Remotely sensed terrestrial open water evaporation

Terrestrial open water evaporation is difficult to measure both in situ and remotely yet is critical for understanding changes in reservoirs, lakes, and inland seas from human management and climatically altered hydrological cycling. Multiple satellite missions and data systems (e.g., ECOSTRESS, OpenET) now operationally produce evapotranspiration (ET), but the open water evaporation data produced over millions of water bodies are algorithmically produced differently than the main ET data and are often overlooked in evaluation. Here, we evaluated the open water evaporation algorithm, AquaSEBS, used by ECOSTRESS and OpenET against 19 in situ open water evaporation sites from around the world using MODIS and Landsat data, making this one of the largest open water evaporation validations to date. Overall, our remotely sensed open water evaporation retrieval captured some variability and magnitude in the in situ data when controlling for high wind events (instantaneous: r2 = 0.71; bias = 13% of mean; RMSE = 38% of mean). Much of the instantaneous uncertainty was due to high wind events (u > mean daily 7.5 m·s-1) when the open water evaporation process shifts from radiatively-controlled to atmospherically-controlled; not accounting for high wind events decreases instantaneous accuracy significantly (r2 = 0.47; bias = 36% of mean; RMSE = 62% of mean). However, this sensitivity minimizes with temporal integration (e.g., daily RMSE = 1.2-1.5 mm·day-1). To benchmark AquaSEBS, we ran a suite of 11 machine learning models, but found that they did not significantly improve on the process-based formulation of AquaSEBS suggesting that the remaining error is from a combination of the in situ evaporation measurements, forcing data, and/or scaling mismatch; the machine learning models were able to predict error well in and of itself (r2 = 0.74). Our results provide confidence in the remotely sensed open water evaporation data, though not without uncertainty, and a foundation by which current and future missions may build such operational data.

Comparison of Machine Learning Methods to Up-Scale Gross Primary Production

Eddy covariance observation is an applicable way to obtain accurate and continuous carbon flux at flux tower sites, while remote sensing technology could estimate carbon exchange and carbon storage at regional and global scales effectively. However, it is still challenging to up-scale the field-observed carbon flux to a regional scale, due to the heterogeneity and the unstable air conditions at the land surface. In this paper, gross primary production (GPP) from ground eddy covariance systems were up-scaled to a regional scale by using five machine learning methods (Cubist regression tree, random forest, support vector machine, artificial neural network, and deep belief network). Then, the up-scaled GPP were validated using GPP at flux tower sites, weighted GPP in the footprint, and MODIS GPP products. At last, the sensitivity of the input data (normalized difference vegetation index, fractional vegetation cover, shortwave radiation, relative humidity and air temperature) to the precision of up-scaled GPP was analyzed, and the uncertainty of the machine learning methods was discussed. The results of this paper indicated that machine learning methods had a great potential in up-scaling GPP at flux tower sites. The validation of up-scaled GPP, using five machine learning methods, demonstrated that up-scaled GPP using random forest obtained the highest accuracy.

Upscaling Northern Peatland CO2 Fluxes Using Satellite Remote Sensing Data

Peatlands play an important role in the global carbon cycle as they contain a large soil carbon stock. However, current climate change could potentially shift peatlands from being carbon sinks to carbon sources. Remote sensing methods provide an opportunity to monitor carbon dioxide (CO2) exchange in peatland ecosystems at large scales under these changing conditions. In this study, we developed empirical models of the CO2 balance (net ecosystem exchange, NEE), gross primary production (GPP), and ecosystem respiration (ER) that could be used for upscaling CO2 fluxes with remotely sensed data. Two to three years of eddy covariance (EC) data from five peatlands in Sweden and Finland were compared to modelled NEE, GPP and ER based on vegetation indices from 10 m resolution Sentinel-2 MSI and land surface temperature from 1 km resolution MODIS data. To ensure a precise match between the EC data and the Sentinel-2 observations, a footprint model was applied to derive footprint-weighted daily means of the vegetation indices. Average model parameters for all sites were acquired with a leave-one-out-cross-validation procedure. Both the GPP and the ER models gave high agreement with the EC-derived fluxes (R2 = 0.70 and 0.56, NRMSE = 14% and 15%, respectively). The performance of the NEE model was weaker (average R2 = 0.36 and NRMSE = 13%). Our findings demonstrate that using optical and thermal satellite sensor data is a feasible method for upscaling the GPP and ER of northern boreal peatlands, although further studies are needed to investigate the sources of the unexplained spatial and temporal variation of the CO2 fluxes.

Integrating eddy fluxes and remote sensing products in a rotational grazing native tallgrass prairie pasture

Eddy covariance (EC) systems provide integrated fluxes within their footprint areas. Spatial heterogeneity of up-scaled areas and spatio-temporal mismatches between EC footprint and remote sensing pixels jeopardize the performance of most satellite-based models. To examine the impact of spatial resolution of satellite products on up-scaling of fluxes, we compared the relationships between measured eddy fluxes and enhanced vegetation index (EVI) derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) at 500 and 250 m spatial resolutions, Visible Infrared Imaging Radiometer Suite (VIIRS) at 500 m spatial resolution, and Landsat at 30 m spatial resolution but integrated at the paddock-scale. The experiment was conducted over a grazed native tallgrass prairie pasture, which was divided into nine paddocks for rotational grazing. The EVI data from all satellites showed consistency in detecting vegetation phenology. Seasonality of EC-measured fluxes corresponded well with remotely-sensed vegetation phenology. Approximately 80% of contribution to eddy fluxes came from within 80 m upwind distance of the 2.7 m tall EC tower. As a result, the major contributing area for the measured fluxes was mostly limited to the paddock containing the EC tower. Different timings and duration of grazing caused some heterogeneity among paddocks within the pasture. The EVI of different spatial scales showed strong relationships with CO₂ fluxes. However, Landsat-derived EVI integrated for the paddock containing the EC tower showed substantially stronger relationships with CO₂ fluxes than did MODIS and VIIRS-derived EVI integrated for multiple paddocks, most likely due to similar spatial resolutions of remote sensing and EC observations. Results illustrate that satellite products of fine-scale spatial resolution that are comparable to EC footprints can help improve the performance of satellite-based models for modeling or up-scaling of eddy fluxes, especially in heterogeneous ecosystems.

Current and emerging methodologies for estimating carbon sequestration in agricultural soils: A review

This review covers the current and emerging analytical methods used in laboratory, field, landscape and regional contexts for measuring soil organic carbon (SOC) sequestration in agricultural soil. Soil depth plays an important role in estimating SOC sequestration. Selecting appropriate sampling design, depth of soil, use of proper analytical methods and base line selection are prerequisites for estimating accurately the soil carbon stocks. Traditional methods of wet digestion and dry combustion (DC) are extensively used for routine laboratory analysis; the latter is considered to be the "gold standard" and superior to the former for routine laboratory analysis. Recent spectroscopic techniques can measure SOC stocks in laboratory and in-situ even up to a deeper depth. Aerial spectroscopy using multispectral and/or hyperspectral sensors located on aircraft, unmanned aerial vehicles (UAVs) or satellite platforms can measure surface soil organic carbon. Although these techniques' current precision is low, the next generation hyperspectral sensor with improved signal noise ratio will further improve the accuracy of prediction. At the ecosystem level, carbon balance can be estimated directly using the eddy-covariance approach and indirectly by employing agricultural life cycle analysis (LCA). These methods have tremendous potential for estimating SOC. Irrespective of old or new approaches, depending on the resources and research needed, they occupy a unique place in soil carbon and climate research. This paper highlights the overview, potential limitations of various scale-dependent techniques for measuring SOC sequestration in agricultural soil.

Monitoring ecosystem reclamation recovery using optical remote sensing: Comparison with field measurements and eddy covariance

Time series remote sensing vegetation indices derived from SPOT 5 data are compared with vegetation structure and eddy covariance flux data at 15 dry to wet reclamation and reference sites within the Oil Sands region of Alberta, Canada. This comprehensive analysis examines the linkages between indicators of ecosystem function and change trajectories observed both at the plot level and within pixels. Using SPOT imagery, we find that higher spatial resolution datasets (e.g. 10 m) improves the relationship between vegetation indices and structural measurements compared with interpolated (lower resolution) pixels. The simple ratio (SR) vegetation index performs best when compared with stem density-based indicators (R² = 0.65; p < 0.00), while the normalised difference vegetation index (NDVI) and soil adjusted vegetation index (SAVI) are most comparable to foliage indicators (leaf area index (LAI) and canopy cover (R² = 0.52-0.78; p > 0.02). Fluxes (net ecosystem production (NEP) and gross ecosystem production (GEP)) are most related to NDVI and SAVI when these are interpolated to larger 20 m × 20 m pixels (R² = 0.44-0.50; p < 0.00). As expected, decreased sensitivity of NDVI is problematic for sites with LAI > 3 m² m^-2, making this index more appropriate for newly regenerating reclamation areas. For sites with LAI < 3 m² m^-2, trajectories of vegetation change can be mapped over time and are within 2.7% and 3.3% of annual measured LAI changes observed at most sites. This study demonstrates the utility of remote sensing in combination with field and eddy covariance data for monitoring and scaling of reclaimed and reference site productivity within and beyond the Oil Sands Region of western Canada.

Regional atmospheric cooling and wetting effect of permafrost thaw-induced boreal forest loss

In the sporadic permafrost zone of North America, thaw-induced boreal forest loss is leading to permafrost-free wetland expansion. These land cover changes alter landscape-scale surface properties with potentially large, however, still unknown impacts on regional climates. In this study, we combine nested eddy covariance flux tower measurements with satellite remote sensing to characterize the impacts of boreal forest loss on albedo, eco-physiological and aerodynamic surface properties, and turbulent energy fluxes of a lowland boreal forest region in the Northwest Territories, Canada. Planetary boundary layer modelling is used to estimate the potential forest loss impact on regional air temperature and atmospheric moisture. We show that thaw-induced conversion of forests to wetlands increases albedo: and bulk surface conductance for water vapour and decreases aerodynamic surface temperature. At the same time, heat transfer efficiency is reduced. These shifts in land surface properties increase latent at the expense of sensible heat fluxes, thus, drastically reducing Bowen ratios. Due to the lower albedo of forests and their masking effect of highly reflective snow, available energy is lower in wetlands, especially in late winter. Modelling results demonstrate that a conversion of a present-day boreal forest-wetland to a hypothetical homogeneous wetland landscape could induce a near-surface cooling effect on regional air temperatures of up to 3-4 °C in late winter and 1-2 °C in summer. An atmospheric wetting effect in summer is indicated by a maximum increase in water vapour mixing ratios of 2 mmol mol^-1 . At the same time, maximum boundary layer heights are reduced by about a third of the original height. In fall, simulated air temperature and atmospheric moisture between the two scenarios do not differ. Therefore, permafrost thaw-induced boreal forest loss may modify regional precipitation patterns and slow down regional warming trends.