Recent advances in soil moisture measurement instrumentation and the potential for online estimation of catchment status for flood and climate forecasting: some experience from semi-arid catchments

1
Recent advances in soil moisture measurement
instrumentation and the potential for online
estimation of catchment status for flood and
climate forecasting some experience from
semi-arid catchments

• Garry Willgoose
• Earth and Biosphere Institute
• University of Leeds

2
Coworkers

• Walker, Rudiger, Grayson, Western U. Melbourne
• Kalma, Hemikara, Hancock, Saco U. Newcastle
  (Aust)
• Houser NASA Hydrology
• Woods NIWA, NZ
• Entekhabi MIT

3
The Core Hydrology Question

• How will emerging microwave remote sensing
  techniques for soil moisture assist in estimating
  the hydrology of catchments
• ERS (early 90s)
• AMSR (current)
• Hydros (planned)
• Can these techniques be integrated with new field
  instrumentation such as TDR?

4
SASMAS Objectives

• To ground validate AMSR-E measurements
• To test data assimilation of SM using AMSR-E or
  surrogate
• To test data assimilation of SM using discharge
  data (in heavily vegetated areas)
• To understand scaling properties of SM from Ha to
  100km2 scale in semi-arid
• To better understanding C, P balance in semi-arid
  catchments
• To understand floodplain as a temp storage for
  sediment from hillslope to river.

5
Time Domain Reflectometry TDR

• Integrated depth measurement at a point
• Difficult to install near surface
• Poor in cracking soils

6
Microwave Remote Sensing

• Typical wavelengths see top few cms of soil water
  and canopy water, impacted by soil surface
  condition (roughness).
• Repeat rate at best
• Radiometer twice/day _at_ low space resolution
  (10-30 km)
• Radar once month _at_ high resolution (20-30m)
• NOT measuring state of interest whole profile
  soil water at catchment scaleET.

7
But we can model profile soil water state

• Frequent measurements of surface soil moisture
  and model to simulate profile.
• Potentially with sufficient soil data can remote
  sense soil depth and water holding capacity.

8
Synthetic Simulations
Assimilation Period

• Surface soil moisture drives the estimation of
  soil moisture down the profile

9
Field Data

• Dotted simulations (surface moisture DA) best
  track the long-term data and the rise in May.

10
What about spatial patterns?

• Tarrawarra site (Grayson, Western, Willgoose,
  McMahon)
• Switch from arid (disorganised) to humid
  (organised).
• Is arid data disorganised or is it
  deterministically linked to spatially random
  soils properties? Single probe calibration.

11
SASMAS 01 Sampling

• 40 x 50km area
• North of Goulburn River within unforested region
• 4 teams over 3 days
• Sampled area about scale of AMSR pixel
• 225 soil moisture samples sites (4 gravimetric, 5
  TDR),
• 194 veg samples

12
Soil Moisture Results (SASMAS1 field campaign)
Gravimetric (0-1cm)
Theta Probe (0-6 cm)
13
The Stanley micro-site

• 1km x 2km for look at hillslope organisation of
  soil moisture. Semi-arid gt not topographic index
  soils, veg?
• 7 permanent TDR sites, 1-3 levels in the soil
• Runoff gauging

14
Sample of a at-a-point time series

• Strong response to rainfall and good correlation
  between depths.

15
Stanley Deep Soil Moisture

• Good correlation over 2km
• Appears likely to be able to calibrate a single
  probe (i.e. difference between sites due to
  permanent effects)
• Soil moisture correlations are parallel gt soil
  moisture process is vertical rather than a
  lateral topographic index type process

16
Stanley Surface Soil Moisture

• Correlation of surface soil moistures not as good
• Cross correlation with deeper soil moistures also
  not as good
• Is /- 10 accuracy good enough?
• Implications for remote sensing
• Soil moisture correlations definitely parallel

17
Short distance (sample scale) correlation

• Significant correlation scale of 0.2-0.5m. None
  up to 10m. Apparently unrelated to vegetation
  patterns. Also unrelated to SM status. Soils?
• Implication Hand held sampling is unrepeatable
  at the hillslope scale, though fixed sites
  indicate significant spatial correlation at this
  scale.
• More handheld sampling planned in March for the
  10-1000m scale.
• If SM correlation can be used as surrogate for
  soil variability what drives the soil
  variability? Implications for hydrology?

18
A tentative Conclusion from field data

• There appears to be a nontrivial spatial
  correlation 1-3 km (from surface soil moisture
  maps). Still processing recent SASMAS field
  campaigns.
• This correlation appears to be consistent through
  time (from correlation between permanent
  stations)
• We can assimilate profile soil moisture from
  surface measurements (whether radar or TDR )
• Conclusion The spatial correlation is a function
  of permanent properties of the catchment (e.g.
  soil, vegetation) rather than temporally
  uncorrelated fns such as rainfall.
• Implications We can (in principle) predict
  catchment scale soil moisture from single site
  TDR measurements (but short correlation scale gt
  permanent sites required not hand held)

19
Results from a synthetic data assimilation study
using stream runoff (for heavy veg sites)

• Root zone soil moisture well assimilated

• Surface soil moisture also well simulated but
  more sensitive to noise

20
Climate Model Initialisation
21
Soil moisture and climate

• Koster (NASA) showed that global climate
  dynamics/forecasts (months-years) sensitive to
  soil moisture (through energy partitioning ET)
• Entekhabi (MIT) showed bimodal continental
  climates as a result of rainfall feedback
• Eltahir (MIT) showed Sahel had three stable
  climate/vegetation states due to feedbacks.

22
Continental feedbacks
Ocean moisture
Rainfall
ET

• Relative strength of ET to ocean moisture
  determines the local feedback

23
How much latent heat transfer from vegetation?
From Choudhury (NASA)
24
Potential role of TDR and RS

• Vegetation extracts from deeper layers so raw
  remote sensing will not capture full behaviour
  profile modelling necessary.
• TDR ground truth soil moisture potentially
  calibratable to regional averages.
• Potential for a network attached to meteorology
  stations.

25
Conclusions

• Point monitoring and telemetering of soil
  moisture now possible and economic.
• Not easy to use upcoming RS data (concentrated on
  surface response).
• TDR point scale data appears to be
  regionalisable. Profile data would complement
  surface imaging.