Search for: All records

Water table depths (WTD) measured by remote sensing (see methods below), calibrated with field logger-based measurements (from https://doi.org/10.5281/zenodo.15046718). This image (wtd_byday_space_time_predictions.tif) is a GeoTIFF with embedded georeferencing information. It includes 119 bands (defined in bands_wtd_byday_space_time_predictions.csv), which correspond to each date from June 10 to Oct. 6, 2014.   METHODS: Water table gauge (point) data QA/QC: In situ measured water table depth (WTD) data (available at https://doi.org/10.5281/zenodo.15046718) includes hourly measurements from 4pm June 9, 2014, until 1pm October 8, 2014, at 36 probe sites.  All WTD with an indication of poor data quality were removed as reviewed by team members. The remaining data was further reduced based on available observations: Sites with more than 98% of all time observations missing where removed, as well as sites missing spatial coordinates. Thus 26 probe sites remained for further modeling. Remote sensing input and processing: Six quad-pol Radarsat-2 images (https://earth.esa.int/eogateway/catalog/radarsat-2-esa-archive) acquired from July 2014 to August 2014 over this study site with their incidence angle ranging from 20 to 45 degrees are used for this project. The images are listed in Table 1. The Sigma Lee filter with the window size 7 by 7 is applied to and geocoded by the range-doppler terrain correction method using a 30m by 30m SRTM DEM (https://doi.org/10.5066/F7PR7TFT) with the final resolution 10m by 10m. To minimize the effects caused by the terrain slope, all images are de-orientated. Finally, a three-component model-based decomposition is applied to decompose the scattering into surface, double-bounce and volume scattering, in which the surface scattering is caused by the underlying ground and vegetation canopy, the double-bounce is from the interaction between the water and vegetation; and the volume scattering is from the multiple scattering in the vegetation canopy. Table 1. Radarsat-2 images Date Mode Incidence Angle Pass Direction 20140702 FQ25 43 Ascending 20140710 FQ05 24 Ascending 20140717 FQ03 21 Ascending 20140726 FQ25 43 Ascending 20140809 FQ20 40 Ascending 20140819 FQ25 43 Ascending In addition, Table 1 depicts that there are three images in FQ25 mode having the same SAR configuration with their baseline approximately 15m, 180m and 165m respectively. This might be used for the InSAR processing, in which the generated unwrapped phase and coherence from the InSAR technique might be useful for the change detection of the water depth in future. Water table depth modeling: The goal of the WTD modeling is to obtain wall-to-wall estimates of daily water depth in the study region from June 10 to October 6, 2014. In the WTD point measurements data (https://doi.org/10.5281/zenodo.15046718), there are 9 sites with complete time series data and 17 sites with varying degrees of completeness. To do the space-time interpolation, we consider an iterative approach, where each iteration consists of two steps, a strictly temporal interpolation and a strictly spatial interpolation. We carry out 3 iterations. In the temporal interpolation step, each of the 26 hourly time series are subset to include the 4pm hour each day, then available data are used to estimate coefficients of 30 linear B-spline basis function using least squares regression, and finally missing time points are predicted based on the linear B-spline models.  In the spatial interpolation step, thin plate spline (TPS) regression is used to estimate WTD over the entire study area for each of the 119 days at 4pm using the non-missing site data and the temporally estimated data at missing sites. The form of the TPS is WTD = b0 + b1*DEM(s) + b2*Pvt1(s) + … + b7*Pvt6(s) + f(s) + ε where b0 + b1*DEM(s) + b2*Pvt1 + … + b7*Pvt6 is a fixed effect regression and f(s) + ε is the estimate of a latent 2-D spatial process.  In the regression part, DEM is a digital elevation map derived from LiDAR where the closest 1m square pixel is matched with each probe location, and Pvt1(s), …, Pvt6(s) are the volume scattering parameters at 6 different overpass dates derived from Radarsat-2, where the closest 10m pixel is matched to each probe location. The reason for using the 4pm hour observation for each day is because the Radarsat-2 overpasses are also obtained at approximately 4pm. At the end of the spatial interpolation step of the first two iterations, only the spatially interpolated WTD values at the missing probe locations having some observed data on other days are retained. These data along with the non-missing observations in the partially observed time series are updated in the following temporal interpolation when the coefficients of the B-splines are re-estimated. An overall goodness of fit measure (R-square) was calculated after the spatial interpolation of each iteration. In order of iterations these were 0.71, 0.84, 0.85. Since there was little improvement between the second and third iteration, the third was taken as the final iteration. This measure of fit does not account for prediction accuracy and does not show the spatially and temporally varying uncertainty. These should be considered in future work. WTD maps of each of the 119 days at 4pm are the result of the third spatial interpolation to every 1m square pixel in the study area.   FUNDING: National Aeronautics and Space Administration, Interdisciplinary Science program: From Archaea to the Atmosphere (award # NNX17AK10G) National Science Foundation, Biology Integration Institutes Program: EMERGE Biology Integration Institute (award # 2022070) United States Department of Energy Office of Biological and Environmental Research, Genomic Science Program: The IsoGenie Project (grant #s DE-SC0004632, DE-SC0010580, and DE-SC0016440) National Science Foundation, MacroSystems program (grant # EF-1241037) We thank the Swedish Polar Research Secretariat and SITES for the support of the work done at the Abisko Scientific Research Station. SITES is supported by the Swedish Research Council's grant 4.3-2021-00164. 

Methane (CH4) emissions in Stordalen Mire (northern Sweden), estimated via two different approaches: "Paint by number" (field ch4_modified_prj.tif): CH4 emission across the landscape calculated via “paint-by-number” approach, using 2014 autochamber-based flux measurements (https://doi.org/10.5281/zenodo.14052690) mapped to landcover classifications (https://doi.org/10.5281/zenodo.15042233). DNDC-modeled (Modeled CH4.tif): CH4 emission across the landscape modeled via Wetland-DNDC (https://www.dndc.sr.unh.edu/), driven by remotely sensed landcover classifications (https://doi.org/10.5281/zenodo.15042233), water table depth (https://doi.org/10.5281/zenodo.15092752), climate data (provided by the Abisko Scientific Research Station), and soil parameters (defined as in Deng et al. 2014, 2017). The DNDC model was run on vegetation and water table clusters (determined by k-means clustering), and model output was spatially assigned to each map pixel. Modeled CH4 emissions account for CH4 production from DOC (Randomforest_stack_epsg32634_extent_kmeansclass10_CH4 prod from DOC.tif) and from CO2 (Randomforest_stack_epsg32634_extent_kmeansclass10_CH4 prod from CO2.tif), minus oxidation (Randomforest_stack_epsg32634_extent_kmeansclass10_CH4 oxid.tif). The model also outputs a map of CH4 isotopic composition (δ13C-CH4) of emissions (Randomforest_stack_epsg32634_extent_kmeansclass10_Delta CH4 flux.tif). The difference between these approaches is provided as a difference map (CH4diff.tif), calculated as the "paint-by-number" (PBN) emissions (field ch4_modified_prj.tif) minus the Wetland-DNDC modeled emissions (Modeled CH4.tif). These images are GeoTIFFs with embedded georeferencing information. FUNDING: National Aeronautics and Space Administration, Interdisciplinary Science program: From Archaea to the Atmosphere (award # NNX17AK10G). National Science Foundation, Biology Integration Institutes Program: EMERGE Biology Integration Institute (award # 2022070). United States Department of Energy Office of Biological and Environmental Research, Genomic Science Program: The IsoGenie Project (grant #s DE-SC0004632, DE-SC0010580, and DE-SC0016440). National Science Foundation, MacroSystems program (grant # EF-1241037). We thank the Swedish Polar Research Secretariat and SITES for the support of the work done at the Abisko Scientific Research Station. SITES is supported by the Swedish Research Council's grant 4.3-2021-00164.