Aside from the global study, the FFR methodology has been applied in a variety of different ways. The FFR methodology can a) be "downscaled", meaning that the global data sources are replaced with local data sources, and that new benchmark rivers are defined, and weighting may be updated based on a sensitivity analysis or b) the results of the FFR analysis (global or local) could be combined with other data layers, supporting new types of analysis.
The CSI, calculated at the river reach scale, can be aggregated to larger spatial scales. For this assessment, we aggregated the CSI to protected area, by calculated a weighted average of all river reaches inside a protected area.
In this application we combined the free-flowing status of rivers with an assessment of freshwater-related ecosystem values, and a water quality pressure assessment. The map below shows area of high/low connectivity, high/low water quality pressures, and the color intensity indicates increasing importance of freshwater values.
In this project, the free-flowing status was used as one of several information layers to prioritize areas for the protection of water resources. For more information, see Lehner et al. 2021 (DOI: 10.1002/aqc.3606)
Here, the CSI and the FFR status is used as indicators of impact as part of a system-scale hydropower planning model. The goal of this assessment was to identify future hydropower portfolios that minimize the impact of hydropower development on the Maranon river in Peru.
For this research on important migratory corridors in the Amazon Basin, we combined the FFR status of rivers with the mapping of migratory corridors of long-distance migratory fishes, dolphins and turtles to identify important freshwater connectivity corridors
Preparation and data review: review of existing (global) data and methodology; discussion with local groups, experts, and stakeholders of possible methods refinements and enhancements to better represent the local context; review of available data layers that could replace global data layers should be conducted with the help of regional experts
Data processing: The data review is followed by data collection and subsequent data processing using Geographic Information Systems (GIS). The result of the data processing are various layers representing pressures on natural connectivity (dams, roads, urban areas, etc.) at river reach scale.
Modeling: Once the relevant datasets and indicators are in place (at the river reach scale), the FRA model, a set of python tools bundled in a Python Toolbox is used to calculate or produce maps and statistical outcome
Refinements: The production of summary statistics and maps is typically followed by several cycles of refinement, based on stakeholder discussion, including local river experts.
The first step, and perhaps the most important one is typically to update the barrier database. Find the most reliable data layers and combine them into a set of point features. Identify and remove duplicate barriers, but review the attribute information first. If the attribute information is different between duplicate projects it is worthwhile choosing the most reliable, or even merge the two duplicates and only maintain reliable attributes form each duplicate to be transferred to the new barrier.
The barrier feature class contains a set of barrier points. The reservoir of a dam is not required to run the FFR analysis, however some attributes of the reservoirs must be linked to the barrier point database, e.g., Storage capacity (STOR_MCM).
Often times, barrier databases do not have precise locations, which can create uncertainty about its location in the river network. Within the FFR framework, barrier points are mapped to corresponding river reaches. It is required to assign each barrier point a values that identifies the river reach (Field: GOID) at which the barrier is located. Since the point is mapped to the river reach as a whole, the barrier point does not have to be located with great geographical precision, nor must its location be snapped to a river reach using GIS. However, it is important to assign the correct river reach (Field: GOID). To ensure this, each barrier must be manually verified, using the available barrier attributes.
For example, if the barrier has a field indicating upstream catchment area, it can be used to determine the correct river reach, using the upstream catchment area attribute of the river network. Similarly, if the barrier database indicates the river name at which the barrier is located, the user can find the river using satellite images and approximate its location.
The waterfalls database stems from Lehner et al, (2016), and aids in calculating the Degree of Fragmentation (DOF). Waterfalls naturally reduce connectivity in river networks by reducing or blocking species movement in river networks. Waterfalls should therefore be considered in the fragmentation analysis. They are blocking species movement upstream, while the downstream connectivity remains intact. As such waterfalls influence the degree of fragmentation (DOF).
Use any waterfall database you have available visualize the points together with the river network. The waterfalls database is not a direct source data layer to the model, but will be used to update the field "HYFALL" in the stream database. The attribute "HYFALL" in the streams feature class indicates the occurrence of at least one waterfall on the river reach (HYFALL = 1); otherwise 'HYFALL = 0". Only waterfalls that you consider to be a true barrier for species should be set to "1". Typically, rapids are included in waterfalls databases, but these are generally passable by most species, and should therefore be excluded as a waterfall.
Review each waterfall and determine if it actually constitutes a barrier for species. Use satellite images and other sources for verification. If in any doubt, for example, the waterfall is not creating a break (e.g. rapids), exclude the waterfall from the assessment.
Select all river reaches that have at least one waterfall, and set the HYFALL attribute to "1", otherwise to "0"
Urban areas are an indicator for dense infrastructure development, which includes sealed surfaces, including parking lots, various types of buildings, as well as difficult to map features, such as weirs, locks channels etc., which may all impact adjacent rivers and streams by inhibiting lateral connectivity.
The indicator is a combination of urban areas and nightlights within urban areas. The original urban dataset used in the FFR study was based on Modis imagery (Schneider et al., 2009). However, new data became available globally developed by the Global Urban Footprint (GUF) project (Esch et al., 2017). These data provide superior urban footprints at an unprecedented spatial resolution of 0.4 arc seconds (~12 m; for more details see https://www.dlr.de/eoc/en/PortalData/60/Resources/dokumente/guf/GUF_Product_Specifications_GUF_DLR_v01.pdf). Urban areas represented by this dataset are far more precise than in previous global maps, resulting in less overall coverage of urban areas compared to the dataset of the global FFR analysis (Schneider et al., 2009) which showed a more blurred representation of urban areas. Also, smaller urban areas are now more clearly distinguished and are found widespread in the study area.
The nightlight dataset used in the global FFR analysis (Doll, 2008) is no longer recommended for further analysis. Instead, we suggest using the dataset by Elvidge et al., (2017) at higher spatial resolution (90 m versus 500 m previously) providing superior imaging accuracy. This new nightlight composite map uses different units and value ranges than the previous dataset. New values are given in nW/area unit reaching up to a maximum of 107,000 nW/area unit (in areas where gas flares are visible year-round). Due to these extreme values an upper cut-off was applied to avoid distortion in our data standardization process caused by outliers. Elvidge et al. (2017) suggested a value of 300 nW as maximum value for city light sources, depending on atmospheric conditions and type of light source. To simplify the standardization with the FFR assessment, we used a threshold where pixels with values larger than 100 were set to 100. This threshold produced well-defined gradients both within cities and between cities, whereas the dataset previously used in the global FFR analysis typically showed maximum light values for most cities across the world, from large to small ones.
Like in the global FFR analysis, the nightlight values should be clipped to urban areas and to a 1-km buffer around the river.
The above described datasets are fairly detailed, and can be used in a local context. However, similarly suitable local datasets can be used instead. The following steps serve a general guideline on calculating the URB attribute for the FFR analysis.
Steps for calculating urban density for each river catchment
Identify and analyze different urban area extents, as well as nightlight intensity maps for the region and create a list of candidate layers.
Choose the most adequate layers for the analysis, considering accuracy (check with satellite images), completeness, consistency and attribute availability.
Reclassify (normalize) the nightlight intensity layer to a scale from 0 to 100.
Overlay the urban areas and the nightlight intensity to nightlight intensity only within urban extents.
Overlay the resulting raster with the buffer raster to eliminate urban areas outside the buffer area (map algebra)
Calculate zonal statistics using the stream catchment to calculate the average urban density in each river reach catchment.
A typical task for updating the road dataset is to use a local dataset that is more updated than the global one. In addition to replacing the road dataset with an updated version, we used new methodological elements to better capture the effects of roads on connectivity, especially for smaller roads. The global FFR analysis used the GRIP v3 dataset (Meijer and Klein Goldewijk, 2009), whereas for this study, we integrated used the polylines from OpenStreetMap as the basis for mapping.
While in the global FFR study the width for all road was kept constant at 50 meters regardless of type (highway, primary roads, etc.), the road width typically vary between 5 and 30 meters depending on road types (see table). Also, calculations of road density should be conducted at 5 m instead of 50 m resolution, which can contributed to a higher accuracy for estimating the road density in each river reach catchment.
The website https://overpass-turbo.eu/ has been used to extract the data for the Mekong region. A specialized script to extract data from OpenStreetMap may facilitate the data collection. Any other dataset available can also be used.
1
Identify and analyze different road layers and create a list.
2
Choose the most adequate road layer, considering accuracy (check with satellite images), completeness, consistency and attribute availability (such as road type)
3
Reclassify the roads into Highway, Primary, Secondary, and other roads.
4
Create appropriate buffers around Highways, Primary, and Secondary roads and other roads
5
Convert the buffered roads to a raster with high resolution (e.g. 50 meters)
6
Create a stream buffer of 2km around the existing stream network in the same resolution as the buffered roads.
7
Overlay the stream buffer raster with the road buffer to eliminate roads outside the buffer area (map algebra)
8
Calculate zonal statistics using the stream catchment layer to calculate the average road density in each river reach catchment. The values range between 0 and 100%.
9
Use the resulting table and create a table join to the stream network using the field GOID.
10
Update the field "RDD" with the road density values.
Benchmark rivers are rivers that were positively identified as free-flowing. The benchmarking dataset of reported FFRs was compiled from literature resources and expert input. For any local assessment, the research team should select their own benchmark rivers specific to the region. The benchmark rivers should be nominated and discussed by a group of experts, ideally as part of a workshop, webinar, or group call. Benchmark rivers can be named, drawn, marked up in existing maps, or provided digitally. The assessment team should then translate the information in to the provided template, using the following steps:
Use the streams feature class to select the benchmark river that should be add to the database.
Export only the selection and merge it with benchmark_rivers feature class. Using this technique, the required fields GOID and BB_ID are derived from the stream feature class, and will be present in the benchmark_rivers feature class.
Update the fields FFRID, BENCH_SRC, and Name_Expert.
The field "BENCH_SRC" indicates the source of the benchmark river. This can be the name of a researcher or a publication, a category such as "Expert" or "Publication" etc. The field is used to analyze the results of the benchmarking separately by source name or category.
A benchmark river can be a full river, i.e. all reaches with same BB_ID are included or part of a river. Only the segments that are benchmark rivers are included in this feature class, the rest are deleted. Each benchmark river must have a unique FFRID and the river name ("Name_Expert") must be consistent across the same FFRID.
