Spatially predicted sedimentation rates, organic carbon densities and organic carbon accumulation rates in the North Sea and Skagerrak
Fuente:
PANGAEA
Guardado en:
| Autor principal: | |
|---|---|
| Formato: | Dataset Open Access |
| Lenguaje: | en |
| Publicado: |
PANGAEA
2021
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1867171025316216832 |
|---|---|
| author | Diesing, Markus |
| author_facet | Diesing, Markus |
| collection | Datos científicos de ciencias marinas y ambientales |
| contents | The dataset contains spatially predicted linear sedimentation rates, organic carbon densities, organic carbon accumulation rates and the associated uncertainties in the predictions of surficial seafloor sediments in the North Sea and Skagerrak. The results are presented as geo-referenced floating-point TIFF-files with a spatial resolution of 500 m and Lambert Azimuthal Equal-Area projection as spatial reference. The same modelling framework was used for predicting sedimentation rates (cm yr-1) based on 210Pb measurements and organic carbon densities (kg m-3). It is based on the quantile regression forest algorithm (Meinshausen, 2006) to make spatial predictions of the target variables and to estimate the uncertainty in the predictions in a spatially explicit way. The quantile regression forest is a generalisation of the random forest algorithm (Breiman, 2001), which aggregates the conditional mean from each tree in a regression forest to make an ensemble prediction. Quantile regression forest additionally returns the whole conditional distribution of the response variable. This allows to determine the underlying variability of an estimate by means of prediction intervals or the standard deviation. The modelling framework addresses uncertainty in the model by calculating the standard deviation of the quantile regression forest predictions. Furthermore, the sensitivity of the model to variations in the available data was estimated by means of resampling. To that end, the response data were split 25 times into training and test subsets at a ratio of 7:3 and 25 models were subsequently built based on these splits. The sensitivity is derived by calculating the standard deviation of the 25 predictions for every pixel. The total uncertainty is the sum of the model uncertainty and the sensitivity. Organic carbon accumulation rates (g m-2 yr-1) were calculated by multiplying predicted organic carbon densities with predicted sedimentation rates. Uncertainties were propagated by taking the square root of the sum of squared relative uncertainties. |
| format | Dataset Open Access |
| id | pangaea_https___doi_org_10_1594_PANGAEA_928272 |
| institution | PANGAEA |
| language | en |
| publishDate | 2021 |
| publisher | PANGAEA |
| record_format | pangaea |
| spellingShingle | Spatially predicted sedimentation rates, organic carbon densities and organic carbon accumulation rates in the North Sea and Skagerrak Diesing, Markus accuracy; map; North_Sea; North Sea; Organic carbon; seafloor; Sediment; sedimentation rate; spatial prediction; Uncertainty The dataset contains spatially predicted linear sedimentation rates, organic carbon densities, organic carbon accumulation rates and the associated uncertainties in the predictions of surficial seafloor sediments in the North Sea and Skagerrak. The results are presented as geo-referenced floating-point TIFF-files with a spatial resolution of 500 m and Lambert Azimuthal Equal-Area projection as spatial reference. The same modelling framework was used for predicting sedimentation rates (cm yr-1) based on 210Pb measurements and organic carbon densities (kg m-3). It is based on the quantile regression forest algorithm (Meinshausen, 2006) to make spatial predictions of the target variables and to estimate the uncertainty in the predictions in a spatially explicit way. The quantile regression forest is a generalisation of the random forest algorithm (Breiman, 2001), which aggregates the conditional mean from each tree in a regression forest to make an ensemble prediction. Quantile regression forest additionally returns the whole conditional distribution of the response variable. This allows to determine the underlying variability of an estimate by means of prediction intervals or the standard deviation. The modelling framework addresses uncertainty in the model by calculating the standard deviation of the quantile regression forest predictions. Furthermore, the sensitivity of the model to variations in the available data was estimated by means of resampling. To that end, the response data were split 25 times into training and test subsets at a ratio of 7:3 and 25 models were subsequently built based on these splits. The sensitivity is derived by calculating the standard deviation of the 25 predictions for every pixel. The total uncertainty is the sum of the model uncertainty and the sensitivity. Organic carbon accumulation rates (g m-2 yr-1) were calculated by multiplying predicted organic carbon densities with predicted sedimentation rates. Uncertainties were propagated by taking the square root of the sum of squared relative uncertainties. |
| title | Spatially predicted sedimentation rates, organic carbon densities and organic carbon accumulation rates in the North Sea and Skagerrak |
| topic | accuracy; map; North_Sea; North Sea; Organic carbon; seafloor; Sediment; sedimentation rate; spatial prediction; Uncertainty |
| url | https://doi.org/10.1594/PANGAEA.928272 |