index (LAI) products have been generated from remote-sensing data. Among them, the Moderate-Resolution Imaging Spectroradiometer (MODIS) LAI product (MOD15A2) is now routinely derived from data acquired by MODIS sensors onboard Terra and Aqua satellite platforms. However, the MODIS LAI product is not spatially and temporally continuous and is inaccurate in many areas for some vegetation types. In this paper, a new algorithm is developed to estimate LAI from time-series MODIS reflectance data (MOD09A1). A radiative-transfer model is coupled with a double-logistic LAI temporal-profile model, and the shuffled complex evolution optimization method, developed at the University of Arizona, is used to estimate the parameters of the coupled model from the temporal signature in a given time window. Preliminary analysis using MODIS surface-reflectance data at flux sites was performed to validate this method. The results show that the new algorithm is able to construct a temporally continuous LAI product efficiently, and the accuracy has been significantly improved over the MODIS LAI product as compared to field-measured LAI data.

A Multiscale and Hierarchical Feature Extraction Method for Terrestrial Laser Scanning Point Cloud Classification

The effective extraction of shape features is an important requirement for the accurate and efficient classification of terrestrial laser scanning (TLS) point clouds. However, the challenge of how to obtain robust and discriminative features from noisy and varying density TLS point clouds remains. This paper introduces a novel multiscale and hierarchical framework, which describes the classification of TLS point clouds of cluttered urban scenes. In this framework, we propose multiscale and hierarchical point clusters (MHPCs). In MHPCs, point clouds are first resampled into different scales. Then, the resampled data set of each scale is aggregated into several hierarchical point clusters, where the point cloud of all scales in each level is termed a point-cluster set. This representation not only accounts for the multiscale properties of point clouds but also well captures their hierarchical structures. Based on the MHPCs, novel features of point clusters are constructed by employing the latent Dirichlet allocation (LDA). An LDA model is trained according to a training set. The LDA model then extracts a set of latent topics, i.e., a feature of topics, for a point cluster. Finally, to apply the introduced features for point-cluster classification, we train an AdaBoost classifier in each point-cluster set and obtain the corresponding classifiers to separate the TLS point clouds with varying point density and data missing into semantic regions. Compared with other methods, our features achieve the best classification results for buildings, trees, people, and cars from TLS point clouds, particularly for small and moving objects, such as people and cars.

The Global Land Surface Satellite (GLASS) Remote Sensing Data Processing System and Products

Using remotely sensed satellite products is the most efficient way to monitor global land, water, and forest resource changes, which are believed to be the main factors for understanding global climate change and its impacts. A reliable remotely sensed product should be retrieved quantitatively through models or statistical methods. However, producing global products requires a complex computing system and massive volumes of multi-sensor and multi-temporal remotely sensed data. This manuscript describes the ground Global LAnd Surface Satellite (GLASS) product generation system that can be used to generate long-sequence time series of global land surface data products based on various remotely sensed data. To ensure stabilization and efficiency in running the system, we used the methods of task management, parallelization, and multi I/O channels. An array of GLASS remote sensing products related to global land surface parameters are currently being produced and distributed by the Center for Global Change Data Processing and Analysis at Beijing Normal University in Beijing, China. These products include Leaf Area Index (LAI), land surface albedo, and broadband emissivity (BBE) from the years 1981 to 2010, downward shortwave radiation (DSR) and photosynthetically active radiation (PAR) from the years 2008 to 2010.

Global Land Surface Fractional Vegetation Cover Estimation Using General Regression Neural Networks From MODIS Surface Reflectance

Fractional vegetation cover (FVC) plays an important role in earth surface process simulations, climate modeling, and global change studies. Several global FVC products have been generated using medium spatial resolution satellite data. However, the validation results indicate inconsistencies, as well as spatial and temporal discontinuities of the current FVC products. The objective of this paper is to develop a reliable estimation algorithm to operationally produce a high-quality global FVC product from the Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance. The high-spatial-resolution FVC data were first generated using Landsat TM/ETM+ data at the global sampling locations, and then, the general regression neural networks (GRNNs) were trained using the high-spatial-resolution FVC data and the reprocessed MODIS surface reflectance data. The direct validation using ground reference data from validation of land European Remote Sensing instruments sites indicated that the performance of the proposed method (R2=0.809, RMSE =0.157) was comparable with that of the GEOV1 FVC product (R2=0.775, RMSE =0.166), which is currently considered to be the best global FVC product from SPOT VEGETATION data. Further comparison indicated that the spatial and temporal continuity of the estimates from the proposed method was superior to that of the GEOV1 FVC product.

Long-Time-Series Global Land Surface Satellite Leaf Area Index Product Derived From MODIS and AVHRR Surface Reflectance

Leaf area index (LAI) is an important vegetation biophysical variable and has been widely used for crop growth monitoring and yield estimation, land-surface process simulation, and global change studies. Several LAI products currently exist, but most have limited temporal coverage. A long-term high-quality global LAI product is required for greatly expanded application of LAI data. In this paper, a method previously proposed was improved to generate a long time series of Global LAnd Surface Satellite (GLASS) LAI product from Advanced Very High Resolution Radiometer (AVHRR) and Moderate Resolution Imaging Spectroradiometer (MOD!S) reflectance data. The GLASS LAI product has a temporal resolution of eight days and spans from 1981 to 2014. During 1981-1999, the LAI product was generated from AVHRR reflectance data and was provided in a geographic latitude/longitude projection at a spatial resolution of 0.05°. During 2000-2014, the LAI product was derived from MODIS surface-reflectance data and was provided in a sinusoidal projection at a spatial resolution of 1 km. The GLASS LAI values derived from MODIS and AVHRR reflectance data form a consistent data set at a spatial resolution of 0.05°. Comparison of the GLASS LAI product with the MODIS LAI product (MOD15) and the first version of the Geoland2 (GEOV1) LAI product indicates that the global consistency of these LAI products is generally good. However, relatively large discrepancies among these LAI products were observed in tropical forest regions, where the GEOV1 LAI values were clearly lower than the GLASS and MOD15 LAI values, particularly in January. A quantitative comparison of temporal profiles shows that the temporal smoothness of the GLASS LAI product is superior to that of the GEOV1 and MODIS LAI products. Direct validation with the mean values of high-resolution LAI maps demonstrates that the GLASS LAI values were closer to the mean values of the high-resolution LAI maps (RMSE = 0.7848 and R2 = 0.8095) than the GEOV1 LAI values (RMSE = 0.9084 and R2 = 0.7939) and the MOD15 LAI values (RMSE = 1.1173 and R2 = 0.6705).

Leaf Area Index

This chapter briefly introduces the inversion algorithms used, discusses the product characteristics and validation results, and presents preliminary analyses and applications. The inversion algorithm was developed to estimate LAI from time-series remote sensing data using general regression neural networks (GRNNs). Unlike existing neural network methods that use remote sensing data acquired only at a specific time to retrieve LAI, the GRNNs used in this study are trained using the fused time-series LAI values from the MODIS and CYCLOPES LAI products and the reprocessed MODIS/AVHRR reflectance. The reprocessed time-series MODIS/AVHRR reflectance values from an entire year were input to the GRNNs to estimate the 1 year LAI profiles. This algorithm has been used to produce the GLASS LAI, one of the longest duration (1981–2012) LAI products in the world. Extensive validations for all biome types have been carried out, and it has been demonstrated that the GLASS LAI presents temporally continuous LAI profiles with much improved quality and accuracy compared with those of the current MODIS and CYCLOPES LAI products.

A Framework for Consistent Estimation of Leaf Area Index, Fraction of Absorbed Photosynthetically Active Radiation, and Surface Albedo from MODIS Time-Series Data

Currently available land-surface parameter products are generated using parameter-specific algorithms from various satellite data and contain several inconsistencies. This paper developed a new data assimilation framework for consistent estimation of multiple land-surface parameters from time-series MODerate Resolution Imaging Spectroradiometer (MODIS) surface reflectance data. If the reflectance data showed snow-free areas, an ensemble Kalman filter (EnKF) technique was used to estimate leaf area index (LAI) for a two-layer canopy reflectance model (ACRM) by combining predictions from a phenology model and the MODIS surface reflectance data. The estimated LAI values were then input into the ACRM to calculate the surface albedo and the fraction of absorbed photosynthetically active radiation (FAPAR). For snow-covered areas, the surface albedo was calculated as the underlying vegetation canopy albedo plus the weighted distance between the underlying vegetation canopy albedo and the albedo over deep snow. The LAI/FAPAR and surface albedo values estimated using this framework were compared with MODIS collection 5 eight-day 1-km LAI/FAPAR products (MOD15A2) and 500-m surface albedo product (MCD43A3), and GEOV1 LAI/FAPAR products at 1/112° spatial resolution and a ten-day frequency, respectively, and validated by ground measurement data from several sites with different vegetation types. The results demonstrate that this new data assimilation framework can estimate temporally complete land-surface parameter profiles from MODIS time-series reflectance data even if some of the reflectance data are contaminated by residual cloud or are missing and that the retrieved LAI, FAPAR, and surface albedo values are physically consistent. The root mean square errors of the retrieved LAI, FAPAR, and surface albedo against ground measurements are 0.5791, 0.0453, and 0.0190, respectively.