FDRI - Upper Severn

Kieran Khamis

Overview

  • My background and motivation
  • Important features of the Upper Severn catchment (USC)
  • Science questions to address in the USC
  • Learning for the wider community

My background

  • Hydro-ecologist and lecturer with experience developing sensor and monitoring networks
  • Stakeholder engagement in the UK and internationally (e.g. WMO Hydrohub)
  • Knowledge transfer working with a range of organisations (instrumentation companies, charities, NGOS and government agencies)

Motivation

Excited by the potential for FDRI to provide step change in:

  1. granularity of hydrological understanding

  2. digital data sharing and new data platforms

  3. hydrological training/teaching opportunities

Great advance in recent years regarding large sample hydrology (Camels) but there is still a need for experimental catchments that are well instrumented and can be used to develop and test new hypotheses and also new data driven models

champion digital data and reproducabe research within GEES and see FDRI as great way to embed new community standards

Open air labs - teaching practioners/students/citizen scientist how to make robust observations of the hydrosphere + new tech

The Upper Severn

  • Plynlimon flow, rainfall and evaporation archive
  • Dense network of flow and rain gauges in headwaters
  • Large water resource and flood mitigation reservoir
  • STW near outlet and CSOs in lower catchment

A diverse catchment

Landcover from UKCEH LCM 20231

I don’t want to dwel on this but just to say from a landcover perspective there is clear potential with a large forester sub catchment (Severn) and large pasture/grassland sub catchment- Afon Dulas

Upper Severn - UK context

Putting the existing gauges into a UK context - clearly some of the more upland catchments with low BFIs however the % forest covers a wide gradient + area spans 2 orders of magnitude

From Brunner et al. (2021)2

A Key paper by Brunner et al highlighted the many and diverse challenges that exsist in modelling and forecasting droughts and floods. From how we define events to lack of understaning about how processes will respond to env. change and how this then feeds through to uncertainty in modelling but I want to focus on 3 key questions that I think can be addressed in the USC

1. How can we optimise information (sensor) value for modelling and forecasting floods and droughts?

Given the large NERC investment a key outcome I would like to see and one that would benefit the community is clear guidance on generating infomation with maximal value for process understanding, modelling and forecasting

From Seibert et al. (2024)3

So I really liked this cartoon from a recent paper by Jan Seibert that highlights the challenges that anyone interesting in understanding the env. faces. How to optimize infomation gatehering on a limiting budget to answer you specific research questions or, in the context of gov. agencies, to fulfill your statutary monitoring requirements. They advocated for the use of model selection process starting with a saturated (overfitted model) and testing subsets to look at information loss

  • Environmental sensors crucial for monitoring the hydrosphere and extremes4

  • Modelling and network/information theory optimise data for flood/drought prediction

  • USC ideal location to generate new understanding of what, where and when to measure

We need robust and reliable env. sensors to capture extremes. To generate new understanding of how best to approach this in the uplands USC is ideal test bed…From a dense network we can use the frame work of siebert to reduce the complexity of the data set (model selection framework). Or more complicated methods relating to network and infomation theory identify overlap/or similarity in nodes and

2. What are the impacts of land cover and management on drought extent and propagation?

  • Assess spatial CCF and spatial extents/patterns of past and future extremes

  • Explain/model patterns based on propagation through the various components of the hydrosphere

  • USC opportunity to instrument sub-catchments spanning a gradient of management and basin attributes

3. Can we better quantify the causal factors in flooding to improve modelling and forecasting?

Key questions for flood crisis managers are still not adequately answered: (a) How high will the river rise? (b) When will the river reach its peak? (c) How long will flooding last?

The missing links

  • How to best use spatiotemporal rainfall information5
  • Understanding the role of catchment wetness and interaction with basin properties
  • Assessing and accounting for routing effects6
  • Unsteady flows and hysteresis generating observation uncertainty7

New understanding from Upper Severn

  • Historical data analysis - 32 storage rain gauges and CEH AWS8,9

  • Build upon existing monitoring infrastructure

  • Assess links to rainfall radar/satellite10 and COSMOS-UK soil moisture11

  • Tracking flood waves and assessing spatial heterogeneity of inputs and (sub)catchment properties

Links to other FDRI sites and wider community

  • New understanding of network design for upland catchments to optimise future monitoring

  • Comparative hydrology (i.e. within FDRI) - transferability of new process understanding and modelling approaches

  • New knowledge of process (spatial) scaling for flood and drought prediction

Considering the combined network in relation to droughts/floods (e.g. rain gauges, soil moisture, ) Network design approaches relied largely on statistical methods, most commonly based on the standard error in estimating regional discharge at ungauged sites.

  • Better constrained models and understanding of variance - bias tradeoffs

  • “Sense checking” outputs from large sample hydrology and complex ML models

As we have more sophisticated modelling tools and a move towards ML and data driven models - well instrumented experimental catchments are more important than ever. Rich data sets can help us understand complexity and transfer ability tradeoffs in modelling. Are we fitting models that can recreate the data well but are correlative rather causative

References

1.
Morton, R. D., Marston, C. G., O’Neil, A. W. & Rowland, C. S. Land Cover Map 2023 (10m classified pixels, N. Ireland). (2024) doi:10.5285/17223091-CA33-41F8-BD5B-BDD2A222CDAE.
2.
Brunner, M. I., Slater, L., Tallaksen, L. M. & Clark, M. Challenges in modeling and predicting floods and droughts: A review. WIREs Water 8, (2021).
3.
Seibert, J., Clerc-Schwarzenbach, F. M. & (Ilja) van Meerveld, H. J. Getting your money’s worth: Testing the value of data for hydrological model calibration. Hydrological Processes 38, (2024).
4.
Agarwal, A. et al. Optimal design of hydrometric station networks based on complex network analysis. Hydrology and Earth System Sciences 24, 2235–2251 (2020).
5.
Saharia, M. et al. On the Impact of Rainfall Spatial Variability, Geomorphology, and Climatology on Flash Floods. Water Resources Research 57, (2021).
6.
Tarasova, L. et al. Causative classification of river flood events. WIREs Water 6, (2019).
7.
Muste, M., Kim, D. & Kim, K. A flood-crest forecast prototype for river floods using only in-stream measurements. Communications Earth & Environment 3, (2022).
8.
Newson, M. The plynlimon floods of august 5th/6th 1973. vols. IH Report No.26 (1975).
9.
Archer, D. R. The use of flow variability analysis to assess the impact of land use change on the paired Plynlimon catchments, mid-Wales. Journal of Hydrology 347, 487–496 (2007).
10.
Biggs, E. M. & Atkinson, P. M. A comparison of gauge and radar precipitation data for simulating an extreme hydrological event in the Severn Uplands, UK. Hydrological Processes 25, 795–810 (2011).
11.
Levy, P. E. Mapping soil moisture across the UK: assimilating cosmic-ray neutron sensors, remotely sensed indices, rainfall radar and catchment water balance data in a Bayesian hierarchical model. Hydrology and Earth System Sciences 28, 4819–4836 (2024).