Modelled erosional impacts on soil and freshwater nitrogen and phosphorus dynamics in Lake Victoria’s catchment area, 2000-2022
https://doi.org/10.5285/02977a5d-00a0-44f8-baee-d2e0eecb5df0
Cite this dataset as:
Sharps, K.; Feeney, C.; Bisht, T.; Thomas, A.; May, L. (2025). Modelled erosional impacts on soil and freshwater nitrogen and phosphorus dynamics in Lake Victoria’s catchment area, 2000-2022. NERC EDS Environmental Information Data Centre. https://doi.org/10.5285/02977a5d-00a0-44f8-baee-d2e0eecb5df0
Download/Access
PLEASE NOTE:
By accessing or using this dataset, you agree to the terms of the relevant licence agreement(s). You will ensure that this dataset is cited in any publication that describes research in which the data have been used.
This dataset is available under the terms of the Open Government Licence 
Bulk download options
You can use WGET to download data. For example
wget --user=YOUR_USERNAME --password=YOUR_PASSWORD --auth-no-challenge https://catalogue.ceh.ac.uk/datastore/eidchub/02977a5d-00a0-44f8-baee-d2e0eecb5df0
This dataset presents model predictions of soil erosion impacts on land and freshwater systems in the Lake Victoria Basin. These impacts include: (1) soil nutrient balances as a function of nitrogen (N) and phosphorus (P) loads to the land surface from fertiliser, manure and atmospheric deposition, minus rates of nutrient loss via soil erosion; and (2) total sediment and nutrient exports to water aggregated per sub-catchment (t/yr), which combines nutrients lost in surface runoff and subsurface flows with additional losses associated with erosion of existing soil nutrient stocks. These data are particularly useful for better understanding terrestrial N and P dynamics, and should provide valuable data for analysis of soil fertility, food security and water quality.
These data were generated for each year (2000, 2005, 2010, 2015 & 2022) using two models from the InVEST ecosystem services platform: the Nutrient Delivery Ratio (NDR) and Sediment Delivery Ratio (SDR) models. Each of these models relies on a set of simple empirical functions, which has the advantage of requiring few model inputs that are easy to provide and allow for rapid modelling of large spatial areas, but come with the disadvantage of missing potentially important processes. For example, the SDR model only considers soil erosion via rills and sheetwash; it does not capture gullies, landslides or river channel change. Additionally, it was not possible to calibrate either model due to a scarcity of observational data. However, the models do produce generally expected patterns of sediment and nutrient dynamics. Hence, the outputs should still be useful for identifying areas of relatively high and low values and for assessing relative changes over time.
The outputs themselves consist of raster gridded maps at 90 x 90 metre resolution, as well as vector polygons representing sub-catchments. The sub-catchments aggregate the total sediment, N and P exports from land to water for all grid cells present within each sub-catchment's boundary. All maps cover the full catchment area (196,883 sq. km excluding the lake itself) of the Lake Victoria Basin.
These data were generated for each year (2000, 2005, 2010, 2015 & 2022) using two models from the InVEST ecosystem services platform: the Nutrient Delivery Ratio (NDR) and Sediment Delivery Ratio (SDR) models. Each of these models relies on a set of simple empirical functions, which has the advantage of requiring few model inputs that are easy to provide and allow for rapid modelling of large spatial areas, but come with the disadvantage of missing potentially important processes. For example, the SDR model only considers soil erosion via rills and sheetwash; it does not capture gullies, landslides or river channel change. Additionally, it was not possible to calibrate either model due to a scarcity of observational data. However, the models do produce generally expected patterns of sediment and nutrient dynamics. Hence, the outputs should still be useful for identifying areas of relatively high and low values and for assessing relative changes over time.
The outputs themselves consist of raster gridded maps at 90 x 90 metre resolution, as well as vector polygons representing sub-catchments. The sub-catchments aggregate the total sediment, N and P exports from land to water for all grid cells present within each sub-catchment's boundary. All maps cover the full catchment area (196,883 sq. km excluding the lake itself) of the Lake Victoria Basin.
Publication date: 2025-08-12
View numbers valid from 12 August 2025 Download numbers valid from 12 August 2025 (information prior to this was not collected)
Formats
TIFF, geopackage
Spatial information
Study area
Spatial representation types
Raster
Vector
Vector
Spatial reference system
ESRI:102022 - Albers for Africa equal area
Temporal information
Temporal extent
2000-01-01 to 2022-12-31
Provenance & quality
The model outputs were ultimately generated from the InVEST platform’s NDR and SDR models. Each of these has a similar structure that uses a digital elevation model (DEM), catchment polygon and threshold flow accumulation (in this case, 1000) to generate a river network and a surrounding topographically varying land surface within the catchment. This is combined with simple formulae that represent land-water connectivity to allow the models to calculate the delivery of matter (here, sediments, N and P) from land to water. Gross loss rates of sediment, N and P from the land surface are governed by combinations of land cover (with each land cover class represented in a map layer and corresponding table of biophysical coefficients), rainfall, soil structure and slope length and steepness. Connectivity parameters govern how much of these gross losses from land ultimately impact the river network (becoming “net” losses).
Topsoil (0-20 cm) nutrient concentration maps from iSDA-soil were combined with erosion rate predictions to estimate rates of soil total nitrogen and total phosphorus fluxes with soil loss. These soil erosion-associated nutrient losses were subtracted from maps produced by the NDR model that represent the total N and P loads from fertilisers, manure and atmospheric deposition to compute nutrient balances on all land grid-cells. Soil nutrient levels were also multiplied by sediment export rates maps produced by the SDR model and aggregated for each subcatchment (i.e. summing values from all grid cells within a subcatchment). These aggregated values were added to aggregated N and P exports to the river network (produced by the NDR model) to obtain more complete total N and P exports from land to water.
QA checks were undertaken to ensure all raster layers were exactly aligned with one another to the same spatial extent, coordinate reference system and resolution (exactly 90 m). Additionally, model outputs were quantitatively evaluated against the limited pool of observational data available to us from catchment outlets all along Lake Victoria’s shoreline. Predicted sediment, and total N and P exports (that each account for the erosion of soil nutrient stocks in addition to losses via runoff captured by the NDR model) were evaluated against available observational measurements from the years 2000 and 2005. In most cases, predicted sediment, N and P losses were predicted to be within 1 order of magnitude of the observations (a widely accepted standard for evaluation of catchment-level exports – see the supporting documentation for more details) and thus our results are shown to be reasonably accurate.
To reduce the overall sizes of each data file, we converted the data from floating point to integer format. Before converting our data into integer format, we multiplied each raster layer (individual band in our GeoTiffs) by 100. Therefore, the user of our data will need to divide the values stored in our raster layers by 100 to get the “true” values (in units of kg/grid cell/year to 2 decimal places).
After converting our data into integer format, each raster band was harmonised to the same coordinate reference system and spatial extent before stacking into multi-band GeoTiff files. The data contained in the geopackage file requires no transformation prior to analysis or other use. The geopackage consists of 5 layers (named after each simulated year: 2000, 2005, 2010, 2015 and 2022). Please see the Supporting Documentation for further details on attributes recorded in the geopackage layers.
Topsoil (0-20 cm) nutrient concentration maps from iSDA-soil were combined with erosion rate predictions to estimate rates of soil total nitrogen and total phosphorus fluxes with soil loss. These soil erosion-associated nutrient losses were subtracted from maps produced by the NDR model that represent the total N and P loads from fertilisers, manure and atmospheric deposition to compute nutrient balances on all land grid-cells. Soil nutrient levels were also multiplied by sediment export rates maps produced by the SDR model and aggregated for each subcatchment (i.e. summing values from all grid cells within a subcatchment). These aggregated values were added to aggregated N and P exports to the river network (produced by the NDR model) to obtain more complete total N and P exports from land to water.
QA checks were undertaken to ensure all raster layers were exactly aligned with one another to the same spatial extent, coordinate reference system and resolution (exactly 90 m). Additionally, model outputs were quantitatively evaluated against the limited pool of observational data available to us from catchment outlets all along Lake Victoria’s shoreline. Predicted sediment, and total N and P exports (that each account for the erosion of soil nutrient stocks in addition to losses via runoff captured by the NDR model) were evaluated against available observational measurements from the years 2000 and 2005. In most cases, predicted sediment, N and P losses were predicted to be within 1 order of magnitude of the observations (a widely accepted standard for evaluation of catchment-level exports – see the supporting documentation for more details) and thus our results are shown to be reasonably accurate.
To reduce the overall sizes of each data file, we converted the data from floating point to integer format. Before converting our data into integer format, we multiplied each raster layer (individual band in our GeoTiffs) by 100. Therefore, the user of our data will need to divide the values stored in our raster layers by 100 to get the “true” values (in units of kg/grid cell/year to 2 decimal places).
After converting our data into integer format, each raster band was harmonised to the same coordinate reference system and spatial extent before stacking into multi-band GeoTiff files. The data contained in the geopackage file requires no transformation prior to analysis or other use. The geopackage consists of 5 layers (named after each simulated year: 2000, 2005, 2010, 2015 and 2022). Please see the Supporting Documentation for further details on attributes recorded in the geopackage layers.
Licensing and constraints
This dataset is available under the terms of the Open Government Licence
Cite this dataset as:
Sharps, K.; Feeney, C.; Bisht, T.; Thomas, A.; May, L. (2025). Modelled erosional impacts on soil and freshwater nitrogen and phosphorus dynamics in Lake Victoria’s catchment area, 2000-2022. NERC EDS Environmental Information Data Centre. https://doi.org/10.5285/02977a5d-00a0-44f8-baee-d2e0eecb5df0
Related
This dataset is included in the following collections
Correspondence/contact details
Authors
https://orcid.org/0000-0003-3265-1505 Other contacts
Rights holder
UK Centre for Ecology & Hydrology
Custodian
NERC EDS Environmental Information Data Centre
info@eidc.ac.uk
Publisher
NERC EDS Environmental Information Data Centre
info@eidc.ac.uk