Satellite Time Series and Google Earth Engine Democratize the Process of Forest-Recovery Monitoring over Large Areas

Contemporary forest-health initiatives require technologies and workflows that can monitor forest degradation and recovery simply and efficiently over large areas. Spectral recovery analysis—the examination of spectral trajectories in satellite time series—can help democratize this process, particularly when performed with cloud computing and open-access satellite archives. We used the Landsat archive and Google Earth Engine (GEE) to track spectral recovery across more than 57,000 forest harvest areas in the Canadian province of Alberta. We analyzed changes in the normalized burn ratio (NBR) to document a variety of recovery metrics, including year of harvest, percent recovery after five years, number of years required to achieve 80% of pre-disturbance NBR, and % recovery the end of our monitoring window (2018). We found harvest areas in Alberta to recover an average of 59.9% of their pre-harvest NBR after five years. The mean number of years required to achieve 80% recovery in the province was 8.7 years. We observed significant variability in pre- and post-harvest spectral recovery both regionally and locally, demonstrating the importance of climate, elevation, and complex local factors on rates of spectral recovery. These findings are comparable to those reported in other studies and demonstrate the potential for our workflow to support broad-scale management and research objectives in a manner that is complimentary to existing information sources. Measures of spectral recovery for all 57,979 harvest areas in our analysis are freely available and browseable via a custom GEE visualization tool, further demonstrating the accessibility of this information to stakeholders and interested members of the public.

Physically Plausible Spectral Reconstruction

Spectral reconstruction algorithms recover spectra from RGB sensor responses. Recent methods—with the very best algorithms using deep learning—can already solve this problem with good spectral accuracy. However, the recovered spectra are physically incorrect in that they do not induce the RGBs from which they are recovered. Moreover, if the exposure of the RGB image changes then the recovery performance often degrades significantly—i.e., most contemporary methods only work for a fixed exposure. In this paper, we develop a physically accurate recovery method: the spectra we recover provably induce the same RGBs. Key to our approach is the idea that the set of spectra that integrate to the same RGB can be expressed as the sum of a unique fundamental metamer (spanned by the camera’s spectral sensitivities and linearly related to the RGB) and a linear combination of a vector space of metameric blacks (orthogonal to the spectral sensitivities). Physically plausible spectral recovery resorts to finding a spectrum that adheres to the fundamental metamer plus metameric black decomposition. To further ensure spectral recovery that is robust to changes in exposure, we incorporate exposure changes in the training stage of the developed method. In experiments we evaluate how well the methods recover spectra and predict the actual RGBs and RGBs under different viewing conditions (changing illuminations and/or cameras). The results show that our method generally improves the state-of-the-art spectral recovery (with more stabilized performance when exposure varies) and provides zero colorimetric error. Moreover, our method significantly improves the color fidelity under different viewing conditions, with up to a 60% reduction in some cases.

Deep Optimal Filter Responses for Multi-Spectral Imaging

Spectral recovery from measured camera signals based on deep learning lead to significant advancements of the potential reconstruction quality. However, most deep learning based approaches only consider RGB cameras and are targeting object classification in particular or remote sensing in general as their final application. Within this work, we analyze the influence of a joint filter optimization and spectral recovery for multi-spectral image acquisition with the underlying goal of capturing high-fidelity color images. An evaluation on the influence of the total camera channel count on the reproduction quality is provided. Finally, a possible normalization of spectral data is discussed.

Colour Fidelity in Spectral Reconstruction from RGB Images

Spectral reconstruction (SR) aims to recover high resolution spectra from RGB images. Recent developments - leading by Convolutional Neural Networks (CNN) - can already solve this problem with low errors. However, those leading methods do not explicitly ensure the predicted spectra will re-integrate (with the underlying camera response functions) into the same RGB colours as the ones they are recovered from, namely the 'colour fidelity' problem. The purpose of this paper is to show, visually and quantitatively, how well (or bad) the existing SR models maintain colour fidelity. Three main approaches are evaluated - regression, sparse coding and CNN. Furthermore, aiming for a more realistic setting, the evaluations are done on real RGB images and the 'end-of-pipe' images (i.e.rendered images shown to the end users) are provided for visual comparisons. It is shown that the state-of-the-art CNN-based model, despite of the superior performance in spectral recovery, introduces significant colour shifts in the final images. Interestingly, the leading sparse coding and the simple linear regression model, both of which are based on linear mapping, best preserve the colour fidelity in SR.

Multi-temporal analysis of post-fire vegetation spectral recovery over European forests

<p>The disturbance and recovery of European Forest ecosystems are greatly affected by wildfires, requiring continued monitoring to observe vegetational structure altered over time. One of the most important parameters is “fire severity” defined as magnitude of environmental change caused by wildfires. Due to correlation between severity and post-fire recovery vegetation, fire severity is an  important indicator to define operations in the burned areas. Satellite based-data is becoming a key information for near real-time mapping and monitoring burned area after wildfire disturbances. Moderate resolution Imaging Spectroradiometer (MODIS) time-series data allows for both the capture of the initial disturbance and the ability to monitor the subsequent vegetation regeneration with spectral vegetation indices. In this study, the Google Earth Engine (GEE) platform, was used to analyse post-fire spectral recovery of European forests through the Normalized Difference Vegetation Index (NDVI) and the Relative Recovery Indicator (RRI) based on the Normalized Burn Ratio (NBR). We assessed Normalized Burn Ratio time series in order to determine trends in the short term rates of spectral recovery for three forest land cover classes and European Biogeographic regions disturbed by wildfire (2004-2013), using a series of 5-year post-disturbance time window. NBR pattern of mixed forests showed a lower variability than broadleaved and coniferous forest, indicating high resilience to environmental disturbances. Results indicate different trends of forest recovery according to different spectral indices analysed for European forest ecosystems. During the analysis period (2004-2013) we found that post-fire spectral recovery rates decreased over ten years of observation in each land cover classes and Biogeographic regions. These trends could be related to on-going climate changes affecting the Mediterranean region.</p><p>Keywords: Fire severity, Forest, Google Earth Engine, Modis (time series), Recovery, Spectral index, Wildfire.</p><p> </p>