A comparison of methods for smoothing and gap filling time series of remote sensing observations – application to MODIS LAI products

[摘要] Moderate resolution satellite sensors including MODIS (Moderate Resolution Imaging Spectroradiometer) already provide morethan 10 yr of observations well suited to describe and understand thedynamics of earth's surface. However, these time series are associated with significant uncertainties and incompletebecause of cloud cover. Thisstudy compares eight methods designed to improve the continuity by fillinggaps and consistency by smoothing the time course. It includes methodsexploiting the time series as a whole (iterative caterpillar singularspectrum analysis (ICSSA), empirical mode decomposition (EMD), low passfiltering (LPF) and Whittaker smoother (Whit)) as well as methods working onlimited temporal windows of a few weeks to few months (adaptive Savitzky–Golayfilter (SGF), temporal smoothing and gap filling (TSGF), and asymmetricGaussian function (AGF)), in addition to the simple climatological LAI yearlyprofile (Clim). Methods were applied to the MODIS leaf area index product forthe period 2000–2008and over 25 sites showed a large range of seasonalpatterns. Performances were discussed with emphasis on the balance achievedby each method between accuracy and roughness depending on the fraction ofmissing observations and the length of the gaps. Results demonstrate thatthe EMD, LPF and AGF methods were failing because of a significant fraction ofgaps (more than 20%), while ICSSA, Whit and SGF were always providingestimates for dates with missing data. TSGF (Clim) was able tofill more than 50% of the gaps for sites with more than 60%(80%) fraction of gaps. However, investigation of the accuracy of thereconstructed values shows that it degrades rapidly for sites with more than20% missing data, particularly for ICSSA, Whit and SGF. In theseconditions, TSGF provides the best performances that are significantly better thanthe simple Clim for gaps shorter than about 100 days. The roughness of thereconstructed temporal profiles shows large differences between the variousmethods, with a decrease of the roughness with the fraction of missing data,except for ICSSA. TSGF provides the smoothest temporal profiles for siteswith a % gap > 30%. Conversely, ICSSA, LPF, Whit, AGF and Climprovide smoother profiles than TSGF for sites with a % gap < 30%.Impact of the accuracy and smoothness of the reconstructed time series wereevaluated on the timing of phenological stages. The dates of start, maximumand end of the season are estimated with an accuracy of about 10 days forthe sites with a % gap < 10% and increases rapidly with the % gap.TSGF provides more accurate estimates of phenological timing up to a% gap < 60%.

[学科分类] 地球化学与岩石