Chapter
To read the full-text of this research, you can request a copy directly from the authors.

Abstract

Time series processing is an important ingredient of a biophysical algorithm in order to get the expected continuous and smooth dynamics required by many applications. Several temporal techniques have been proposed to reduce noise and fill gaps in the time series of satellite data. The choice of the compositing method may have a large impact on the accuracy of the phenology extracted from the reconstructed time series. This chapter presents a comparison of six methods to improve the temporal coherence and continuity of leaf area index (LAI) time series. The temporal smoothing gap filling (TSGF) method which is based on an adaptive Savitzky-Golay filter combined with a linear interpolation approach for filling gaps over a limited temporal window showed the best performance when applied to time series with less than 60 % of gaps. A climatology based approach outperformed other approaches for filling gaps in time series with more than 60 % of missing data or when the period of missing data is longer than 100 days. Based on these findings, a dedicated approach combining the local TSGF filter with a climatology gap filling technique was developed. It constitutes the basis of the algorithm for the operational production of continuous and smooth time series of biophysical variables from VEGETATION data within the European Copernicus Global Land Service.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

... It ranges between − 1 and +1 with high values close to + 1 indicating dense and phenologically active vegetation surfaces and values near to 0 and lower indicating non-vegetated or phenologically non-active surfaces. NDVI observations collected as time series facilitate the measurement of spectral-temporal profiles (e.g., Förster et al., 2012;Verger et al., 2016), phenological chronologies (e.g., Gerstmann et al., 2016) or trends and abrupt changes (e.g., Verbesselt et al., 2010), which is especially useful for crop monitoring on the pixel-or parcel-levels. ...
... The parameter corresponds to the similarity of the observed time series with an averaged (i.e., simulated) time series for catch crop cultivation (Verger et al., 2016;Heupel et al., 2019). The value is expected to be high with a successful catch crop cultivation. ...
... The parameter corresponds to the similarity of the observed time series with an averaged (i.e., simulated) time series for catch crop cultivation (Verger et al., 2016; (continued on next page) development and early die-off of catch crops (Deutscher Wetterdienst, 2016) especially influenced the quality of the BB16 training data set. Thus, under specific annual weather conditions such as heat waves or early frost, phenological stages and growth rates become less pronounced in the NDVI signal and the chosen method may work more weakly. ...
Article
The European legislative guidelines on the Common Agricultural Policy (CAP) lead to an obligation of area-wide information about cross-compliant Greening measures in EU countries. But by on-site controlling, agricultural authorities can monitor only a small portion of all registered parcels. With the Copernicus Programme, the freely available Sentinel-1 and Sentinel-2 imagery increasingly supports remote sensing-based agricultural monitoring on a large-scale. However, most prototypes lack the topic of winter catch cropping. Therefore, we developed a new machine learning method for catch crop monitoring and detection at the parcel-level. To gain training and test data for supervised machine learning, we collected winter catch crop parcel data from different years (2016–2019) and four federal states in Germany. Normalized Difference Vegetation Index (NDVI) time series were calculated for each parcel from Sentinel-2 data within the typical winter catch crop cultivation season (July-April). We revealed distinctive temporal patterns of catch cropping and developed nineteen descriptive features for automatized prediction. Then, we trained fifteen Random Forest classifiers comprising different regions and years and conducted a multi-level validation to identify the model with the highest robustness on new data. The Random Forest classifier trained with the input data from all federal states and years outperformed the other models. It reached a mean prediction accuracy of 84% for both classes (catch crop and non-catch crop) across eleven different spatio-temporal domains. Under optimal annual weather conditions it reached accuracies close to 90%. Anomalies caused by heat waves and early frost events were found to have a high influence on the phenology of catch crops and thus lead to reduced prediction accuracies. From the set of predictors, those features with the highest importance measured the correlation between the observed time series and a simulated NDVI phenological profile. We concluded that catch cropping parcels are automatically separable from parcels with other winter cultivations (e.g., winter cereals, grasslands, fallow) with Sentinel-2 NDVI time series data. Different catch crop subgroups (i.e., seed mixes) could not be differentiated by our approach due to very similar phenological profiles. Nonetheless, the used approach allows for large-scale winter catch crop monitoring and supports authorities in the selection of parcels with high demand for on-site controlling. By merging the training datasets from different federal states and years, we could overcome the typical spatial and temporal overfitting problem in machine learning. Therefore, the study’s final classifier can be reliably transferred to new datasets in Germany and other regions with similar bio-geographical conditions.
... First, we calculated a phenological reference time series (TS_ref) from a selected subset of 100 reference fields covered with catch crops in RP from which an spatial average value was calculated. This phenological reference profile (TS_ref) for specific catch crops can be used for the classification of other time series within the same time frame [7]. Second, we calculated time series stacks for a subset of field borders data in RP. ...
Conference Paper
Full-text available
On-site controlling of subsidy measures in rural areas are cost-and time-intensive for authorities and can only be done for a small subset of subsidized areas within short time frames. The use of high resolution satellite image time series covers many applications in the field of agricultural monitoring. We present a method for an automated control system based on vegetation indices from Sentinel-2 and Sentinel-1 image time series. We calculated a catch crop phenological reference profile (TS_ref) from a subset of sampling sites in the German federal state Rhineland-Palatinate and compared it with observed Sentinel time series (TS_obs) via Pearson's correlation coefficient. The utilization of the classification results reduces on-site controls noticeably and helps to define areas with higher and lower control demands.
Article
Full-text available
The ESA-GlobCover initiative has delivered to the international community the very first 300 m global land cover map for 2005. This product is derived from 40 terabytes of imagery acquired by Envisat's MERIS instrument for the period from December 2004 to June 2006. This product has been successfully validated by an independent network of international experts through a global accuracy assessment. Unlike other land cover initiatives, the GlobCover system was developed as an operational service to be automated from the level 1b imagery to the final land cover map, including all the pre-processing steps and the classification process. A globally consistent but regionally-tuned classification process was designed to take most advantage of the available data. The 22- classes legend based on the FAO Land Cover Classification System is compatible with the previous 1-km GLC2000 land cover map.
Article
Full-text available
The NRT algorithm for near-real time estimation of global LAI, FAPAR, and FCOVER variables from SPOT/VEGETATION (VGT) satellite data is described here. It consists of three steps: 1) neural networks (NNT) (one for each variable) to provide instantaneous estimates from daily VGT-P reflectances; 2) a multistep filtering approach to eliminate data mainly affected by atmospheric effects and snow cover; and 3) Savitzky-Golay and climatology temporal smoothing and gap filling techniques to ensure consistency and continuity as well as short-term projection of the product dynamics. Performances of NRT estimates were evaluated by comparing with other products over the 2005-2008 period: 1) the offline estimates from the application of the algorithm over historical time series (HIST); 2) the geoland2 version 1 products also issued from VGT (GEOV1/VGT); and 3) ground data. NRT rapidly converges closely to the HIST processing after six dekads (10-day period) with major improvement after two dekads. Successive reprocessing will, therefore, correct for some instabilities observed in the presence of noisy and missing data. The root-mean-square error (RMSE) between NRT and HIST LAI is lower than 0.4 in all cases. It shows a rapid exponential decay with the number of observations in the composition window with convergence when 30 observations are available. NRT products are in good agreement with ground data (RMSE of 0.69 for LAI, 0.09 for FAPAR, and 0.14 for FCOVER) and consistent with GEOV1/VGT products with a significant improvement in terms of continuity (only 1% of missing data) and smoothness, especially at high latitudes, and Equatorial areas.
Article
Full-text available
Long term global terrestrial vegetation monitoring from satellite Earth Observation system is a critical issue within global climate and earth science modeling applications. This contribution describes the algorithm and a first validation of the global LAI, FAPAR and FCOVER products derived from AVHRR Long Term Data Record (LTDR) for the 1981-2000 period within geoland2 project. GEOV1_AVHRR algorithm aims to ensure robustness and consistency of the derived biophysical products with the ones developed in the recent years, and particularly with the GEOV1_VGT products derived from SPOT/VGT sensor. The approach is based on the use of neural networks and dedicated temporal smoothing and gap filing techniques. The comparison of GEOV1_AVHRR with GEOV1_VGT and with similar products derived from AVHRR GIMMS shows that GEOV1_AVHRR outperforms the current existing products in terms of smoothness and continuity (fraction of valid data of 99.7%). At the same time, GEOV1_AVHRR shows high consistency with GEOV1_VGT and a similar accuracy (RMSE=0.80) as evaluated with ground-based observations. Combination of GEOV1_AVHRR and GEOV1_VGT realizes thus the objective of producing continuous and consistent time series of global observations of LAI, FAPAR and FCOVER for the last three decades. In the next months GEOV1_AVHRR products will be freely accessible at geoland2 portal at 0.05º spatial resolution and 10 days temporal sampling frequency.
Article
Full-text available
Consistent, continuous, and long time series of global biophysical variables derived from satellite data are required for global change research. A novel climatology fitting approach called CACAO (Consistent Adjustment of the Climatology to Actual Observations) is proposed to reduce noise and fill gaps in time series by scaling and shifting the seasonal climatological patterns to the actual observations. The shift and scale CACAO parameters adjusted for each season allow quantifying shifts in the timing of seasonal phenology and inter-annual variations in magnitude as compared to the average climatology. CACAO was assessed first over simulated daily Leaf Area Index (LAI) time series with varying fractions of missing data and noise. Then, performances were analyzed over actual satellite LAI products derived from AVHRR Long-Term Data Record for the 1981–2000 period over the BELMANIP2 globally representative sample of sites. Comparison with two widely used temporal filtering methods—the asymmetric Gaussian (AG) model and the Savitzky-Golay (SG) filter as implemented in TIMESAT—revealed that CACAO achieved better performances for smoothing AVHRR time series characterized by high level of noise and frequent missing observations. The resulting smoothed time series captures well the vegetation dynamics and shows no gaps as compared to the 50–60$%$ of still missing data after AG or SG reconstructions. Results of simulation experiments as well as confrontation with actual AVHRR time series indicate that the proposed CACAO method is more robust to noise and missing data than AG and SG methods for phenology extraction.
Article
Full-text available
The Best Index Slope Extraction (BISE) is proposed as an alternative to the Maximum Value Composite (MVC) as a way of reducing noise in Normalized Difference Vegetation Index (NDVI) time-series from Advanced Very High Resolution Radiometer (AVHRR) data. The BISE method retains more valuable elements of an NDVI profile than the MVC. NDVI profiles prepared using the BISE method show strong differences in the development characteristics of various West African vegetation types.
Article
Full-text available
Long-term global data sets of vegetation Leaf Area Index (LAI) and Fraction of Photosynthetically Active Radiation absorbed by vegetation (FPAR) are critical to monitoring global vegetation dynamics and for modeling exchanges of energy, mass and momentum between the land surface and planetary boundary layer. LAI and FPAR are also state variables in hydrological, ecological, biogeochemical and crop-yield models. The generation, evaluation and an example case study documenting the utility of 30-year long data sets of LAI and FPAR are described in this article. A neural network algorithm was first developed between the new improved third generation Global Inventory Modeling and Mapping Studies (GIMMS) Normalized Difference Vegetation Index (NDVI3g) and best-quality Terra Moderate Resolution Imaging Spectroradiometer (MODIS) LAI and FPAR products for the overlapping period 2000–2009. The trained neural network algorithm was then used to generate corresponding LAI3g and FPAR3g data sets with the following attributes: 15-day temporal frequency, 1/12 degree spatial resolution and temporal span of July 1981 to December 2011. The quality of these data sets for scientific research in other disciplines was assessed through (a) comparisons with field measurements scaled to the spatial resolution of the data products, (b) comparisons with broadly-used existing alternate satellite data-based products, (c) comparisons to plant growth limiting climatic variables in the northern latitudes and tropical regions, and (d) correlations of dominant modes of interannual variability with large-scale circulation anomalies such as the EI Niño-Southern Oscillation and Arctic Oscillation. These assessment efforts yielded results that attested to the suitability of these data sets for research use in other disciplines. The utility of these data sets is documented by comparing the seasonal profiles of LAI3g with profiles from 18 state-of-the-art Earth System Models: the models consistently overestimated the satellite-based estimates of leaf area and simulated delayed peak seasonal values in the northern latitudes, a result that is consistent with previous evaluations of similar models with ground-based data. The LAI3g and FPAR3g data sets can be obtained freely from the NASA Earth Exchange (NEX) website.
Article
Full-text available
This study investigates the performances of four major global Leaf Area Index (LAI) products at 1/11.2° spatial sampling and a monthly time step: ECOCLIMAP climatology, GLOBCARBON (from SPOT/VEGETATION and ATSR/AATSR), CYCLOPES (from SPOT/VEGETATION) and MODIS Collection 4 (main algorithm, from MODIS/TERRA). These products were intercompared during the 2001-2003 period over the BELMANIP network of sites. Their uncertainty was assessed by comparison with 56 LAI reference maps derived from ground measurements. CYCLOPES and MODIS depict realistic spatial variations at continental scale, while ECOCLIMAP poorly captures surface spatial heterogeneity, and GLOBCARBON tends to display erratic variations. ECOCLIMAP and GLOBCARBON show the highest frequency of successful retrievals while MODIS and CYCLOPES retrievals are frequently missing in winter over northern latitudes and over the equatorial belt. CYCLOPES and MODIS describe consistent temporal profiles over most vegetation types, while ECOCLIMAP does not show any interannual variations, and GLOBCARBON can exhibit temporal instability during the growing season over forests. The CYCLOPES, MODIS, and GLOBCARBON LAI values agree better over croplands and grasslands than over forests, where differences in vegetation structure representation between algorithms and surface reflectance uncertainties lead to substantial discrepancies between products. CYCLOPES does not reach high enough LAI values to properly characterize forests. In contrast, the other products have sufficient dynamic range of LAI to describe the global variability of LAI. Overall, CYCLOPES is the most similar product to the LAI reference maps. However, more accurate ground measurements and better representation of the global and seasonal variability of vegetation are required to refine this result.
Conference Paper
Full-text available
This paper proposes empirical mode decomposition (EMD) based pre-process to increase classification accuracy of hyperspectral images. EMD is an adaptive and non-linear signal decomposition approach and decomposes the data into intrinsic mode functions (IMFs) and a residue. In this paper, EMD is applied to each hyperspectral image band to obtain IMFs. After EMD is performed to each band, new bands are reconstructed as the sum of higher level IMFs and classification is executed over these new bands. Support vector machine (SVM) is used to show the classification performance of the proposed approach. Experimental results show that, utilization of the first two IMFs significantly increases the classification accuracy compared to applying SVM directly to the original data set.
Article
Full-text available
Annual, inter-annual and long-term trends in time series derived from remote sensing can be used to distinguish between natural land cover variability and land cover change. However, the utility of using NDVI-derived phenology to detect change is often limited by poor quality data resulting from atmospheric and other effects. Here, we present a curve fitting methodology useful for time series of remotely sensed data that is minimally affected by atmospheric and sensor effects and requires neither spatial nor temporal averaging. A two-step technique is employed: first, a harmonic approach models the average annual phenology; second, a spline-based approach models inter-annual phenology. The principal attributes of the time series (e.g., amplitude, timing of onset of greenness, intrinsic smoothness or roughness) are captured while the effects of data drop-outs and gaps are minimized. A recursive, least squares approach captures the upper envelope of NDVI values by upweighting data values above an average annual curve. We test this methodology on several land cover types in the western U.S., and find that onset of greenness in an average year varied by less than 8 days within land cover types, indicating that the curve fit is consistent within similar systems. Between 1990 and 2002, temporal variability in onset of greenness was between 17 and 35 days depending on the land cover type, indicating that the inter-annual curve fit captures substantial inter-annual variability. Employing this curve fitting procedure enhances our ability to measure inter-annual phenology and could lead to better understanding of local and regional land cover trends.
Article
Full-text available
Current models of vegetation dynamics using the normalized vegetation index (NDVI) time series perform poorly for high-latitude environments. This is due partly to specific attributes of these environments, such as short growing season, long periods of darkness in winter, persistence of snow cover, and dominance of evergreen species, but also to the design of the models. We present a new method for monitoring vegetation activity at high latitudes, using Moderate Resolution Imaging Spectroradiometer (MODIS) NDVI. It estimates the NDVI of the vegetation during winter and applies a double logistic function, which is uniquely defined by six parameters that describe the yearly NDVI time series. Using NDVI data from 2000 to 2004, we illustrate the performance of this method for an area in northern Scandinavia (35 × 162 km2, 68° N 23° E) and compare it to existing methods based on Fourier series and asymmetric Gaussian functions. The double logistic functions describe the NDVI data better than both the Fourier series and the asymmetric Gaussian functions, as quantified by the root mean square errors. Compared with the method based on Fourier series, the new method does not overestimate the duration of the growing season. In addition, it handles outliers effectively and estimates parameters that are related to phenological events, such as the timing of spring and autumn. This makes the method most suitable for both estimating biophysical parameters and monitoring vegetation phenology.
Article
Full-text available
Vegetation typically elicits dynamics at the seasonal and annual level. Time-series of normalized difference vegetation index (NDVI) datasets, such as the pathfinder AVHRR land (PAL) NDVI dataset, have proven to be appropriate for the detection of long-term vegetation cover changes. It has been applied in modelling experiments for terrestrial ecosystems at the global, continental, and regional scales. But some PAL NDVI time series remain significant residual effects and noise levels. A simple method, the mean-value iteration filter (MVI), has been developed to reduce the noise and to enable the reconstruction of high quality NDVI time-series. A comparison between the newly developed method and other existing methods (the modified BISE algorithm and a fast Fourier transform algorithm) indicates that the newly developed method is an effective tool for reconstructing high-quality time series of PAL NDVI time series.
Article
Full-text available
Although the Normalized Difference Vegetation Index (NDVI) time-series data, derived from NOAA/AVHRR, SPOT/VEGETATION, TERRA or AQUA/MODIS, has been successfully used in research regarding global environmental change, residual noise in the NDVI time-series data, even after applying strict pre-processing, impedes further analysis and risks generating erroneous results. Based on the assumptions that NDVI time-series follow annual cycles of growth and decline of vegetation, and that clouds or poor atmospheric conditions usually depress NDVI values, we have developed in the present study a simple but robust method based on the Savitzky–Golay filter to smooth out noise in NDVI time-series, specifically that caused primarily by cloud contamination and atmospheric variability. Our method was developed to make data approach the upper NDVI envelope and to reflect the changes in NDVI patterns via an iteration process. From the results obtained by applying the newly developed method to a 10-day MVC SPOT VGT-S product, we provide optimized parameters for the new method and compare this technique with the BISE algorithm and Fourier-based fitting method. Our results indicate that the new method is more effective in obtaining high-quality NDVI time-series.
Article
Full-text available
Satellite-derived NDVI time series are fundamental to the remote sensing of vegetation phenology, but their application is hindered by prevalent noise resulting chiefly from varying atmospheric conditions and sun-sensor-surface viewing geometries. A model-based empirical comparison of six selected NDVI time series noise-reduction techniques revealed the general superiority of the double logistic and asymmetric Gaussian function-fitting methods over four alternative filtering techniques. However, further analysis demonstrated the strong influence of noise level, strength, and bias, and the extraction of phenological variables on technique performance. Users are strongly cautioned to consider both their ultimate objectives and the nature of the noise present in an NDVI data set when selecting an approach to noise reduction, particularly when deriving phenological variables.
Article
Full-text available
A new method for analysing nonlinear and non-stationary data has been developed. The key part of the method is the `empirical mode decomposition' method with which any complicated data set can be decomposed into a finite and often small number of 'intrinsic mode functions' that admit well-behaved Hilbert transforms. This decomposition method is adaptive, and, therefore, highly efficient. Since the decomposition is based on the local characteristic time scale of the data, it is applicable to nonlinear and non-stationary processes. With the Hilbert transform, the 'instrinic mode functions' yield instantaneous frequencies as functions of time that give sharp identifications of imbedded structures. The final presentation of the results is an energy-frequency-time distribution, designated as the Hilbert spectrum. In this method, the main conceptual innovations are the introduction of `intrinsic mode functions' based on local properties of the signal, which make the instantaneous frequency meaningful; and th
Article
Full-text available
This paper describes the harmonic analysis of time series (HANTS) algorithm. It performs two tasks: screening and removal of cloud-affected observations, and temporal interpolation of the remaining observations to reconstruct gapless images at a prescribed time. HANTS was applied to 36 AVHRR 10-days-maximum-NDVI composites covering most of Europe. The results show that cloud-affected data are recognized successfully and replaced. Up to half the data points were rejected with no consequence for the successful reconstruction of seasonal NDVI profiles.
Article
Full-text available
Ecological and climate models require high-quality consistent biophysical parameters as inputs and validation sources. NASA's moderate resolution imaging spectroradiometer (MODIS) biophysical products provide such data and have been used to improve our understanding of climate and ecosystem changes. However, the MODIS time series contains occasional lower quality data, gaps from persistent clouds, cloud contamination, and other gaps. Many modeling efforts, such as those used in the North American Carbon Program, that use MODIS data as inputs require gap-free data. This letter presents the algorithm used within the MODIS production facility to produce temporally smoothed and spatially continuous biophysical data for such modeling applications. We demonstrate the algorithm with an example from the MODIS-leaf-area-index (LAI) product. Results show that the smoothed LAI agrees with high-quality MODIS LAI very well. Higher R-squares and better linear relationships have been observed when high-quality retrieval in each individual tile reaches 40% or more. These smoothed products show similar data quality to MODIS high-quality data and, therefore, can be substituted for low-quality retrievals or data gaps.
Article
Full-text available
This study investigates the representativeness of land cover and leaf area index (LAI) sampled by a global network of sites to be used for the evaluation of land biophysical products, such as LAI or fAPAR, derived from current satellite systems. The networks of sites considered include 100 sites where ground measurements of LAI or fAPAR have been performed for the validation of medium resolution satellite land biophysical products, 188 FLUXNET sites and 52 AERONET sites. All the sites retained had less than 25% of water bodies within a 8times8 km <sup>2</sup> window, and were separated by more than 20 km. The ECOCLIMAP global classification was used to quantify the representativeness of the networks. It allowed describing the Earth's surface with seven main types and proposed a climatology for monthly LAI values at a spatial resolution around 1 km. The site distribution indicates a large over representation of the northern midlatitudes relative to other regions, and an under-representation of bare surfaces, grass, and evergreen broadleaf forests. These three networks represent all together 295 sites after elimination of sites that were too close. They were thus completed by 76 additional sites to improve the representativeness in latitude, longitude, and surface type. This constitutes the BELMANIP network proposed as a benchmark for intercomparison of land biophysical products. Suitable approaches to conducting intercomparison at the sites are recommended
Article
Land-surface modelling would benefit significantly from improved characterisation of the seasonal variability of vegetation at a global scale. GEOCLIM, a global climatology of leaf area index (LAI), fraction of absorbed photosynthetically active radiation (FAPAR)—both essential climate variables—and fraction of vegetation cover (FCOVER), is here derived from observations from the SPOT VEGETATION programme. Interannual average values from the GEOV1 Copernicus Global Land time series of biophysical products at 1-km resolution and 10-day frequency are computed for 1999 to 2010. GEOCLIM provides the baseline characteristics of the seasonal cycle of the annual vegetation phenology for each 1-km pixel on the globe. The associated standard deviation characterises the interannual variability. Temporal consistency and continuity is achieved by the accumulation of multi-year observations and the application of techniques for temporal smoothing and gap filling. Specific corrections are applied over cloudy tropical regions and high latitudes in the Northern Hemisphere where the low number of available observations compromises the reliability of estimates. Artefacts over evergreen broadleaf forests and areas of bare soil are corrected based on the expected limited seasonality. The GEOCLIM data set is demonstrated to be consistent, both spatially and temporally. GEOCLIM shows absolute differences lower than 0.5 compared with MODIS (GIMMS3g) climatology of LAI for more than 80% (90%) of land pixels, with higher discrepancies in tropical and boreal latitudes. ECOCLIMAP systematically produces higher LAI values. The phenological metric for the date of maximum foliar development derived from GEOCLIM is spatially consistent (correlation higher than 0.9) with those of MODIS, GIMMS3g, ECOCLIMAP and MCD12Q2 with average differences within 14 days at the global scale.
Article
In attempting to analyze, on digital computers, data from basically continuous physical experiments, numerical methods of performing familiar operations must be developed. The operations of differentiation and filtering are especially important both as an end in themselves, and as a prelude to further treatment of the data. Numerical counterparts of analog devices that perform these operations, such as RC filters, are often considered. However, the method of least squares may be used without additional computational complexity and with considerable improvement in the information obtained. The least squares calculations may be carried out in the computer by convolution of the data points with properly chosen sets of integers. These sets of integers and their normalizing factors are described and their use is illustrated in spectroscopic applications. The computer programs required are relatively simple. Two examples are presented as subroutines in the FORTRAN language.
Article
Several models have been fitted in the past to smooth time-series vegetation index data from different satellite sensors to estimate vegetation phenological parameters. However, differences between the models and fine tuning of model parameters lead to potential differences, uncertainty and bias between the results amongst users. The current research assessed four techniques: Fourier analysis, asymmetric Gaussian model, double logistic model and the Whittaker filter for smoothing multi-temporal satellite sensor observations with the ultimate purpose of deriving an appropriate annual vegetation growth cycle and estimating phenological parameters reliably. The research used Level 3 Medium Resolution Imaging Spectrometer (MERIS, spatial resolution ~ 4.6 km) Terrestrial Chlorophyll Index (MTCI) data over the years 2004 to 2006 composited at eight day intervals covering the Indian sub-continent. First, the four models were fitted to representative sample time-series of the major vegetation types in India, and the quality of the fit was analysed. Second, the effect of noise on model fitting was analysed by adding Gaussian noise to a standard profile. Finally, the four models were fitted to the whole study area to characterise variation in the quality of model fitting as a function of single and double vegetation seasons. These smoothed data were used to estimate the onset of greenness (OG), a major phenological parameter. The models were evaluated using the root mean square error (RMSE), Akaike Information Criteria (AIC), and Bayesian Information Criteria (BIC). The first test (fitting to representative sample time series) revealed the consistently superior performance of the Whittaker and Fourier approaches in most cases. The second test (fitting after the addition of Gaussian noise) revealed the superior performance of the double logistic and Fourier approaches. Finally, when the approaches were applied to the whole study, thus, including vegetation with different phenological profiles and multiple growing seasons (third test), it was found that it was necessary to tune each of the models according to the number of annual growing seasons to produce reliable fits. The double logistic and asymmetric Gaussian models did not perform well for areas with more than one growing season per year. The mean absolute deviation in OG derived from these models was a maximum (3 to 4 weeks) within the dry deciduous vegetation type and minimum (1 week) in evergreen vegetation. All techniques yielded consistent results over the south-western and north-eastern regions of India characterised by tropical climate.
Article
Nonlinear data smoothers provide a practical method of finding smooth traces for data confounded with possibly long-tailed or occasionally spikey noise. They are resistant to the effects of extreme observations that are not part of the local pattern, yet they are able to respond rapidly to well-supported patterns. This article defines a collection of nonlinear smoothers based upon running medians and presents methods for describing and comparing their performance. The characterizations of the smoothers presented here reveal some with excellent low-pass transfer characteristics, negligible Gibbs rebound, and resistance to the effects of non-Gaussian disturbances.
Article
Atmospheric CO2 measurements from the NOAA Geophysical Monitoring for Climatic Change program at the Mauna Loa Observatory are presented for the period 1974-1985. Hourly and daily variations in CO2 concentration due to local sources and sinks and the long-term increase in CO2 are examined. IT is shown that the seasonal cycle was increasing at a rate of 0.05 + or - 0.02 ppm/yr. The average CO2 growth rate for the period is 1.42 + or - 0.02 ppm/yr. The fraction of CO2 remaining in the atmosphere from fossil fuel combustion is found to be 59 percent.
Article
The well-known and popular Savitzky-Golay filter has several disadvantages. A very attractive alternative is a smoother based on penalized least squares, extending ideas presented by Whittaker 80 years ago. This smoother is extremely fast, gives continuous control over smoothness, interpolates automatically, and allows fast leave-one-out cross-validation. It can be programmed in a few lines of Matlab code. Theory, implementation, and applications are presented.
Article
A synergistic algorithm for producing global leaf area index and fraction of absorbed photosynthetically active radiation fields from canopy reflectance data measured by MODIS (moderate resolution imaging spectroradiometer) and MISR (multiangle imaging spectroradiometer) instruments aboard the EOS-AM 1 platform is described here. The proposed algorithm is based on a three-dimensional formulation of the radiative transfer process in vegetation canopies. It allows the use of information provided by MODIS (single angle and up to 7 shortwave spectral bands) and MISR (nine angles and four shortwave spectral bands) instruments within one algorithm. By accounting features specific to the problem of radiative transfer in plant canopies, powerful techniques developed in reactor theory and atmospheric physics are adapted to split a complicated three-dimensional radiative transfer problem into two independent, simpler subproblems, the solutions of which are stored in the form of a look-up table. The theoretical background required for the design of the synergistic algorithm is discussed.
Article
To eliminate the confusion in the definition of leaf area index (L) for non-flat leaves, the projection coefficients of several objects including spheres, cylinders, hemicircular cylinders, and triangular and square bars are investigated through mathematical derivation and numerical calculation for a range of ellipsoidal angular distributions. It is shown that the projection coefficient calculated based on half the total intercepting area is close to a constant of 0.5 when the inclination angle of the objects is randomly (spherically) distributed, whereas the calculated results based on the object's largest projected area are strongly dependent on the shape of the objects. Therefore, it is suggested that the leaf area index of non-flat leaves be defined as half the total intercepting area per unit ground surface area and that the definition of L based on the projected leaf area be abandoned.
Article
This paper describes a correction of directional effects in AVHRR reflectance time series. The method relies on a priori directional signatures, in the red and near infrared, for various surface covers. The reflectance normalization is applied to the historical Pathfinder AVHRR Land (PAL) data set acquired between 1981 and 1999. The high frequency variability in the reflectance time series, which is interpreted as a noise due to varying observation geometry and directional effects, is reduced by a factor greater than 1.7 for most vegetated surfaces. For the analysis of the vegetation annual cycle and its inter-annual variations, we recommend the use of the Difference Vegetation Index (DVI), following a prior correction of the directional effects, instead of the commonly used NDVI. Indeed, DVI appears more robust than NDVI to noise resulting from atmospheric perturbations, which makes it a good candidate for long-term ecological surveys and climate change studies.
Article
Leaf Area Index (LAI) is one of the most important variables characterizing land surface vegetation and dynamics. Many satellite data, such as the Moderate Resolution Imaging Spectroradiometer (MODIS), have been used to generate LAI products. It is important to characterize their spatial and temporal variations by developing mathematical models from these products. In this study, we aim to model MODIS LAI time series and further predict its future values by decomposing the LAI time series of each pixel into several components: trend, intra-annual variations, seasonal cycle, and stochastic stationary or irregular parts. Three such models that can characterize the non-stationary time series data and predict the future values are explored, including Dynamic Harmonics Regression (DHR), STL (Seasonal-Trend Decomposition Procedure based on Loess), and Seasonal ARIMA (AutoRegressive Intergrated Moving Average) (SARIMA). The preliminary results using six years (2001–2006) of the MODIS LAI product indicate that all these methods are effective to model LAI time series and predict 2007 LAI values reasonably well. The SARIMA model gives the best prediction, DHR produces the smoothest curve, and STL is more sensitive to noise in the data. These methods work best for land cover types with pronounced seasonal variations.
Article
This article describes the algorithmic principles used to generate LAI, fAPAR and fCover estimates from VEGETATION observations. These biophysical variables are produced globally at 10 days temporal sampling interval under lat–lon projection at 1/112° spatial resolution. After a brief description of the VEGETATION sensors, radiometric calibration process, based on vicarious desertic targets is first presented. The cloud screening algorithm was then fine tuned using a global network of cloudiness observations. Atmospheric correction is then achieved using the SMAC code with inputs coming from meteorological values of pressure, ozone and water vapour. Aerosol optical thickness is derived from MODIS climatology assuming continental aerosol type. The Roujean BRDF model is then adjusted for red, near infrared and short wave infrared bands used to the remaining cloud free observations collected over a time window of ± 15 days. Outliers due to possible cloud contamination or residual atmospheric correction are iteratively eliminated and prior information is used to get more robust estimates of the three BRDF kernel coefficients. Nadir viewing top of canopy reflectance in the three bands is input to the biophysical algorithm to compute the products at 10 days sampling interval. This algorithm is based on training neural networks over SAIL + PROPSPECT radiative transfer model simulations for each biophysical variable. Details on the way the training data base was generated and the neural network designed and calibrated are presented. Finally, theoretical performances are discussed. Validation over ground measurement data sets and inter-comparison with other similar biophysical products are presented and discussed in a companion paper. The CYCLOPES products and associated detailed documentation are available at http://postel.mediasfrance.org.
Article
Accurate measurements of regional to global scale vegetation dynamics (phenology) are required to improve models and understanding of inter-annual variability in terrestrial ecosystem carbon exchange and climate–biosphere interactions. Since the mid-1980s, satellite data have been used to study these processes. In this paper, a new methodology to monitor global vegetation phenology from time series of satellite data is presented. The method uses series of piecewise logistic functions, which are fit to remotely sensed vegetation index (VI) data, to represent intra-annual vegetation dynamics. Using this approach, transition dates for vegetation activity within annual time series of VI data can be determined from satellite data. The method allows vegetation dynamics to be monitored at large scales in a fashion that it is ecologically meaningful and does not require pre-smoothing of data or the use of user-defined thresholds. Preliminary results based on an annual time series of Moderate Resolution Imaging Spectroradiometer (MODIS) data for the northeastern United States demonstrate that the method is able to monitor vegetation phenology with good success.
Article
Leaf Area Index (LAI) is an important biophysical variable for characterizing the land surface vegetation. Global LAI product has been routinely produced from the MODerate resolution Imaging Spectroradiometer (MODIS) aboard the Terra and Aqua satellite platforms. However, the MODIS standard LAI product is not continuous both spatially and temporally. To fill the gaps and improve the quality, we have developed a data filtering algorithm. This filter, called the temporal spatial filter (TSF), integrates both spatial and temporal characteristics for different plant functional types. The spatial gaps are first filled with the multi-year averages of the same day. If the values are missing over all years, the pixel is filled with a new estimate using the vegetation continuous field–ecosystem curve fitting method. The TSF integrates both the multi-seasonal average trend (background) and the seasonal observation. We implement this algorithm using the MODIS Collection 4 LAI product over North America. Comparison of the TSF results with the Savitzky–Golay filter indicates that the TSF performs much better in restoring the spatial and temporal distribution of seasonal LAI trends. The new LAI product has been validated by comparing with field measurements and the derived LAI maps from ETM+ data at a broadleaf forest site and an agricultural site. The validation results indicate that the new LAI product agrees better with both the field measurements and LAI values obtained from the ETM+ than does the MODIS LAI standard product, which usually shows higher LAI values.
Article
The generation of multi-decade long Earth System Data Records (ESDRs) of Leaf Area Index (LAI) and Fraction of Photosynthetically Active Radiation absorbed by vegetation (FPAR) from remote sensing measurements of multiple sensors is key to monitoring long-term changes in vegetation due to natural and anthropogenic influences. Challenges in developing such ESDRs include problems in remote sensing science (modeling of variability in global vegetation, scaling, atmospheric correction) and sensor hardware (differences in spatial resolution, spectral bands, calibration, and information content). In this paper, we develop a physically based approach for deriving LAI and FPAR products from the Advanced Very High Resolution Radiometer (AVHRR) data that are of comparable quality to the Moderate resolution Imaging Spectroradiometer (MODIS) LAI and FPAR products, thus realizing the objective of producing a long (multi-decadal) time series of these products. The approach is based on the radiative transfer theory of canopy spectral invariants which facilitates parameterization of the canopy spectral bidirectional reflectance factor (BRF). The methodology permits decoupling of the structural and radiometric components and obeys the energy conservation law. The approach is applicable to any optical sensor, however, it requires selection of sensor-specific values of configurable parameters, namely, the single scattering albedo and data uncertainty. According to the theory of spectral invariants, the single scattering albedo is a function of the spatial scale, and thus, accounts for the variation in BRF with sensor spatial resolution. Likewise, the single scattering albedo accounts for the variation in spectral BRF with sensor bandwidths. The second adjustable parameter is data uncertainty, which accounts for varying information content of the remote sensing measurements, i.e., Normalized Difference Vegetation Index (NDVI, low information content), vs. spectral BRF (higher information content). Implementation of this approach indicates good consistency in LAI values retrieved from NDVI (AVHRR-mode) and spectral BRF (MODIS-mode). Specific details of the implementation and evaluation of the derived products are detailed in the second part of this two-paper series.
Article
This study compared aspatial and spatial methods of using remote sensing and field data to predict maximum growing season leaf area index (LAI) maps in a boreal forest in Manitoba, Canada. The methods tested were orthogonal regression analysis (reduced major axis, RMA) and two geostatistical techniques: kriging with an external drift (KED) and sequential Gaussian conditional simulation (SGCS). Deterministic methods such as RMA and KED provide a single predicted map with either aspatial (e.g., standard error, in regression techniques) or limited spatial (e.g., KED variance) assessments of errors, respectively. In contrast, SGCS takes a probabilistic approach, where simulated values are conditional on the sample values and preserve the sample statistics. In this application, canonical indices were used to maximize the ability of Landsat ETM+ spectral data to account for LAI variability measured in the field through a spatially nested sampling design. As expected based on theory, SGCS did the best job preserving the distribution of measured LAI values. In terms of spatial pattern, SGCS preserved the anisotropy observed in semivariograms of measured LAI, while KED reduced anisotropy and lowered global variance (i.e., lower sill), also consistent with theory. The conditional variance of multiple SGCS realizations provided a useful visual and quantitative measure of spatial uncertainty. For applications requiring spatial prediction methods, we concluded KED is more useful if local accuracy is important, but SGCS is better for indicating global pattern. Predicting LAI from satellite data using geostatistical methods requires a distribution and density of primary, reference LAI measurements that are impractical to obtain. For regional NPP modeling with coarse resolution inputs, the aspatial RMA regression method is the most practical option.
Article
A neural network is developed to operationally estimate biophysical variables over land surfaces from the observations of the ENVISAT-MERIS instrument: the leaf area index (LAI), the fraction of absorbed photosynthetically active radiation (fAPAR), the fraction of vegetation cover (fCover), and the canopy chlorophyll content (LAI×Cab). The neural network requires as input the geometry of observation and the top of canopy reflectances, corrected from the atmospheric effects, in eleven spectral bands. It is trained on a reflectance database made of radiative transfer model simulations. The principles underlying the generation of the database and the design of the network are first presented. The estimated variables are then compared to other existing products, LAI- and fAPAR-MODIS and MGVI-MERIS, and validated against ground measurements performed in the framework of the VALERI project. Results show remarkable consistency of the temporal dynamics between the several products with however some differences in the range of variation. When compared to actual VALERI ground measurements, the proposed algorithm shows the best performances for LAI (RMSE = 0.47) and fAPAR (RMSE = 0.09).
Article
A multi-resolution analysis (MRA) based on the wavelet transform (WT) has been implemented to study NDVI time series. These series, which are non-stationary and present short-term, seasonal and long-term variations, can be decomposed using this MRA as a sum of series associated with different temporal scales. The main focus of the paper is to check the potential of this MRA to capture and describe both intra- and inter-annual changes in the data, i.e., to discuss the ability of the proposed procedure to monitor vegetation dynamics at regional scale. Our approach concentrates on what wavelet analysis can tell us about a NDVI time series. On the one hand, the intra-annual series, linked to the seasonality, has been used to estimate different key features related to the vegetation phenology, which depend on the vegetation cover type. On the other hand, the inter-annual series has been used to identify the trend, which is related to land-cover changes, and a Mann-Kendall test has been applied to confirm the significance of the observed trends. NDVI images from the MEDOKADS (Mediterranean Extended Daily One-km AVHRR Data Set) imagery series over Spain are processed according to a per-pixel strategy for this study. Results show that the wavelet analysis provides relevant information about vegetation dynamics at regional scale, such as the mean and minimum NDVI value, the amplitude of the phenological cycle, the timing of the maximum NDVI and the magnitude of the land-cover change. The latter, in combination with precipitation data, has been used to interpret the observed land-cover changes and identify those subtle changes associated to land degradation.
Article
The main objective of this paper is the validation of CYCLOPES version 3.1 LAI and fAPAR products. It is achieved by the comparison with MODIS collection 4 and 4.1 products and ECOCLIMAP LAI climatology over the BELMANIP representative set of sites, and with ground measurements over a limited set of sites. Great attention is paid to the consistency of the comparison: for the spatial dimension, product PSF appears to be the main aspect governing the spatial resolution at which the comparison has to be achieved. For CYCLOPES, a minimal size of the sites should be 3 km × 3 km2, while the optimal one is 10 km × 10 km2; regarding the temporal sampling interval and resolution, the problem is much easier to solve when assuming a relatively smooth time course of vegetation characteristics (8–16 days). Great care was also paid to the departure of products from the nominal definition, particularly for LAI where different scales of clumping have to be considered.Results showed that CYCLOPES and MODIS products have generally consistent seasonality, CYCLOPES being however characterized by a smoother temporal evolution as expected. Differences are mainly concentrated on the magnitude of products values, CYCLOPES achieving better performances both for LAI (RMSE = 0.73) and fAPAR (RMSE = 0.10) over the limited number of sites where ground measurements were available. This study also sets a framework to the validation exercise that could be used to evaluate other products or future versions of the same products and contribute to associate quantitative uncertainties as required by the user community.
Article
Red and near-infrared satellite data from the Advanced Very High Resolution Radiometer sensor have been processed over several days and combined to produce spatially continuous cloud-free imagery over large areas with sufficient temporal resolution to study green-vegetation dynamics. The technique minimizes cloud contamination, reduces directional reflectance and off-nadir viewing effects, minimizes sun-angle and shadow effects, and minimizes aerosol and water-vapor effects. The improvement is highly dependent on the state of the atmosphere, surface-cover type, and the viewing and illumination geometry of the sun, target and sensor. An example from southern Africa showed an increase of 40 percent from individual image values tothe final composite image. Limitations associated with the technique are discussed, and recommendations are given to improve this approach.
Article
Leaf area index (LAI) is one of the most important Earth surface parameters in modeling ecosystems and their interaction with climate. Based on a geometrical optical model (Four-Scale) and LAI algorithms previously derived for Canada-wide applications, this paper presents a new algorithm for the global retrieval of LAI where the bidirectional reflectance distribution function (BRDF) is considered explicitly in the algorithm and hence removing the need of doing BRDF corrections and normalizations to the input images. The core problem of integrating BRDF into the LAI algorithm is that nonlinear BRDF kernels that are used to relate spectral reflectances to LAI are also LAI dependent, and no analytical solution is found to derive directly LAI from reflectance data. This problem is solved through developing a simple iteration procedure. The relationships between LAI and reflectances of various spectral bands (red, near infrared, and shortwave infrared) are simulated with Four-Scale with a multiple scattering scheme. Based on the model simulations, the key coefficients in the BRDF kernels are fitted with Chebyshev polynomials of the second kind. Spectral indices - the simple ratio and the reduced simple ratio - are used to effectively combine the spectral bands for LAI retrieval. Example regional and global LAI maps are produced. Accuracy assessment on a Canada-wide LAI map is made in comparison with a previously validated 1998 LAI map and ground measurements made in seven Landsat scenes
Article
A new method for extracting seasonality information from time-series of satellite sensor data is presented. The method is based on nonlinear least squares fits of asymmetric Gaussian model functions to the time-series. The smooth model functions are then used for defining key seasonality parameters, such as the number of growing seasons, the beginning and end of the seasons, and the rates of growth and decline. The method is implemented in a computer program TIMESAT and tested on Advanced Very High Resolution Radiometer (AVHRR) normalized difference vegetation index (NDVI) data over Africa. Ancillary cloud data [clouds from AVHRR (CLAVR)] are used as estimates of the uncertainty levels of the data values. Being general in nature, the proposed method can be applied also to new types of satellite-derived time-series data.
Climatology of LAI, FAPAR and FCOVER products (V1) INRA-EMMAH
  • F Baret
  • M Weiss
Smoothing of NDVI time series curves for monitoring of vegetation changes in time
  • D Filipova-Racheva
  • M Hall-Beyer
GIO global land component - algorithm theorethical basis document - Leaf Area Index (LAI), Fraction of Absorbed Photosynthetically Active Radiation (FAPAR) and Fraction of green Vegetation Cover (FCover) Version 2.0 (GEOV2), Issue I1
  • A Verger
  • F Baret
  • M Weiss
Empirical mode decomposition pre-process for higher accuracy hyperspectral image classification. In: Geoscience and remote sensing symposium
  • B Demir
  • S Erturk
A comparison of methods for smoothing and gap filling time series of remote sensing observations: application to MODIS LAI products
  • S Kandasamy
  • F Baret
  • A Verger
  • P Neveux
  • M Weiss