Published November 1, 2026 | Version Submission 2026 v1.3

Upscaling riverine GHG fluxes in Germany with machine learning | Submission 2026 v1.3

  • 1. ROR icon Karlsruhe Institute of Technology

Description

This River GHG emission Dataset for Germany is based on the publication of Mwanake et al. 2025, which provides GHG emissions per day and m² of river surface within a catchment. To use these upscaled data, they must be scaled to the specific waterbody. The dataset consists of catchment-based GHG emissions, combined with geodata provided by the Thünen Institute, which calculated a first estimate of GHG emissions from aquatic natural and anthropogenic sources within the framework of the ITMS Germany project (Submission 2025).

References:

Mwanake, R.M., Wangari, E.G., Winkler, K., Gettel, G.M., Butterbach-Bahl, K., Kiese, R. (2025) From data to insights: Upscaling riverine GHG fluxes in Germany with machine learning. Science of the Total Environment 958: 177984. https://doi.org/10.1016/j.scitotenv.2024.177984

NFR Sectors (according to the IPCC guidelines)

Gases

Temporal resolution

Spatial resolution

Uncertainties

wet 2 rivers including artificial and natural running waters

CO2, CH4, N2O

Yearly (currently only the year 2020 is available)

Vector line and surface layers plus aggregated UBA Grid at 1 km × 1 km resolution

Yes, from the original publication with global numbers

Update Submission 2026 v1.2

  • Updated the geographical extent to match the CoastGEm dataset outlines. More specifically, the extent of the estuaries provided from the ITMS-Q&S_II project CoastGEm.

Update Submission 2026 v1.3:

  • Generation of a 1 x 1 km² gridded layer aggregating Channels, Ditches and Rivers.

General Remarks

  • Rivers < 3 m are currently not included to allow comparison to TI estimates from Submission 2025.

  • Some country line problems occur when a river crosses the German border, for example the Rhine (border of Baden-Württemberg & France).

Differences to Submission 2025

  • Not applicable.

Planned modifications for 2027

  • Replace ditch emissions by ditch surface and IPCC emission factor.

  • Include rivers < 3 m.

  • Improve country borders splitting rivers.

  • Extend to further years.

  • Long-term goal: increase temporal resolution to months.

Uncertainties

Based on original publication:

 

CO2

CH4

N2O

Uncertainty total based on error propagation

+-34.35%

+-38.98%

+-36.05%

License

  • The line and polygon features are ITMS-only and provided to UBA for the Gridded National Dataset.

  • The gridded data to the UBA Grid (1 x 1 km²) is OpenData.

Documentation

  • Original Publication: Mwanake, R.M., Wangari, E.G., Winkler, K., Gettel, G.M., Butterbach-Bahl, K., Kiese, R. (2025) From data to insights: Upscaling riverine GHG fluxes in Germany with machine learning. Science of the Total Environment 958: 177984. https://doi.org/10.1016/j.scitotenv.2024.177984

Technical info (English)

User Guide & Quality Assurance

Uncertainty total based on error propagation

Based on original publication: https://doi.org/10.1016/j.scitotenv.2024.177984

 

CO2

CH4

N2O

Uncertainty total based on error propagation

+-34.35%

+-38.98%

+-36.05%

Table Description Aggregated to UBA Grid (1 km x 1 km)

Technical Field Name

Alias (Display Name)

Unit

TOTAL_em1

TOTAL_EM_FCO2 [t/yr]_Riverine_GHGF_2020_All

t CO2/yr

TOTAL_em2

TOTAL_EM_FCH4 [t/yr]_Riverine_GHGF_2020_All

t CH4/yr

TOTAL_em3

TOTAL_EM_FN2O [t/yr]_Riverine_GHGF_2020_All

t N2O/yr

TOTAL_AREA

TOTAL_AREA [km2]_Riverine_GHGF_2020_All

km²

Quality Assurance

Sums over Germany

Detailed analyses in the following Excel sheet: QualityCheck_Subm2026_v3.xlsx

Preview

see: Riverine Fluxes for Germany 2020_Submission 2026v13.pdf

Totals of the line and surface datasets

Dataset

Area

CO2 emissions [t CO2/yr]

CH4 emissions [t CH4/yr]

N2O emissions [t N2O/yr]

River (line)

174,801,495.10 m²

4,068,781.874 t

940.044 t

91.312 t

River (surface)

1,321,557,107.50 m²

33,723,131.084 t

9,115.946 t

2,662.625 t

Channel (line)

415,486.13 m²

13,284.085 t

2.898 t

0.757 t

Channel (surface)

139,730,475.54 m²

3,854,199.814 t

840.664 t

335.730 t

Ditch (line)

398,291,568.17 m²

8,717,811.687 t

1,777.143 t

167.373 t

Total [km², t/yr]

2,034.80 km²

50,377,208.543 t

12,676.695 t

3,257.796 t

Totals of the aggregated UBA grid dataset (1 km × 1 km)

Waterbody type

Area [km²]

CO2 emissions [t CO2/yr]

CH4 emissions [t CH4/yr]

N2O emissions [t N2O/yr]

Rivers (line + surface)

1,496.22

37,788,069.28

10,054.96

2,753.81

Channels (line + surface)

140.15

3,867,483.90

843.56

336.49

Ditches (line)

398.29

8,717,811.69

1,777.14

167.37

Total (all waterbodies)

2,034.65

50,373,364.87

12,675.66

3,257.67

Note: The small difference between the totals of the line and surface datasets and the aggregated UBA grid dataset is expected and results from the aggregation to the 1 km × 1 km grid and rounding effects. The difference is below 0.1% for all reported totals.

Comparison to Other Literature

 

Mwanake 2025 (v0.1)

Submission 2026 v1.3

Rocher-Ros

NIR (Submission 2025)

Area

167 km²***

2,035 km²

 

~2,800 km²

CH4-C

0.8 Gg CH4-C = 800 t CH4-C

9,508 t CH4-C = 9.508 Gg CH4-C

283.543 Mg CH4-C = 0.284 Gg CH4-C

≈ 50.78 Gg CH4-C

CH4

1.07 Gg CH4* ≈ 6.41 g CH4/m²/yr

12,677 t CH4 = 12.677 Gg CH4 ≈ 6.23 g CH4/m²/yr

0.378 Gg CH4

67.71 kt CH4 = 67.71 Gg CH4** ≈ 24.18 g CH4/m²/yr

CO2-C

 

13,739,239 t CO2-C = 13,739.239 Gg CO2-C ≈ 6.75 kg CO2-C/m²/yr

 

 

CO2

 

≈ 50,377.21 Gg CO2 ≈ 24.76 kg CO2/m²/yr

 

 

N2O-N

 

2,073.1 t N2O-N = 2.073 Gg N2O-N ≈ 1.019 g N2O-N/m²/yr

 

 

N2O

 

≈ 3.26 Gg N2O ≈ 1.60 g N2O/m²/yr

 

 

*The lower national total in Mwanake 2025 is mainly explained by the coarse upscaling dataset used in the publication, which covers only part of the flowing waters present in Germany. This also explains the smaller reported water area.

**The NIR value refers to an unpublished first estimate prepared by the Thünen Institute for ITMS in Submission 2025. It was calculated based on the geodata used here and IPCC emission factors, but is not described in detail or published separately. This estimate motivated the use of the same geodata in Submission 2026, combined with the upscaling approach based on Mwanake 2025. The larger area originally included estuaries, which Submission 2026 v1.3 of the here described dataset excludes.

***The 167 km² area is an approximate, reconstructed comparison value from the Submission 2025 v1 reprocessing of the Mwanake 2025 data. Mwanake 2025 did not report one national river-surface area directly; the publication used the HydroRIVERS dataset and modelled river widths to derive catchment-level river surface areas. The HydroRIVERS-based area is of the same order of magnitude, but the reconstructed value may be slightly low because the HydroRIVERS map was clipped differently in the reprocessing. It is used here only to make the per-area CH4 values comparable.

References

  • Mwanake, R.M., Wangari, E.G., Winkler, K., Gettel, G.M., Butterbach-Bahl, K., Kiese, R. (2025) From data to insights: Upscaling riverine GHG fluxes in Germany with machine learning. Science of the Total Environment 958: 177984. https://doi.org/10.1016/j.scitotenv.2024.177984

  • Rocher-Ros, G., Stanley, E.H., Loken, L.C., Casson, N.J., Raymond, P.A., Liu, S., Amatulli, G., Sponseller, R.A. (2023) Global methane emissions from rivers and streams. Nature 621(7979): 530-535. https://doi.org/10.1038/s41586-023-06344-6

 

Methods (English)

Data Processing

The following describes how the line and area layers linking GHG emissions to the respective geodata were derived from the originally published dataset, and how they were aggregated to a 1 x 1 km² grid.

Original Publication

Publication: https://doi.org/10.1016/j.scitotenv.2024.177984

Raw Data (GeoTIFF: https://nextcloud.imk-ifu.kit.edu/s/C2cAsAaYDggtBiS) for the year 2020

The original publication provides the data on a catchment level in the following units:

  • FCO2: CO2-C mg m-2 d-1 (x/12*44 to get CO2 total molecule)

  • FCH4: CH4-C mg m-2 d-1 (x/12*16 to get CH4 total molecule)

  • N2OmgF: N2O-N mg m-2 d-1 (x/28*44 to get N2O total molecule)

  • FCO2e: CO2-eq mg m-2 d-1

  • FCH4e: CO2-eq mg m-2 d-1

  • N2OmgFe: CO2-eq mg m-2 d-1

  • Total CO2e yr: CO2-eq kg m-2 y-1

The original dataset provides GHG emissions on the catchment level. Therefore, it was necessary to scale them by rivers, channels and ditches.

Used Additional Products:

Thünen - Geodata for localisation of German waterbodies used in Submission 2025:

Data sets used by the Thünen Institute for 2025 to determine the water area of Germany.
Data is based on the German Basis DLM (BDLM) from 2023 and 2011 and is based on area objects (polygons) and line objects.

Rivers Areas (bdlm2023_fluss)

Rivers as areas, including estuaries; these were removed for the calculation, contrary to the GHG calculation of the TI.
The original shapefile contained rivers with a width of >= 12 m (bdlm2023_fluss), including estuaries. According to the BDLM, the following object type applies: Objektart=44001 UND FKT is null.
During processing, we removed polygons without GHG emissions.
To complement the data with the ITMS-Q&S_II CoastGEm dataset (estuarine emissions, Submission 2025 v1.3), we removed only polygons that overlapped with the CoastGEm dataset. Polygons that had been removed in Submission 2026 v1.1 were added back, as otherwise there would be a gap.

Class

Area

with Estuary

2,126,177,428.95 m² -> 2,126.2 km²

without Estuary

1,337,831,123.3 m² -> 1,337.8 km²

without Estuary and polygons with null values in the emission map removed

1,320,320,705.2 m² -> 1,320.3 km²

without Estuary and polygons with null values in the emission map removed, adapted to the CoastGEm dataset

1,321,329,287 m² -> 1,321.3 km²

Rivers Line (bdlm2023_fluss_linie)

Rivers from the layer named in the heading, with widths of > 3 m and < 12 m. The class 0 - 3 m was not included in the original calculation due to the high number of elements.

Class (BRG)

Count

Area*

Length*

Width

0 - 3 m (3)

3 million

0

0

1.5 m

3 - 6 m (6)

149,633

109,700,320.4 m² -> 109.7 km²

36,553,873.9 m -> 36,553.9 km

3 m

6 - 12 m (12)

37,553

65,101,556.64 m² -> 65.1 km²

10,859,626.45 m -> 10,859.6 km

6 m

*Areas and lengths were derived from the shapefile geometry in ArcGIS.

Channels Areas (bdlm2023_kanal_fläche)

This dataset contains channels with a width >= 12 m and contains the following object type: Objektart=44001 UND FKT=8300.

Class

Area

all

139,730,475.54 m² -> 139.7 km²

Channels Line (bdlm2023_kanal_linie)

Channels from the layer named in the heading, with a width < 12 m and a total area of ~41.5 ha. The area and the number of elements are rather small.

Class (BRG)

Count

Area

Length

Width

0 - 3 m (3)

39

7,720.5 m²

5,147.02 m

1.5 m

3 - 6 m (6)

75

43,409.94 m² -> 0.043 km²

14,469.98 m -> 14.5 km

3 m

6 - 12 m (12)

151

364,355.67 m² -> 0.364 km²

60,725.95 m -> 60.73 km

6 m

Ditches Line (bdlm2011_graben_linie)

This line dataset from the BDLM of 2011 contains the ditches in Germany. Definition according to OK (Objektkatalog) 3.0.2: "Natural or artificial elongated depression in the ground for irrigation and drainage as well as supply and discharge."
Since the ditches were only recorded according to object type catalog 3.0.2 (and are no longer included in 6.0.1 from 2008 onwards), the data is from 2011.
All classes were used here, but since we assumed that there are more narrow ditches than wide ones and did not want to overestimate the ditches, the widths were reduced: Class 3 (0-3 m) was calculated with an average width of 1.5 m, class 6 (3-6 m) was calculated with an average width of 3 m, and class 6-12 m was calculated with 9 m (as they were rather few compared to class 3). Additionally, the dataset includes some more classes.

Columns

Technical Field Name

Description

Unit

Breite_m

Width of the line object

m

LAENGE

Length of the line object

m

Area

Area of the line object according to Breite_m and LAENGE

m²

Area, Length & Counts

Class (BRG)

Count

Area

Length

Width (Breite_m)

0

1

744 m²

124 m

6 m

3 (0 - 3 m)

1,159,735

344,390,345 m² -> 344 km²

229,593,563 m -> 229,594 km

1.5 m

6 (3 - 6 m)

8,158

29,252,942 m² -> 29 km²

9,750,981 m -> 9,751 km

3 m

12 (6 - 12 m)

11,631

20,496,071 m² -> 20.5 km²

2,277,341 m -> 2,277 km

9 m

9998

8,157

3,919,010 m² -> 3.9 km²

653,168 m -> 653 km

6 m

9997

196

202,374 m² -> 0.2 km²

16,864 m -> 16.9 km

12 m

9

16

30,175 m²

3,352 m

9 m

Processing with ESRI ArcGIS

Step 1 | Pre-Processing of bdlm2023_fluss

  1. Cut estuaries from bdlm2023_fluss by hand to adapt it to the CoastGEm dataset.

  2. Formerly removed polygons that were now missing were added back.

Step 2 | Assigning Emissions to Areas and Lines

  1. Feature to Point, with Inside checked.

  2. Extract Multi Values to Points, to add all values to the points.

  3. Remove unnecessary columns.

  4. Spatial Join, Closest.

Step 3 | Calculating Areas for Line Elements

  • Ditches:

    • AREA calculation according to the Breite_m column.

  • Channels:

    • Classes 0-3 m, 3-6 m, and 6-12 m were calculated with average widths of 1.5 m, 3 m, and 6 m, respectively.

  • Rivers:

    • Classes 0-3 m, 3-6 m, and 6-12 m were calculated with average widths of 1.5 m, 3 m, and 6 m, respectively.

  • AREA is geodesic area.

  • To adapt directly to the CoastGEm dataset, some polygons were cut, and the area of the Shape_Area was calculated anew by ArcGIS. The old area field was removed.

Step 4 | Calculating GHG Emissions, Transfer to Yearly Emissions, Transfer to total molecule

  • Multiply AREA with the GHG emissions. Column prefix is "EM_".

  • Multiply all EM_ columns by 365 to convert daily emissions to yearly emissions.

  • Calculate total GHG emissions by summing up the GHG emissions of the single elements and of rivers, ditches and channels individually.
    Clarification on units: Individual line and polygon feature layers retain emissions in mg/yr. The conversion from mg/yr to t/yr was applied only to the summary and total fields in the aggregated UBA grid dataset.

  • Transfer CO2-C to molecule, transfer N2O-N to molecule and transfer CH4-C to molecule.

    1. CO2-C to CO2: /12*44

    2. CH4-C to CH4: /12*16

    3. N2O-N to N2O: /28*44

  • Remove unnecessary columns for transfer dataset.

Step 5 | Aggregation to UBA Grid Cells

  1. Use SummarizeWithinWeigh in GeoAnalytics Desktop Tools to calculate the weighted GHG emissions per UBA grid cell for the line features. In contrast to the standard SummarizeWithin tool, the emissions are weighted by the water surface within each UBA grid cell.
    Statistics: Sum; Quantity Type: Count.

  2. Use Apportion Polygon to calculate the weighted GHG emissions per UBA grid cell for the polygon features.

Step 6 | Aggregate to One Shapefile

  1. Spatial Join (Analysis Tool) to aggregate all summarized single files into one polygon. Match option = identical.

Step 7 | Finalise Geodatabase for Transfer for Lines and Polygons

  1. Remove unnecessary columns.

  2. Ignore Shape_Length and Shape_Area.

  3. Area is provided by AREA_rl, AREA_rs, AREA_cl, AREA_cs and AREA_dl.

Files

Riverine Fluxes for Germany 2020_Submission 2026v13.pdf

Files (315.4 MB)

Name Size
md5:56b47c6362fec9a9e70225a11d9cb008
38.4 kB Download
md5:57863193409cdd59abbbea9cb58c1a58
13.7 MB Preview Download
md5:73e1b5b692719110efb76aa7a3448e66
283.5 MB Preview Download
md5:3a7dcc6cea770ceeba8815bef78c1b8f
18.1 MB Preview Download