a platform for evaluating the latest hydrological, climatic and water quality sensors and in parallel evaluating the latest dynamic models of these resultant high frequency data
Continuous high-frequency monitoring is ongoing at four small watersheds that drain into the Llyn Brianne reservoir in the headwaters of the River Towy in the Cambrian Mountains, mid-Wales.
The watersheds were instrumented with identical stream monitoring equipment in late 2012, initially as part of the NERC-funded Diversity in Upland Rivers for Ecosystem Service Sustainability (DURESS) project. These replicated, paired watersheds consist of two grassland sites (LI6 & LI7) and two coniferous forest sites (LI3 & LI8).
High-frequency (15-minute) monitoring of hydrometric, water quality and climatic variables began in January 2013 and continues for selected variables. Instrumentation included SMART sensors for measuring variables such as nitrate and dissolved organic carbon in-situ. This has yielded novel hydrochemical data spanning many months, and is suitable for SMART modelling approaches.
Location of the site can be found on our interactive map.
Site map (from Jones et al., 2014, Hydrol Res, 45.6, 868)
Llyn Brianne dam and spillway
Coordinates: N 52° 08′ 32.0″, W 003° 44′ 02.0″
Area: 81 ha
Land Cover: Coniferous plantation (Sitka Spruce dominant)
Geology: Lower Palaeozoic shales, mudstones, greywackes and grits.
Soils: 47% podzol, 48% gleysol, 5% histosol
Mean Annual Precip: 2,100 mm (8 yr average, 2002-2010)
SMART sensors: Campbell Scientific CR1000 with temperature, water level, turbidity, electrical conductivity and pH sensors.
S::CAN spectro::lyser measuring turbidity, nitrate-N, DOC, TOC, apparent colour, true colour.
SMART models: RIVC algorithm (Rainfall-hydrogen ion; Rainfall-DOC).
Landowner: Natural Resources Wales
Coordinates: N 52° 07′ 57.2″, W 003° 43′ 19.8″
Area: 69 ha
Land Cover: Grassland
Geology: Lower Palaeozoic shales, mudstones, greywackes and grits.
Soils: 45% podzol, 15% gleysol, 40% histosol
Mean Annual Precip: 2,100 mm (8 yr average, 2002-2010)
SMART sensors: Campbell Scientific CR1000 with temperature, water level, turbidity, electrical conductivity and pH sensors.
S::CAN spectro::lyser measuring turbidity, nitrate-N, DOC, TOC, apparent colour, true colour.
Automatic Weather Station (CR1000) measuring rainfall, temperature, incoming solar radiation, photosynthetically active radiation, relative humidity, wind speed, wind direction.
SMART models: RIVC algorithm (Rainfall-hydrogen ion; Rainfall-DOC).
Landowner: Mr Roger Davies
Coordinates: N 52° 07′ 44.8″, W 003° 43′ 39.7″
Area: 69 ha
Land Cover: Grassland
Geology: Lower Palaeozoic shales, mudstones, greywackes and grits.
Soils: 33% podzol, 18% gleysol, 49% histosol
Mean Annual Precip: 2,100 mm (8 yr average, 2002-2010)
SMART sensors: Campbell Scientific CR1000 with temperature, water level, turbidity, electrical conductivity and pH sensors.
S::CAN spectro::lyser measuring turbidity, nitrate-N, DOC, TOC, apparent colour, true colour.
EXO1 sonde measuring borehole pH, temperature, electrical conductivity, DO.
SMART models: RIVC algorithm (Rainfall-hydrogen ion; Rainfall-DOC).
Landowner: Mr Roger Davies
Coordinates: N 52° 07′ 32.0″, W 003° 44′ 50.7″
Area: 121 ha
Land Cover: Coniferous plantation (Sitka Spruce dominant)
Geology: Lower Palaeozoic shales, mudstones, greywackes and grits.
Soils: 44% podzol, 9% gleysol, 48% histosol
Mean Annual Precip: 2,100 mm (8 yr average, 2002-2010)
SMART sensors: Campbell Scientific CR1000 with temperature, water level, turbidity, electrical conductivity and pH sensors.
S::CAN spectro::lyser measuring turbidity, nitrate-N, DOC, TOC, apparent colour, true colour.
EXO2 sonde measuring fluorescent dissolved organic matter (fDOM), blue-green algae, chlorophyll-a, dissolved oxygen, pH, turbidity, and oxidation reduction potential (ORP)
SMART models: RIVC algorithm (Rainfall-hydrogen ion; Rainfall-DOC).
Landowner: Mr Roger Davies
Jones, T.D., Chappell, N.A. and Tych, W. 2014. First dynamic model of dissolved organic carbon derived directly from high frequency observations through contiguous storms.Environmental Science & Technology, 48: 13289-13297. View online.
Jones, T.D. and Chappell, N.A. 2014. Streamflow and hydrogen ion interrelationships identified using Data-Based Mechanistic modelling of high frequency observations through contiguous storms.Hydrology Research, 45(6): 868-892. View online.
Littlewood, L.G. 1989. The Dynamics of Acid Runoff from Moorland and Conifer Afforested Catchments Draining into Llyn Brianne, Wales. PhD Thesis, University of Wales.
The four sites have identical instrumentation housed in water quality boxes with in-stream protective manifolds. The following images show the infrastructure and sensor system set up:
Water quality box to house and protect equipment (e.g. autosampler, dataloggers, batteries) – LI8
Water samples collected using an ISCO autosampler to allow calibration of water quality sensors to laboratory measured values
Manifold to hold and protect water quality sensors in the stream – LI8
Trapezoidal flume (for accurate measurement of stream discharge), raingauge and tripod-mounted datalogger with solar panel – LI6
S::CAN spectro::lyser smart sensor removed from housing for cleaning – LI3
Automatic Weather Station – LI6
EXO2 multiparameter sonde removed from housing for cleaning – LI8
Dataloggers and sensor interfaces housed on water quality box door – LI8
The datasets for these four sites currently include:
January 2013 – present: rainfall, streamflow, stream temperature, electrical conductivity, plus all AWS variables.
January 2013 – August 2014: turbidity, nitrate-N, DOC, TOC, apparent & true colour, pH.
Example data (right): First 6 months of 15-min sampled & calibrated DOC (mg/L) from watersheds LI3-8 (8 Jan – 25 Jun 13). Collected using the S::CAN spectro::lyser and con:nect interface with twice weekly manual cleaning of instrument lenses with 10% HCl. Rainfall for the period also shown in blue.
Please contact us if you are interested in using our datasets.
The deployment of SMART sensors and subsequent SMART modelling at the Llyn Brianne sites has been an important aspect of the NERC-funded DURESS project. This project brings together many institutions and disciplines to assess how biodiversity provides the ecosystem services on which people depend.
The Lancaster University team are quantifying the link between distributed changes in catchment land-use and management, climate, river ecological variables and river biota for a range of physical characters, and at a range of scales.
Three key questions are being addressed by the team using data obtained from SMART sensors and using SMART models:
SMART model identification routines have been applied to the high-frequency observations (recorded through contiguous storms) from these four watersheds. The models are able to simulate the rapid dynamics observed in the hydrometric behaviour (rainfall-streamflow) and in the water quality behaviour (e.g. rainfall-H+ load and rainfall-DOC load). The latter is the first dynamic model of stream DOC export from such high-frequency observations.
Rainfall time-series (blue) plus observed DOC (black) and simulated DOC (red) for all four basins for a period of three contiguous storms in 2013
The model identification routine, known as the Refined Instrumental Variable in Continuous Time, (RIVC) algorithm is based on transfer functions, and gives high simulation efficiencies for these analyses with constrained uncertainty.
The modelled water chemistry response and the hydrometric response display two separate response components, i.e. second-order dynamics, with a fast pathway (e.g. soil-water) and slow pathway (e.g. movement through underlying drift strata). The similarity of the model structures enables preliminary physical interpretation of the dynamic responses.
For example, most DOC export at these sites was through the faster hydrometric pathway and was exhausted in the slower pathway, while as temperature increased from winter to spring there was increased delay between the initial hydrometric response and the response of the DOC load.
This work has highlighted the critical need for high-frequency (sub-hourly) monitoring of water quality variables in order to avoid distortion of the true dynamics by models (‘temporal aliasing’).
Full details of the modelling results and implications are given in Jones and Chappell (2014) and Jones et al. (2014).
DOC load simulated from rainfall at LI3 shows a second-order response, with 56% of DOC delivery following a fast pathway, and 44% of response following a slow pathway.
©2015 Lancaster University