Remote sensing and surface water hydrology

Remote sensing for surface water hydrology
• RS applications for assessment of hydrometeorological
states and fluxes
– Soil moisture, snow cover, snow water equivalent,
evapotranspiration, vegetation cover and water content, land
surface energy balance, water quality
• The above parameterize numerous physical, conceptual,
and empirical models of surface water dynamics, such
as runoff, infiltration, and streamflow
• Can runoff/streamflow be directly observed and
quantified with RS?
Not with any current technology
NRCS* Curve number method
Data and Parameters
• Digital Elevation model
• Watershed delineation
• Land use / land cover
• Soil hydrologic group
}=C
• Precipitation data
• Streamflow record
• Stream baseflow estimation
• Antecedent moisture condition
* NRCS – Natural Resources Conservation Service
urve
Number
Essential observations of a surface water system
Precipitation (rainfall)
Runoff
Streamflow
RS direct
quantification
Infiltration
Soil
moisture
Passive microwave
methods
very coarse spatial resolution
poor temporal resolution
expensive data
RS proxy
characterization
Landscape state
and energy flux
Infiltration
moderate spatial resolution
excellent temporal resolution
free data
Data and Methodology
• Remote Sensing Data
MODIS NASA’s Moderate Resolution Imaging Spectroradiometer
- Surface temperature (LST)
- Albedo
- Vegetation state
- NDVI (Normalized Difference Vegetation Index)
- EVI (Enhanced Vegetation Index)
- User derived MSI (Moisture Stress Index) and others
AMSR-E Advanced Microwave Scanning Radiometer
- Soil Moisture (resolution issues?)
- Vegetation water content and roughness
General methodology
Model parameterization based on:
• MODIS time-series landscape biophysicals
– High temporal resolution (daily but composited as 8 and 16 day
products)
– Moderate spatial resolution (0.25 - 1km2 pixel dim)
• NEXRAD radar (Stage III, MPE) precipitation estimates
• USGS gauged streamflow records
http://malibusurfsidenews.com/blog/uploaded_images/
USGS_Pic2488r-764415.jpg
NEXRAD MPE radar estimate of hourly precipitation rate for 4 July 2006 (21:00
GMT) for Sandies Creek watershed and surrounding region. Rates ranged from
0.0 mm/hr (black pixel) to 14.6 mm/hr (white pixel) for cells within the watershed
Daytime LST (8 day composite) for the Sandies Creek watershed for the
period 18 - 25 February 2002. Mean temperatures for this period ranged
from 24.9 C (dark pixels) to 29.3 C (light pixels).
NDVI (16 day composite) image of the Sandies Creek watershed for the period 18
February – 6 March 2002. Dark-toned and light-toned pixels represent low and
high NDVI values (stressed vegetation vs healthy), respectively.
How is LST coupled to soil
moisture (or vice versa)
• Heat flux from the earth’s surface
– Radiative flux (long wave thermal 9-13 μm)
– Sensible heat flux (convection and conduction)
– Latent heat flux (phase change)
• Is soil surface emissivity affected by soil
moisture? would this affect radiative, sensible, or
latent heat loss?
• Would a loss or gain of near-surface soil
moisture likely impact sensible or latent heat
flux?
From: http://upload.wikimedia.org/wikipedia/en/6/69/LWRadiationBudget.gif
Coupling vegetation to soil moisture
A Typical Vegetation Reflectance Spectra
visible
0.6
near infrared
middle infrared
Leaf water content
Leaf structure
0.4
0.3
0.2
Leaf chemistry
0.1
Wavelength
2424
2302
2180
2058
1936
1814
1692
1570
1448
1326
1204
1082
960
838
716
594
472
0
350
Reflectance
0.5
 nir   red
1  L
EVI 
 nir  C1  red  C2  blue  L
   red
NDVI  nir
 nir   red
Spectral response of leaf drydown as % water loss
red nir
0.8
0%
15%
0.6
25%
0.5
32%
0.4
41%
0.3
55%
0.2
100%
0.1
Wavelength (nm)
2495
2352
2209
2066
1923
1780
1637
1494
1351
1208
1065
922
779
636
493
0
350
Reflectance
0.7
 Mb2   Mb6
NDWI 
 Mb 2   Mb6
MSI 
 Mb6
 Mb 2
MSI mod 
Spectral response of leaf drydown as % water loss
Band 6
Band 7
Band 2
0.8
0%
0.7
15%
0.6
25%
0.5
32%
0.4
41%
0.3
55%
0.2
100%
0.1
Wavelength (nm)
2495
2352
2209
2066
1923
1780
1637
1494
1351
1208
1065
922
779
636
493
0
350
Reflectance
 Mb7
 Mb 2
Development of a benchmark model (CN) for
Sandies Creek for 2004
RS Model Development (2004)
• 6 MODIS parameters (LSTday, LSTnight, NDVI, EVI,
NDWI, MSI) x 2 states (raw, deseasoned) x 3 antecedent
offsets (0, 8, 16 days) = 36 regressors evaluated (plus
precipitation)
• Streamflow log transformed (normality assumptions)
• Final model: Prec, LSTdayr(1), EVIr(0)
4
3
log Q  0.957  0.439 P  0.192T  7.331I
where Q = streamflow, P = precipitation, T = LST, and I = EVI
All β1,2,3 estimates significant at P < 0.0001
β0 estimate significant at P < 0.04
2
logQ Actual
Final equation:
1
0
-1
-2
-3
-6
-4
-2
0
2
4
6
8
logQ Predicted P<.0001 RSq=0.84
RMSE=0.6876
10 12
Date
Jan-07
Sep-06
May-06
Jan-06
Sep-05
May-05
Jan-05
Sep-04
May-04
Jan-04
Sep-03
May-03
Jan-03
Sep-02
May-02
Jan-02
Temperature (C)
60
0
50
10
40
20
30
30
20
10
40
0
50
-10
60
Mean daily precipitation (mm)
2002 – 07* time series of daytime LST and
precipitation
precipitation
LST-day
LST-day deseasoned
seasonal mean
Sandies Creek calibration and validation results
Model period
E log series
Bias
Calibration All (n = 174)
0.677
0.207 (-0.471)*
2002 (n = 43)
0.616
1.037 (-0.399)*
2003 (n = 42)
0.477
-0.467
2004 (n = 45)
0.705
-0.516
2005 (n = 44)
0.785
-0.627
Validation All (n = 57)
0.453
-0.322
2006 (n = 46)
-0.028
-0.593
2007 (n = 11)
0.871
-0.293
Calibration
* Exclusion of July 2002
flood event
Validation
Sandies Creek validation results (linear space)