Assessment of Workflow Feature Selection on Forest LAI Prediction with Sentinel-2A MSI, Landsat 7 ETM+ and Landsat 8 OLI.

The European Space Agency (ESA)'s Sentinel-2A (S2A) mission is providing time series that allow the characterisation of dynamic vegetation, especially when combined with the National Aeronautics and Space Administration (NASA)/United States Geological Survey (USGS) Landsat 7 (L7) and Landsat 8 (L8) missions. Hybrid retrieval workflows combining non-parametric Machine Learning Regression Algorithms (MLRAs) and vegetation Radiative Transfer Models (RTMs) were proposed as fast and accurate methods to infer biophysical parameters such as Leaf Area Index (LAI) from these data streams. However, the exact design of optimal retrieval workflows is rarely discussed. In this study, the impact of five retrieval workflow features on LAI prediction performance of MultiSpectral Instrument (MSI), Enhanced Thematic Mapper Plus (ETM+) and Operational Land Imager (OLI) observations was analysed over a Dutch beech forest site for a one-year period. The retrieval workflow features were the (1) addition of prior knowledge of leaf chemistry (two alternatives), (2) the choice of RTM (two alternatives), (3) the addition of Gaussian noise to RTM produced training data (four and five alternatives), (4) possibility of using Sun Zenith Angle (SZA) as an additional MLRA training feature (two alternatives), and (5) the choice of MLRA (six alternatives). The features were varied in a full grid resulting in 960 inversion models in order to find the overall impact on performance as well as possible interactions among the features. A combination of a Terrestrial Laser Scanning (TLS) time series with litter-trap derived LAI served as independent validation. The addition of absolute noise had the most significant impact on prediction performance. It improved the median prediction Root Mean Square Error (RMSE) by 1.08 m2 m-2 when 5 % noise was added compared to inversions with 0 % absolute noise. The choice of the MLRA was second most important in terms of median prediction performance, which differed by 0.52 m2 m-2 between the best and worst model. The best inversion model achieved an RMSE of 0.91 m2 m-2 and explained 84.9% of the variance of the reference time series. The results underline the need to explicitly describe the used noise model in future studies. Similar studies should be conducted in other study areas, both forest and crop systems, in order to test the noise model as an integral part of hybrid retrieval workflows.

Memory effects of climate and vegetation affecting net ecosystem CO2 fluxes in global forests.

Forests play a crucial role in the global carbon (C) cycle by storing and sequestering a substantial amount of C in the terrestrial biosphere. Due to temporal dynamics in climate and vegetation activity, there are significant regional variations in carbon dioxide (CO2) fluxes between the biosphere and atmosphere in forests that are affecting the global C cycle. Current forest CO2 flux dynamics are controlled by instantaneous climate, soil, and vegetation conditions, which carry legacy effects from disturbances and extreme climate events. Our level of understanding from the legacies of these processes on net CO2 fluxes is still limited due to their complexities and their long-term effects. Here, we combined remote sensing, climate, and eddy-covariance flux data to study net ecosystem CO2 exchange (NEE) at 185 forest sites globally. Instead of commonly used non-dynamic statistical methods, we employed a type of recurrent neural network (RNN), called Long Short-Term Memory network (LSTM) that captures information from the vegetation and climate's temporal dynamics. The resulting data-driven model integrates interannual and seasonal variations of climate and vegetation by using Landsat and climate data at each site. The presented LSTM algorithm was able to effectively describe the overall seasonal variability (Nash-Sutcliffe efficiency, NSE = 0.66) and across-site (NSE = 0.42) variations in NEE, while it had less success in predicting specific seasonal and interannual anomalies (NSE = 0.07). This analysis demonstrated that an LSTM approach with embedded climate and vegetation memory effects outperformed a non-dynamic statistical model (i.e. Random Forest) for estimating NEE. Additionally, it is shown that the vegetation mean seasonal cycle embeds most of the information content to realistically explain the spatial and seasonal variations in NEE. These findings show the relevance of capturing memory effects from both climate and vegetation in quantifying spatio-temporal variations in forest NEE.

Assessing the influence of historic net and gross land changes on the carbon fluxes of Europe.

Legacy effects of land cover/use on carbon fluxes require considering both present and past land cover/use change dynamics. To assess past land use dynamics, model-based reconstructions of historic land cover/use are needed. Most historic reconstructions consider only the net area difference between two time steps (net changes) instead of accounting for all area gains and losses (gross changes). Studies about the impact of gross and net land change accounting methods on the carbon balance are still lacking. In this study, we assessed historic changes in carbon in soils for five land cover/use types and of carbon in above-ground biomass of forests. The assessment focused on Europe for the period 1950 to 2010 with decadal time steps at 1-km spatial resolution using a bookkeeping approach. To assess the implications of gross land change data, we also used net land changes for comparison. Main contributors to carbon sequestration between 1950 and 2010 were afforestation and cropland abandonment leading to 14.6 PgC sequestered carbon (of which 7.6 PgC was in forest biomass). Sequestration was highest for old-growth forest areas. A sequestration dip was reached during the 1970s due to changes in forest management practices. Main contributors to carbon emissions were deforestation (1.7 PgC) and stable cropland areas on peaty soils (0.8 PgC). In total, net fluxes summed up to 203 TgC yr(-1) (98 TgC yr(-1) in forest biomass and 105 TgC yr(-1) in soils). For areas that were subject to land changes in both reconstructions (35% of total area), the differences in carbon fluxes were about 68%. Overall for Europe the difference between accounting for either gross or net land changes led to 7% difference (up to 11% per decade) in carbon fluxes with systematically higher fluxes for gross land change data.

Gross changes in reconstructions of historic land cover/use for Europe between 1900 and 2010.

Historic land-cover/use change is important for studies on climate change, soil carbon, and biodiversity assessments. Available reconstructions focus on the net area difference between two time steps (net changes) instead of accounting for all area gains and losses (gross changes). This leads to a serious underestimation of land-cover/use dynamics with impacts on the biogeochemical and environmental assessments based on these reconstructions. In this study, we quantified to what extent land-cover/use reconstructions underestimate land-cover/use changes in Europe for the 1900-2010 period by accounting for net changes only. We empirically analyzed available historic land-change data, quantified their uncertainty, corrected for spatial-temporal effects and identified underlying processes causing differences between gross and net changes. Gross changes varied for different land classes (largest for forest and grassland) and led to two to four times the amount of net changes. We applied the empirical results of gross change quantities in a spatially explicit reconstruction of historic land change to reconstruct gross changes for the EU27 plus Switzerland at 1 km spatial resolution between 1950 and 2010. In addition, the reconstruction was extended back to 1900 to explore the effects of accounting for gross changes on longer time scales. We created a land-change reconstruction that only accounted for net changes for comparison. Our two model outputs were compared with five commonly used global reconstructions for the same period and area. In our reconstruction, gross changes led in total to a 56% area change (ca. 0.5% yr(-1)) between 1900 and 2010 and cover twice the area of net changes. All global reconstructions used for comparison estimated fewer changes than our gross change reconstruction. Main land-change processes were cropland/grassland dynamics and afforestation, and also deforestation and urbanization.

Detecting leaf pulvinar movements on NDVI time series of desert trees: a new approach for water stress detection.

Heliotropic leaf movement or leaf 'solar tracking' occurs for a wide variety of plants, including many desert species and some crops. This has an important effect on the canopy spectral reflectance as measured from satellites. For this reason, monitoring systems based on spectral vegetation indices, such as the normalized difference vegetation index (NDVI), should account for heliotropic movements when evaluating the health condition of such species. In the hyper-arid Atacama Desert, Northern Chile, we studied seasonal and diurnal variations of MODIS and Landsat NDVI time series of plantation stands of the endemic species Prosopis tamarugo Phil., subject to different levels of groundwater depletion. As solar irradiation increased during the day and also during the summer, the paraheliotropic leaves of Tamarugo moved to an erectophile position (parallel to the sun rays) making the NDVI signal to drop. This way, Tamarugo stands with no water stress showed a positive NDVI difference between morning and midday (ΔNDVI mo-mi) and between winter and summer (ΔNDVI W-S). In this paper, we showed that the ΔNDVI mo-mi of Tamarugo stands can be detected using MODIS Terra and Aqua images, and the ΔNDVI W-S using Landsat or MODIS Terra images. Because pulvinar movement is triggered by changes in cell turgor, the effects of water stress caused by groundwater depletion can be assessed and monitored using ΔNDVI mo-mi and ΔNDVI W-S. For an 11-year time series without rainfall events, Landsat ΔNDVI W-S of Tamarugo stands showed a positive linear relationship with cumulative groundwater depletion. We conclude that both ΔNDVI mo-mi and ΔNDVI W-S have potential to detect early water stress of paraheliotropic vegetation.

A laboratory goniometer system for measuring reflectance and emittance anisotropy.

In this paper, a laboratory goniometer system for performing multi-angular measurements under controlled illumination conditions is described. A commercially available robotic arm enables the acquisition of a large number of measurements over the full hemisphere within a short time span making it much faster than other goniometers. In addition, the presented set-up enables assessment of anisotropic reflectance and emittance behaviour of soils, leaves and small canopies. Mounting a spectrometer enables acquisition of either hemispherical measurements or measurements in the horizontal plane. Mounting a thermal camera allows directional observations of the thermal emittance. This paper also presents three showcases of these different measurement set-ups in order to illustrate its possibilities. Finally, suggestions for applying this instrument and for future research directions are given, including linking the measured reflectance anisotropy with physically-based anisotropy models on the one hand and combining them with field goniometry measurements for joint analysis with remote sensing data on the other hand. The speed and flexibility of the system offer a large added value to the existing pool of laboratory goniometers.