Bingwen Qiu

Characterizing spatiotemporal non-stationarity in vegetation dynamics in China using MODIS EVI dataset

This paper evaluated the spatiotemporal non-stationarity in the vegetation dynamic based on 1-km resolution 16-day composite Moderate Resolution Imaging Spectroradiometer (MODIS) Enhanced Vegetation Index (EVI) datasets in China during 2001–2011 through a wavelet transform method. First, it revealed from selected pixels that agricultural crops, natural forests, and meadows were characterized by their distinct intra-annual temporal variation patterns in different climate regions. The amplitude of intra-annual variability generally increased with latitude. Second, parameters calculated using a per-pixel strategy indicated that the natural forests had the strongest variation pattern from seasonal to semiannual scales, and the multiple-cropping croplands typically showed almost equal variances distributed at monthly, seasonal, and semiannual scales. Third, spatiotemporal non-stationarity induced from cloud cover was also evaluated. It revealed that the EVI temporal profiles were significantly distorted with regular summer cloud cover in tropical and subtropical regions. Nevertheless, no significant differences were observed from those statistical parameters related to the interannual and interannual components between the de-clouded and the original MODIS EVI datasets across the whole country. Finally, 12 vegetation zones were proposed based on spatiotemporal variability, as indicated by the magnitude of interannual and intra-annual dynamic components, normalized wavelet variances of detailed components from monthly to semiannual scale, and proportion of cloud cover in summer. This paper provides insightful solutions for addressing spatiotemporal non-stationarity by evaluating the magnitude and frequency of vegetation variability using monthly, seasonal, semiannual to interannual scales across the whole study area.

A new methodology to map double-cropping croplands based on continuous wavelet transform

Abstract Cropping intensity is one of the major factors in crop production and agricultural intensification. A new double-cropping croplands mapping methodology using Moderate Resolution Imaging Spectroradiometer (MODIS) time series datasets through continuous wavelet transform was proposed in this study. This methodology involved four steps. First, daily continuous MODIS Enhanced Vegetation Index (EVI) time series datasets were developed for the study year. Next, the EVI time series datasets were transformed into a two dimensional (time–frequency) wavelet scalogram based on continuous wavelet transform. Third, a feature extraction process was conducted on the wavelet scalogram, where the characteristic spectra were calculated from the wavelet scalogram and the feature peak within two skeleton lines was obtained. Finally, a threshold was determined for feature peak values to discriminate double-cropping croplands within each pixel. The application of the proposed procedure to Chinas Henan Province in 2010 produced an objective and accurate spatial distribution map, which correlated well with in situ observation data (over 90% agreement). The proposed new methodology efficiently handled complex variability that might be caused by regional variation in climate, management practices, growth peaks by winter weed or winter wheat, and data noise. Therefore, the methodology shows promise for future studies at regional and global scales.

Effect of topography and accessibility on vegetation dynamic pattern in Mountain-hill Region

Knowledge of both vegetation distribution pattern and phenology changes is very important. Their complicated relationship with elevation and accessibility were explored through a geographically weighted regression (GWR) framework in Fujian province, China. The 16-day time series of 250 m Moderate Resolution Imaging Spectroradiometer (MODIS) Enhanced Vegetation Index (EVI) dataset from 2000 to 2010 was applied. Wavelet transform method was adopted to decompose the original time series and construct the annual maximum EVI and amplitude of the annual phenological cycle (ΔEVI). Candidate explaining factors included topographic conditions, accessibility variables and proportions of primary vegetation types. Results revealed very strong positive influence from parameters of elevation and accessibility to big rivers and negative effect from accessibility to resident on both maximum EVI and phenological magnitude through ordinary linear least square (OLS) regression analysis. GWR analysis revealed that spatially, the parameters of topography and accessibility had a very complex relationship with both maximum EVI and phenology magnitude, as a result of the various combinations of environmental factors, vegetation composition and also intensive anthropogenic impact. Apart from the continuously increasing trend of phenology magnitude with increasing altitude, the influence of topography and accessibility on maximum EVI and phenological magnitude generally decreased, even from strongly positive to negative, with increasing altitude or distance. Specially, the most rapid change of correlation coefficient between them was observed within a low elevation or close distance; less variation was discovered within a certain range of medium altitude or distance and their relationship might change above this range. Non-stationary approaches are needed to better characterize the complex vegetation dynamic pattern in Mountain-hill Region.

A simple smoother based on continuous wavelet transform: Comparative evaluation based on the fidelity, smoothness and efficiency in phenological estimation

Abstract This study proposed a simple Smoother without any local adjustments based on Continuous Wavelet Transform (SCWT). And then it evaluated its performance together with other commonly applied techniques in phenological estimation. These noise reduction methods included Savitzky–Golay filter (SG), Double Logistic function (DL), Asymmetric Gaussian function (AG), Whittaker Smoother (WS) and Harmonic Analysis of Time-Series (HANTS). They were evaluated based on fidelity and smoothness, and their efficiencies in deriving phenological parameters through the inflexion point-based method with the 8-day composite Moderate Resolution Imaging Spectroradiometer (MODIS) 2-band Enhanced Vegetation Index (EVI2) in 2013 in China. The following conclusions were drawn: (1) The SG method exhibited strong fidelity, but weak smoothness and spatial continuity. (2) The HANTS method had very robust smoothness but weak fidelity. (3) The AG and DL methods performed weakly for vegetation with more than one growth cycle (i.e., multiple crops). (4) The WS and SCWT smoothers outperformed others with combined considerations of fidelity and smoothness, and consistent phenological patterns (correlation coefficients greater than 0.8 except evergreen broadleaf forests (0.68)). (5) Compared with WS methods, the SCWT smoother was capable in preservation of real local minima and maxima with fewer inflexions. (6) Large discrepancy was examined from the estimated phenological dates with SG and HANTS methods, particularly in evergreen forests and multiple cropping regions (the absolute mean deviation rates were 6.2–17.5 days and correlation coefficients less than 0.34 for estimated start dates).

Vegetation distribution pattern along altitudinal gradient in subtropical mountainous and hilly river basin, China

Knowledge of vegetation distribution patterns is very important. Their relationships with topography and climate were explored through a geographically weighted regression (GWR) framework in a subtropical mountainous and hilly region, Minjiang River Basin of Fujian in China. The HJ-1 satellite image acquired on December 9, 2010 was utilized and NDVI index was calculated representing the range of vegetation greenness. Proper analysis units were achieved through segregation based on small sub-basins and altitudinal bands. Results indicated that the GWR model was more powerful than ordinary linear least square (OLS) regression in interpreting vegetation-environmental relationship, indicated by higher adjusted R2 and lower Akaike information criterion values. On one side, the OLS analysis revealed dominant positive influence from parameters of elevation and slope on vegetation distribution. On the other side, GWR analysis indicated that spatially, the parameters of topography had a very complex relationship with the vegetation distribution, as results of the various combinations of environmental factors, vegetation composition and also anthropogenic impact. The influences of elevation and slope generally decreased, from strongly positive to nearly zero, with increasing altitude and slope. Specially, most rapid changes of coefficients between NDVI and elevation or slope were observed in relatively flat and low-lying areas. This paper confirmed that the non-stationary analysis through the framework of GWR could lead to a better understanding of vegetation distribution in subtropical mountainous and hilly region. It was hoped that the proposed scale selection method combined with GWR framework would provide some guidelines on dealing with both spatial (horizontal) and altitudinal (vertical) non-stationarity in the dataset, and it could easily be applied in characterizing vegetation distribution patterns in other mountainous and hilly river basins and related research.

Developing soil indices based on brightness, darkness, and greenness to improve land surface mapping accuracy

Soil, as one of the three basic biophysical components, has been understudied using remote sensing techniques compared to vegetation and impervious surface areas (ISA). This study characterized land surfaces based on the brightness–darkness–greenness model. These three dimensions, brightness, darkness, and greenness, were represented by the first Tasseled Cap Transformation (TC1), Normalize Difference Snow Index (NDSI), and Normalized Difference Vegetation Index (NDVI), respectively. The Ratio Index for Bright Soil (RIBS) was developed based on TC1 and NDSI, and the Product Index for Dark Soil (PIDS) was established by TC1 and NDVI. Their applications to the Landsat 8 Operational Land Imager images and 500 m 8-day composite Moderate Resolution Imaging Spectroradiometer (MODIS) in China revealed the efficiency. The two soil indices proficiently highlighted soil covers with consistently the smallest values, due to larger TC1 and smaller NDSI values in bright soil, and smaller NDVI and TC1 values in dark soil. The RIBS is capable of distinguishing bright soil from ISA without masking vegetation and water body. The spectral separability bright soil and ISA were perfect, with a Jeffries–Matusita distance of 1.916. And the PIDS was the only soil index that could discriminate dark soil from other land covers including ISA. The soil areas in China were classified using a simple threshold method based on MODIS images. An overall accuracy of 94.00% was obtained, with the kappa index of 0.8789. This study provided valuable insights into developing indices for characterizing land surfaces from different perspectives.

Automated cropping intensity extraction from isolines of wavelet spectra

Display Omitted Obtain positive wavelet isolines from wavelet spectra.Derive CI from number of strong bright centers and skeleton width.Handle complex intra-class variability through time and frequency.Guarantee great generality, comparability and consistency.Applicable to VI time series from either coarse or high images. Timely and accurate monitoring of cropping intensity (CI) is essential to help us understand changes in food production. This paper aims to develop an automatic Cropping Intensity extraction method based on the Isolines of Wavelet Spectra (CIIWS) with consideration of intra-class variability. The CIIWS method involves the following procedures: (1) characterizing vegetation dynamics from time-frequency dimensions through a continuous wavelet transform performed on vegetation index temporal profiles; (2) deriving three main features, the skeleton width, maximum number of strong brightness centers and the intersection of their scale intervals, through computing a series of wavelet isolines from the wavelet spectra; and (3) developing an automatic cropping intensity classifier based on these three features. The proposed CIIWS method improves the understanding in the spectral-temporal properties of vegetation dynamic processes. To test its efficiency, the CIIWS method is applied to Chinas Henan province using 250m 8days composite Moderate Resolution Imaging Spectroradiometer (MODIS) Enhanced Vegetation Index (EVI) time series datasets. An overall accuracy of 88.9% is achieved when compared with in-situ observation data. The mapping result is also evaluated with 30m Chinese Environmental Disaster Reduction Satellite (HJ-1)-derived data and an overall accuracy of 86.7% is obtained. At county level, the MODIS-derived sown areas and agricultural statistical data are well correlated (r2=0.85). The merit and uniqueness of the CIIWS method is the ability to cope with the complex intra-class variability through continuous wavelet transform and efficient feature extraction based on wavelet isolines. As an objective and meaningful algorithm, it guarantees easy applications and greatly contributes to satellite observations of vegetation dynamics and food security efforts.

Spatiotemporal analysis of vegetation variability and its relationship with climate change in China

This paper investigated spatiotemporal dynamic pattern of vegetation, climate factor, and their complex relationships from seasonal to inter-annual scale in China during the period 1982–1998 through wavelet transform method based on GIMMS data-sets. First, most vegetation canopies demonstrated obvious seasonality, increasing with latitudinal gradient. Second, obvious dynamic trends were observed in both vegetation and climate change, especially the positive trends. Over 70% areas were observed with obvious vegetation greening up, with vegetation degradation principally in the Pearl River Delta, Yangtze River Delta, and desert. Overall warming trend was observed across the whole country (>98% area), stronger in Northern China. Although over half of area (58.2%) obtained increasing rainfall trend, around a quarter of area (24.5%), especially the Central China and most northern portion of China, exhibited significantly negative rainfall trend. Third, significantly positive normalized difference vegetation index (NDVI)–climate relationship was generally observed on the de-noised time series in most vegetated regions, corresponding to their synchronous stronger seasonal pattern. Finally, at inter-annual level, the NDVI–climate relationship differed with climatic regions and their long-term trends: in humid regions, positive coefficients were observed except in regions with vegetation degradation; in arid, semiarid, and semihumid regions, positive relationships would be examined on the condition that increasing rainfall could compensate the increasing water requirement along with increasing temperature. This study provided valuable insights into the long-term vegetation–climate relationship in China with consideration of their spatiotemporal variability and overall trend in the global change process.

Identifying scale-location specific control on vegetation distribution in mountain-hill region

The scale-location specific control on vegetation distribution was investigated through continuous wavelet transforms approaches in subtropical mountain-hill region, Fujian, China. The Normalized Difference Vegetation Index (NDVI) was calculated as an indicator of vegetation greenness using Chinese Environmental Disaster Reduction Satellite images along latitudinal and longitudinal transects. Four scales of variations were identified from the local wavelet spectrum of NDVI, with much stronger wavelet variances observed at larger scales. The characteristic scale of vegetation distribution within mountainous and hilly regions in Southeast China was around 20 km. Significantly strong wavelet coherency was generally examined in regions with very diverse topography, typically characterized as small mountains and hills fractured by rivers and residents. The continuous wavelet based approaches provided valuable insight on the hierarchical structure and its corresponding characteristic scales of ecosystems, which might be applied in defining proper levels in multilevel models and optimal bandwidths in Geographically Weighted Regression.

Automatic and adaptive paddy rice mapping using Landsat images: Case study in Songnen Plain in Northeast China

Spatiotemporal explicit information on paddy rice distribution is essential for ensuring food security and sustainable environmental management. Paddy rice mapping algorithm through the Combined Consideration of Vegetation phenology and Surface water variations (CCVS) has been efficiently applied based on the 8day composites time series datasets. However, the great challenge for phenology-based algorithms introduced by unpromising data availability in middle/high spatial resolution imagery, such as frequent cloud cover and coarse temporal resolution, remained unsolved. This study addressed this challenge through developing an automatic and Adaptive paddy Rice Mapping Method (ARMM) based on the cloud frequency and spectral separability. The proposed ARMM method was tested on the Landsat 8 Operational Land Imager (OLI) image (path/row 118/028) in the Songnen Plain in Northeast China in 2015. First, the whole study region was automatically and adaptively subdivided into undisturbed and disturbed regions through a per-pixel strategy based on Landsat image data availability during key phenological stage. Second, image objects were extracted from approximately cloud-free images in disturbed and undisturbed regions, respectively. Third, phenological metrics and other feature images from individual or multiple images were developed. Finally, a flexible automatic paddy rice mapping strategy was implemented. For undisturbed region, an object-oriented CCVS method was utilized to take the full advantages of phenology-based method. For disturbed region, Random Forest (RF) classifier was exploited using training data from CCVS-derived results in undisturbed region and feature images adaptively selected with full considerations of spectral separability and the spatiotemporal coverage. The ARMM method was verified by 473 reference sites, with an overall accuracy of 95.77% and kappa index of 0.9107. This study provided an efficient strategy to accommodate the challenges of phenology-based approaches through transferring knowledge in parts of a satellite scene with finer time series to targets (other parts) with deficit data availability.