Using Window Regression to Gap-Fill Landsat ETM+ Post SLC-Off Data

Abstract

The continued development of algorithms using multitemporal Landsat data creates opportunities to develop and adapt imputation algorithms to improve the quality of that data as part of preprocessing. One example is de-striping Enhanced Thematic Mapper Plus (ETM+, Landsat 7) images acquired after the Scan Line Corrector failure in 2003. In this study, we apply window regression, an algorithm that was originally designed to impute low-quality Moderate Resolution Imaging Spectroradiometer (MODIS) data, to Landsat Analysis Ready Data from 2014-2016. We mask Operational Land Imager (OLI; Landsat 8) image stacks from five study areas with corresponding ETM+ missing data layers, using these modified OLI stacks as inputs. We explored the algorithm's parameter space, particularly window size in the spatial and temporal dimensions. Window regression yielded the best accuracy (and moderately long computation time) with a large spatial radius (a 7 × 7 pixel window) and a moderate temporal radius (here, five layers). In this case, root mean square error for deviations from the observed reflectance ranged from 3.7-7.6% over all study areas, depending on the band. Second-order response surface analysis suggested that a 15 × 15 pixel window, in conjunction with a 9-layer temporal window, may produce the best accuracy. Compared to the neighborhood similar pixel interpolator gap-filling algorithm, window regression yielded slightly better accuracy on average. Because it relies on no ancillary data, window regression may be used to conveniently preprocess stacks for other data-intensive algorithms. [13], which classify each image and impute based on the relationships between pixels sharing a class; and the Profile-Based Interpolator [14], which leverages multitemporal trajectories and geospatial similarities to impute. The third category of gap-filling algorithms, wherein the source images include non-ETM+ data, includes the algorithms from the second category when they employ additional data from other Landsat platforms. Fusion with other sensor platforms such as the Moderate Resolution Imaging Spectroradiometer (MODIS) [15] falls into this category as well. Additionally, Wijedasa [16] used post-classification mosaics to generate a seamless map of peat swamp disturbance based on ETM+ data, though this process required supervised classification. A complete review of these algorithms is beyond the scope of this study, but each of them has advantages and disadvantages, the latter typically centering around the tradeoff of slow computation times against per-pixel interpolation accuracy.However, since the Landsat archive was made freely available to the public [17], it has become common to incorporate the full history, from the Thematic Mapper era to the present day, in time series research [18][19][20]. Many of these studies rely on algorithms that leverage the temporal richness of the Landsat archive to predict image data in other dates (i.e., between-images interpolation), but they do not necessarily perform well at imputing missing...

A Google Earth Engine Tool to Investigate, Map and Monitor Volcanic Thermal Anomalies at Global Scale by Means of Mid-High Spatial Resolution Satellite Data

“…Figure 13b displays the lava flow emitted by the Eyjafjallajökull volcano on 3 April 2010 and generated by an intense effusive eruption which started the previous month after about two centuries of quiescence, previously investigated from space by means of infrared MODIS data [75]. It should be noted that, despite the striping in Landsat-7 ETM+ imagery (e.g., [76]), the analyzed time series (from 1999 to 2011) did not reveal any false detection (see quiescence periods correctly marked by the tool before and after the end of eruption). Thus, the NHI tool seems to perform well in using also TM/ETM+ data, although the impact of other issues (e.g., oversaturation effects [77]) needs to be better assessed.…”

Landsat 4, 5 and 7 (1982 to 2017) Analysis Ready Data (ARD) Observation Coverage over the Conterminous United States and Implications for Terrestrial Monitoring

“…Research to assess and consider more sophisticated processing using the ARD is ongoing, including correction for bi-directional reflectance and topographic effects, gap-filling and improved cloud masking [23,28,41,42]. A recommendation of this study is to improve the Landsat geolocation processing.…”

Science of Landsat Analysis Ready Data

“…Brooks et al [16] applied a window regression method that was originally developed to impute low-quality moderate resolution imaging spectroradiometer (MODIS) to Landsat ARD between 2014 and 2016 to gap-fill Landsat ETM+ SLC-off data. For the five study areas, root mean square error derived from the observed reflectance ranged from 3.7%-7.6% for different spectral bands, and it outperformed other algorithms such as the neighborhood similar pixel interpolator gap-filling algorithm.…”