MEMORANDUM - epa.wa.gov.au D6... · Lake DisapPointment: NDVI, NDWI AND ET calculations p3...

STR E ET PO STA L C ONT A CT

7 Forrest AvenueEast Perth 6004WESTERN AUSTRALIA

PO Box 6917East Perth 6892WESTERN AUSTRALIA

+61 (0)8 6218 0900 P+61 (0)8 6218 0934 [email protected]

AB N 68 120 964 650 www.hydrobiology.biz

MEMORANDUMT O : Dan Tenardi; Lisa Chandler

C C :

S E N D E R : Phil Whittle

D A T E : 15th August 2017

LAKE DISAPPOINTMENT: NDVI, NDWI AND ET CALCULATIONSHydrobiology has completed the GIS tasks and extraction of summary data for Normalised DifferenceVegetation Index (NDVI), Normalised Difference Wetness Index (NDWI or Wetness Index) and ET (ActualEvapotranspiration) data for Lake Disappointment. The years 2004, 2006 and 2008 were chosen for analysis ofLandsat 4-5 imagery (NDVI and NDWI) based on suitable dry season conditions and availability of high-qualitycloud-free imagery. ET data was obtained from the NCI WIRADA dataset.

NDVI provides a reliable measure of chlorophyll content or greenness of the vegetation. It is suggested thatunvarying NDVI values are typically experienced in vegetation that has access to groundwater, and thisrelationship can often be more apparent at the end of the dry season when water is limited e.g. Barron et al2012).

An assessment was also conducted of evapotranspiration (ET) for the same years (2004, 2006 and 2008) usingthe CSIRO MODIS reflectance-based scaling evapotranspiration (CMRSET) data set. Groundwater-dependentvegetation (GDE) is commonly associated with higher rates of ET, hence by calculating ET it may be possible tohighlight potential GDEs, especially when taken in concert with the NDVI and NDWI measures (e.g.Guerschman et al. 2009).

The use of remote sensing to assess vegetation function has recently become an established technique.“Remote sensing provides a robust and spatially explicit means to assess not only vegetation structure andfunction but also relationships amongst these and climate variables” (Eamus et al. 2015).

Brief MethodsNDVI and NDWI Method

The general approach to identification of potential GDEs followed Barron et al. (2012) – “Mapping groundwater-dependent ecosystems using remote sensing measures of vegetation and moisture dynamics”. This involvedusing multi-spectral imagery to derive Normalised Difference Vegetation Index (NDVI) and NormalisedDifference Wetness Index (NDWI) parameters using the red, near infrared and short-wave infrared bands (asdescribed in Barron et al. 2012). Landsat imagery at a spatial resolution of 30 m × 30 m was obtained from theUSGS Earth Explorer web service for the Lake Disappointment study area (Figure 1). Images from the end ofthe wet season (Feb-April) until the end of the dry season (Sept-Nov) were obtained for three years (2004,2006 and 2008). These years were chosen based on a combination of an extended dry spell of several months(rainfall records from Telfer Aero) and suitable cloud-free imagery available for the whole study area. Imagerywas obtained for a temporal frequency of monthly where possible.

Landsat imagery from the Landsat 4-5 archive was determined to be the most appropriate for this analysis as itcontained a suitable temporal and spatial coverage. The Landsat 7 imagery contains a sensor error that may

Lake DisapPointment: NDVI, NDWI AND ET calculations www.hydrobiology.biz p2

have made comparison between the older Landsat 4-5 and Landsat 7 imagery (which covers years 2009 to2013) subject to error. The current Landsat 8 imagery is only processed for the study area for 2016 to present,which represents an unusually wet period.

Raw imagery in GeoTIFF format was downloaded from the USGS website and Bands 3, 4 and 5 (Table 1) wereimported into the Manifold GIS software package for processing. Each image was clipped to a standardcoverage area (Figure 1) and the NDVI and Wetness values calculated using Python scripting within theManifold software. A vegetation community (Floristic community) map provided by Reward Minerals (producedby Botanica) was used to select zones for generation of statistics by vegetation type. Full descriptive statisticswere generated for the 2006 study year and average values for the 2004 and 2008 study years, for each image(Table 2). Descriptive statistics were generated by exporting the NDVI/Wetness values for each vegetationtype, for each image, into Excel. Averages were generated within Manifold GIS.

Figure 1 and Table 4 provide background information on the vegetation communities analysed. Note that themap colour and graph colour for each vegetation community correspond, to allow easier cross referencing.

Table 1 Landsat 4-5 Thematic mapper band information

Bands Wavelength(micrometers)

Resolution(meters)

Band 3 - Red 0.63-0.69 30

Band 4 - Near Infrared (NIR) 0.76-0.90 30

Band 5 - Shortwave Infrared (SWIR) 1 1.55-1.75 30

Table 2 Number of images processed by year

Year Number of NDVI images Number of Wetness images Total

2004 8 8 16

2006 9 9 18

2008 8 8 16

TOTAL 25 25 50


Evapotranspiration Method

Estimates of actual evapotranspiration (AET) was calculated for the study area using satellite imagery from the‘CSIRO MODIS reflectance based scaling evapotranspiration’ (CMRSET) data set (250 m resolution). This dataset was developed by Guerschman et al. (2009) and it provides an estimate of AET across Australia, based onMODIS reflectance and short wave infra-red data, and gridded meteorological surfaces.

In brief, the CMRSET algorithm uses reflectance data from the MODIS satellite to calculate ET across theAustralian continent. AET is calculated from potential ET (PET) by applying a ‘crop factor’ which incorporatesthe enhanced vegetation index (EVI) and global vegetation moisture index (GVMI). The algorithm wascalibrated by comparing estimated AET with measured AET from seven eddy covariance towers aroundAustralia covering a variety of landscapes (forest, savannah, grassland, floodplain and lake). CMRSET wasfurther validated by comparing estimated AET with ‘surrogate AET’ (precipitation minus streamflow) in 227unimpaired catchments around Australia Guerschman et al. (2009).

A cautious approach is required when attempting to make inference about the presence of GDE from AET forseveral reasons. The first being that the amount of ET for a given vegetation type can be influenced by otherfactors such as vegetation health, leaf area index and how water tolerant the vegetation type is (Gonzalez2015, Woods et al. 2016). Secondly the calibration method used for the CMRSET was conducted in areas withrainfall of greater than 250 mm and not in low rainfall areas like the study area. Thirdly, Van Dijik et al. 2015found that this method has a tendency to overestimate ET from salt lakes, however he also suggested thatresults for areas other than salt lakes are more robust and the ET values potentially more representative ofwhat is actually happening.

Raw imagery in .nc format was downloaded from the NCI (National Computational Infrastructure) website andimported into QGIS software package for processing. A vegetation community (Floristic community) mapprovided by Reward Minerals (produced by Botanica) was used to generate statistics by vegetation type for2004, 2006 and 2008 (Table 3). Descriptive statistics were generated by exporting the ET values for eachvegetation type, for each image, into Excel.

Table 3 Number of images processed by year

Year Number of ET images

2004 11

2006 10

2008 11

TOTAL 32

Calculation of Estimated Groundwater Evapotranspiration

Groundwater evapotranspiration (ETg) refers to the water losses from groundwater due to transpiration, directwater uptake through roots from GDEs, and direct evaporation (e.g. from any wet surface including soil or landsurface). Groundwater-dependent vegetation is commonly associated with a comparatively higher rate ofevapotranspiration (ETg), hence by identifying areas where ETg exceeds rainfall on an annual basis it ispossible to predict potential GDEs (O’Grady et al. 2011). It is important to know that this method is asimplification of the system and does not include a direct measure of evaporation and assumes that 100% ofthe ETg comes from transpiration. Eamus et al. 2015 estimated that the average error associated with thismethod was about 12%, however it is likely to be much greater in environments where groundwater isexpressed at the surface and or moist soil i.e. salt lakes and wetlands. In these types of environments there willbe greater groundwater expression and hence higher ET, and it is highly likely that these high ET are not due to


the presence of GDE but due to limitations of the method. Hence caution needs to be applied when makinginference about GDEs associated with groundwater expressed at the surface. The rainfall data used for thiscalculation came from the Telfer rain gauge station which is 200 km from the study site, this data was usedbecause it is the closest and most complete data set that was available. The distance of the rainfall data fromthe study site is another limitation of this method.

The spatial resolution of the ET data allows for a pixel size of 250m2. The vegetation in this area can be highlypatchy and may not completely fill a pixel, hence other components will be incorporated into the calculations.This limitation needs to be considered when interpreting the ET results.

Groundwater evapotranspiration (ETg) can be calculated from satellite imagery using NDVI and rainfall usingthe following formula in which NDVI* is the peak season normalised NDVI. It is important to remember thatthese ETg figures are estimates and to get a more accurate result it is suggested that the model is calibratedusing sites with known ET.

= ( − ) × ∗NDVI* was calculated by subtracting the NDVI for the area that had no vegetation (NDVIz , i.e. lake bed) fromthe summer peak season NDVI for each vegetation unit and dividing this by NDVI at saturation (NDVIm ; themaximum value obtained by any vegetation unit), minus the NDVI for the area that had no vegetation (i.e. lakebed) (Eamus et al. 2015).

∗= −−


The following figures have been provided as a graphical presentation of the results:

Figure 1 Map of vegetation communities used in the NDVI/Wetness/ET analysis ............................................6

Figure 2 Comparison of Lake Disappointment and Barron et al. (2012) study area NDVI values.......................8

Figure 3 Ratio of NDVI values from late wet to end of dry season .................................................................9

Figure 4 Ratio of NDWI values from late wet to end of dry season ..............................................................10

Figure 5 Ratio of NDVI values from late wet to end of dry season– distance from 1:1 (no change) ................11

Figure 6 2004 NDVI – average values over the dry season by vegetation type .............................................12

Figure 7 2004 Wetness Index – average values over the dry season by vegetation type ...............................13

Figure 8 2006 NDVI – average values over the dry season by vegetation type .............................................14

Figure 9 2006 Wetness Index – average values over the dry season by vegetation type ...............................15

Figure 10 Histograms of NDVI relative frequency (0.05 unit bins) for 8th June 2006 (max. avg. NDVI) .........16

Figure 11 Histograms of NDVI relative frequency (0.05 unit bins) for 28th Sept. 2006 (End dry season) .......17

Figure 12 Histograms NDVI for wet (8th June; blue) and dry (28th Sept.; red) by vegetation community.......18

Figure 13 2008 NDVI – average values over the dry season by vegetation type ..........................................19

Figure 14 2008 Wetness Index – average values over the dry season by vegetation type............................20

Figure 15 Cumulative ET for each vegetation unit type for 2004 ................................................................22

Figure 16 Cumulative ET for each vegetation unit type for 2006. ...............................................................23

Figure 17 Cumulative ET for each vegetation unit type for 2008. ...............................................................24

Figure 18 Estimated groundwater evapotranspiration for each vegetation unit (2006). ................................25


Figure 1Map of vegetation communities used in the NDVI/Wetness/ET analysis

Floristic communities 2017Areas : veg_communCD-CSSSF1CD-OGHSR1D-HG1D-HG2IslandsLake bedOD-AFW1OD-EW1OD-OS1P-HG1RH-MWS1

NDVI 200605230.40.30.20.150.10.050-0.5

460000 m 470000 m 480000 m 490000 m 500000 m

460000 m 470000 m 480000 m 490000 m 500000 m

-262

0000

...-2

6100

00...

-260

0000

...-2

5900

00...

-258

0000

...-2

5700

00...

-256

0000

...-2

5500

00...

-262

0000

...-2

6100

00...

-260

0000

...-2

5900

00...

-258

0000

...-2

5700

00...

-256

0000

...-2

5500

00...


Table 4 Vegetation community descriptions

Landform Major Vegetation Group Floristic Community VegetationCode

Clos

ed D

epre

ssio

n Chenopod Shrublands,Samphire Shrublands and

Forblands (MVG22)

Heath of mixed Tecticornia spp. on salt lakeedge CD-CSSSF1

N/A Salt Lake CD-SL1

Other Grasslands, Herblands,Sedgelands and Rushlands

(MVG21)Open mixed herbs in clay-loam depression CD-OGHSR1

Dun

efie

ld

Casuarina Forests andWoodlands (MVG 8)

Low forest of Allocasuarina decaisneana overopen scrub of Acacia/ Grevillea and mid-densehummock grass of Triodia basedowii on sand

dunes/ swales

D-CFW1

Hummock Grasslands(MVG20)

Open low woodland of Corymbia opaca overlow scrub of Acacia/Grevillea spp. and mid-

dense hummock grass of Triodia basedowii onsand dunes/ swales

D-HG1

Scrub of Acacia/Eremophila/Grevillea spp. overmid-dense hummock grass of Triodia

basedowii on sand dunes/ swalesD-HG2

Ope

n D

epre

ssio

n

Acacia Forests andWoodlands (MVG 6)

Low woodland of Acacia spp. over low scrub ofSenna artemisioides and mixed dwarf scrub in

drainage depressionOD-AFW1

Eucalypt Woodland (MVG 5)Open low woodland of Eucalyptus

camaldulensis/ Corymbia spp. over mid-densehummock grass of Triodia spp. in creekline

OD-EW1

Other Shrublands (MVG 17)Low woodland of Hakea lorea/ Melaleucaglomerata over low heath of Fimbristylis

eremophila in drainage depressionOD-OS1

Plai

n Hummock Grasslands(MVG20)

Open low woodland of Corymbia spp./ Hakealorea over low scrub of Acacia spp. and mid-

dense hummock grass of Triodia spp. insandplain

P-HG1

Open shrub mallee of Eucalyptus gamophylla/E. kingsmillii subsp. kingsmillii over low scrubof Acacia bivenosa and mid-dense hummock

grass of Triodia basedowii in sandplain

P-HG2

Roc

ky H

illsl

ope Acacia Forests and

Woodlands (MVG 6)

Scrub of Acacia spp. over mixed low scrub andmid-dense hummock grass of Triodia pungens

on rocky hillslopeRH-AFW1

Mallee Woodlands andShrublands (MVG 14)

Open shrub mallee of Eucalyptus gamophylla/E. kingsmillii subsp. kingsmillii over low scrub

of Acacia/ Grevillea spp. and mid-densehummock grass of Triodia spp. on rocky

hillslope

RH-MWS1

Roc

kyPl

ain Acacia Forests and

Woodlands (MVG 6)

Low woodland of Acacia spp. over low scrub ofEremophila/ Senna spp. And mid-dense

hummock grass of Triodia basedowii on rockyplain

RP-AFW1


Landform Major Vegetation Group Floristic Community VegetationCode

Hummock Grasslands(MVG20)

Open low woodland of Corymbia aspera overlow scrub of Acacia spp. and mid-dense

hummock grass of Triodia basedowii on rockyplain

RP-HG1

While the present study has used methods consistent with Barron et al. (2012), it should be noted that theNDVI values returned between the two studies will differ due to vastly different vegetation communities andseasonal conditions. Figure 2 provides an example of NDVI maps for Lake Disappointment and the Barron et al.(2012) study areas for August 2004, showing considerably “greener” conditions for the Swan Coastal Plain areawhen compared to dry marginal landscape of Lake Disappointment.

Figure 2Comparison of Lake Disappointment and Barron et al. (2012) study area NDVI values

ResultsNDVI and NDWI

Barron et al. (2012) assessed the presence of potential GDEs by plotting changes in NDVI values over the dryseason, hypothesizing that those vegetation communities with the least change in greenness are most likely tobe supplemented by water sources other than rainfall (i.e. groundwater or perched surface water). Followingthis method, Figure 3 provides a plot of the late wet season NDVI values (x-axis) against the late dry seasonNDVI (y-axis) for each floristic community identified in the study area for each of three years (2004, 2006 and2008). Figure 5 shows the ‘end of dry season greenness index (NDVI)’ plotted against the ‘end of wet season

Lake Disappointment (5th A ug 2004) Swan coastal plain - Barron et al. study area (10th A ug 2004)

NDVI0.40.30.20.15

0.10.050-0.5


greenness index’. Vegetation units that deviate most from the 1:1 line are classified as ‘fast-drying vegetation’(Barron et al, 2012) and are very unlikely to be groundwater dependent. Vegetation units with relatively highand unvarying NDVI values, which closely follow the 1:1 plot line are inferred to have a continuing source ofwater (ie, are considered to be more likely to be groundwater dependent). Units with consistently low andunvarying NDVI may represent permanent water or wetland surfaces (if they also show high and unvaryingNDWI signatures and high ET) or may correspond to sparse vegetation or bare soil (if they have lower NDWIand low cumulative ET).

This method has identified CD-CSSSF1 (Tecticornia spp. on salt lake edge) as having the least variable NDVIvalues across the dry season, followed by CD-OGHSR1 (open mixed herbs in clay-loam depression. Theserankings were relatively consistent across years (Error! Reference source not found.). There is a distinctpossibility that the sparseness of vegetation, particularly in the CD-CSSSF1 community, is lowering the NDVIresponse over the dry season. The NDVI pixels are an average of 30 m × 30 m, which includes any bareground between plants. The lake bed signature showed the least NDVI variability as it represented bare groundwith a complete lack of vegetation (Figure 3). Sparse vegetation would comprise a greater degree of non-variable substrate in the form of bare ground and/or dead litter material. Therefore the Barron et al. (2012)method is likely to require greater botanic interpretation when applied to the Lake Disappointment study area.

Figure 3Ratio of NDVI values from late wet to end of dry season

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

0.20

0.25

-0.15 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20 0.25

End

of d

ry N

DVI

Max. late wet NDVI

Ratio of NDVI by Floristic Community

CD-CSSSF1

CD-OGHSR1

D-HG1

D-HG2

Islands

Lake bed

OD-AFW1

OD-EW1

OD-OS1

P-HG1

RH-MWS1

Average image

one to one

Linear (one to one)Lake bed


Figure 4Ratio of NDWI values from late wet to end of dry season

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

-0.4 -0.35 -0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1

End

of d

ry N

DWI

Max. late wet NDWI

Ratio of NDWI by Floristic Community

CD-CSSSF1

CD-OGHSR1

D-HG1

D-HG2

Islands

Lake bed

OD-AFW1

OD-EW1

OD-OS1

P-HG1

RH-MWS1

one to one

Linear (one to one)

Lake bed


Figure 5Ratio of NDVI values from late wet to end of dry season– distance from 1:1 (no change)

00.010.020.030.040.050.060.070.080.09

0.1

CD-CSSSF1 CD-OGHSR1 D-HG1 OD-OS1 RH-MWS1 D-HG2 P-HG1 Islands OD-EW1 OD-AFW1

Dist

ance

(rel

ativ

e uni

ts)

Floristic community

2004

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

CD-CSSSF1 CD-OGHSR1 D-HG1 P-HG1 Islands RH-MWS1 D-HG2 OD-OS1 OD-EW1 OD-AFW1

Dist

ance

(rel

ativ

e uni

ts)

Floristic community

2006

00.005

0.010.015

0.020.025

0.030.035

0.040.045

CD-CSSSF1 Islands CD-OGHSR1 D-HG1 OD-EW1 OD-AFW1 P-HG1 RH-MWS1 D-HG2 OD-OS1

Dist

ance

(rel

ativ

e uni

ts)

Floristic community

2008


Figure 62004 NDVI – average values over the dry season by vegetation type

ND

VI


Figure 72004 Wetness Index – average values over the dry season by vegetation type

ND

WI(

Wet

ness

Inde

x)


Figure 82006 NDVI – average values over the dry season by vegetation type

ND

VI


Figure 92006 Wetness Index – average values over the dry season by vegetation type

ND

WI(

Wet

ness

Inde

x)


Figure 10 Histograms of NDVI relative frequency (0.05 unit bins) for 8th June 2006 (max. avg. NDVI)

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

CD-CSSSF1

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

CD-OGHSR1

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

D-HG1

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

D-HG2

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

Islands

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

Lake bed

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

OD-AFW1

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

OD-EW1

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

OD-OS1

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

P-HG1

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

RH-MWS1

8th June 2006 (max. NDVI)Histograms by vegetation community


Figure 11 Histograms of NDVI relative frequency (0.05 unit bins) for 28th Sept. 2006 (End dry season)

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

CD-CSSSF1

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

CD-OGHSR1

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

D-HG1

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

D-HG2

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

Islands

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

Lake bed

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

OD-AFW1

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

OD-EW1

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

OD-OS1

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

P-HG1

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5NDVI (0.05 unit bins)

00.20.40.60.8

RH-MWS1

28th September 2006 (End dry season)Histograms by vegetation community


Figure 12 Histograms NDVI for wet (8th June; blue) and dry (28th Sept.; red) by vegetation community


Figure 13 2008 NDVI – average values over the dry season by vegetation type

Feb-08 Mar-08 Apr-08 May-08 Jun-08 Jul-08 Jul-08 Aug-08 Sep-08 Oct-08 Nov-08Date

-0.12

-0.08

-0.04

0

0.04

0.08

ND

VI

CD-CSSSF1CD-OGHSR1D-HG1

D-HG2IslandsLake bed

OD-AFW1OD-EW1OD-OS1

P-HG1RH-MWS1

Average NDVI by Vegetation Community - 2008


Figure 14 2008 Wetness Index – average values over the dry season by vegetation type

ND

WI(

Wet

ness

Inde

x)


Evapotranspiration

Cumulative ET

Estimates of actual evapotranspiration (ET) were calculated for the study area using satellite imagery from theCMRSET data set. Cumulative ET was plotted for each year (2004, 2006 and 2008) to see how ET differed overa year for each of the vegetation types. According to O’Grady et al. (2011) GDE-related vegetation classes arelikely to have greater ET losses over the dry period than non-GDE-related classes. The results, shown in Figure15, Figure 16 and Figure 17 showed that the highest estimated losses to evaporation are associated with thelake bed and island, which you would expect because there is water associated with these units in terms ofopen water or moist soil for a much greater proportion of the dry season than other vegetation units, so ETrates are expected to be higher. It is important to note that the rainfall values used in these figures are forTelfer, which is approximately 175km North West of Lake Disappointment. Rainfall can be highly variable acrossthis area, so this data may not be a truly representative of what is happening and any conclusion drawn fromthis data should be done so with this caveat in mind.

There was one vegetation type that displayed consistently high ET rates (CD CSSSF1) which is the ‘heath ofmixed Tecticornia spp. on salt lake edge’. This result was consistent across all three years (2004, 2006 and2008). This may be indicative of groundwater dependence. However, it is important to note that evaporationfrom groundwater is dependent on the water table depth and hence ET is expected to be greater where thewater table is shallower. O’Grady et al. (2009) showed that more groundwater is taken up by vegetation that islocated where the water table is shallower. The vegetation class CD CSSSF1 sits on the edge of the salt lakewhere the water table is expected to be at its shallowest and this may explain why ET rates are higher.

It is also important to remember the limited spatial resolution of this data which allows for a pixel size of250m2. The typical width of the CD-CSSSF1 is in the order of 100m to 3000m and is extremely patchy;consequently it is highly likely that ET values will be overestimated because lake bed components will beinadvertently included in the ET calculations for that pixel. Consequently caution needs to be taken whenmaking assumptions that this vegetation unit is groundwater dependant based solely on estimated ET. Todetermine conclusively that the class is a GDE further botanic interpretation should be applied to the studyarea.


Figure 15 Cumulative ET for each vegetation unit type for 2004

0

200

400

600

800

1000

1200

1400

Cum

ulat

ive

ET (m

m)

Cumulative Evapotranspiration for 2004

OD-OS1

RH-MWS1

OD-AFW1

CD-OGHSR1

CD-CSSSF1

P-HG1

D-HG1

D-HG2

OD-EW1

Lake bed

Islands

Rainfall (Telfer)


Figure 16 Cumulative ET for each vegetation unit type for 2006.

0

200

400

600

800

1000

1200

1400

Cum

ulat

ive

ET (m

m)


OD-OS1

RH-MWS1

OD-AFW1

CD-OGHSR1

CD-CSSSF1

P-HG1

D-HG1

D-HG2

OD-EW1

Lake bed

Islands

Rainfall (Telfer)


Figure 17 Cumulative ET for each vegetation unit type for 2008.

0

200

400

600

800

1000

1200

Cum

ulat

ive

ET (m

m)


OD-OS1

RH-MWS1

OD-AFW1

CD-OGHSR1

CD-CSSSF1

P-HG1

D-HG1

D-HG2

OD-EW1

Lake bed

Islands

Rainfall (Telfer)


Groundwater Evapotranspiration

The results of the groundwater evapotranspiration assessment are shown in Figure 18. No results are providedfor the lake surface, for vegetation on islands on the lake or for vegetation units occurring mainly within 250mof the playa edge, as recent work by van Dijk et al (2015) has shown that the CMRSET method is unreliable forsalt lake systems and is known to overestimate evapotranspiration.

An extension of this method was devised recently by Doody et al. 2017 in which they calculated the probabilityof a vegetation unit using groundwater during dry seasons, which was derived from a ratio of ET to rainfall andis referred to as the potential inflow dependent landscape (pIDE). Table 5 shows the ratio and probability ofinflow dependence of each vegetation unit (with the exception of salt lake units, for the reasons given above).These results indicate that there is low likelihood of pronounced groundwater dependency in the vegetationunits listed in the table.

Figure 18 Estimated groundwater evapotranspiration for each vegetation unit (2006).

-160

-140

-120

-100

-80

-60

-40

-20

0OD-OS1 CD-OGHSR1 P-HG1 RH-MWS1 D-HG1 D-HG2 OD-EW1 OD-AFW1

Groundwater Evapotranspiration for 2006


Table 5 The probability of inflow dependence for each vegetation unit for 2006

Floristic Community ET/Rainfall Ratio pIDE (%)

OD-OS1 0.59 0%

CD-OGHSR1 0.53 0%

P-HG1 0.62 5%

RH-MWS1 0.68 6%

D-HG1 0.63 5%

D-HG2 0.65 6%

OD-EW1 0.83 10%

OD-AFW1 0.92 20%

Summary and Conclusions

Spectral data were analysed for three separate years at Lake Disappointment.

No vegetation unit showed consistently high and unvarying NDVI and NDWI indices (the spectralsignature typically associated with groundwater dependent vegetation).

One vegetation unit (OD-CSSSF1) showed low, but relatively constant NDVI values and moderate, butvariable, NDWI values (but with lower and less variable wetness than the playa surface). Typically, thissignature would indicate areas of sparse vegetation or bare soil.

There are methodological issues that limit the application of ET estimation on salt lakes. Theselimitations constrained the use of ET methods in estimating the likelihood of groundwater dependenceof vegetation on islands or in close proximity to the playa.

For vegetation units not closely associated with the salt lake, estimated evapotranspiration andgroundwater evaporation amounts did not exceed rainfall, further supporting the conclusion that theseunits are unlikely to be strongly groundwater dependent.

References

Barron, O. E., Emelyanova, I., Van Niel, T.G., Pollock, D. and Hodgson, G. (2012). Mapping groundwater-dependent ecosystems using remote sensing measures of vegetation and moisture dynamics. Hydrol. Process.(2012).

Doody, T.M., Barron, O.V., Kate Dowsley, K., Emelyanova,I., Fawcett,J., Overton,I.C., Pritchard,J.L. AlbertI.J.M. Van Dijk,V. and Warren, G. (2017). Continental mapping of groundwater dependent ecosystems: Amethodological framework to integrate diverse data and expert opinion. Journal of Hydrology: RegionalStudies 10: 61-81.

Eamus D., Zolfaghar S., Villalobos-Vega R., Cleverly J. and Huete A. (2015). Groundwater-dependentecosystems: recent insights from satellite and field-based studies. Hydrology and Earth System Science, 19,4229-4256.


Gonzalez, D. 2015, ‘Preliminary analysis of remotely sensed actual evapotranspiration data,’ in Woods, J (Eds),Modelling salt dynamics on the River Murray in South Australia: Conceptual model, data review and salinity riskapproaches. Goyder Institute for Water Research Technical Report Series No. 15/x.

Gu, Y., Brown, J.F., Verdin, J. and Wardlow, B. (2007). Five-year analysis of MODIS NDVI and NDWI forgrassland drought assessment over the central Great Plains of the United States. Geophysical Research Letters,VOL. 34, 2007.

Guerschman J.P., van Dijk, A.I.J.M., Mattersdorf, G., Beringer, J., Hutley, L.B., Leuning, R., Pipunic, R.C. andSherman, B.S. (2009). Scaling of potential evapotranspiration with MODIS data reproduces flux observationsand catchment water balance observations across Australia, Journal of Hydrology 369: 107-119.

O’Grady, A., Carter, J. and Holland, K. (2009). Review of Australian Groundwater Discharge Studies ofTerrestrial Systems. CSIRO: Water for a Healthy Country National Research Flagship, Australia.

O’Grady A.P., Carter J.L. and Bruce J. (2011). Can we predict groundwater discharge from terrestrialecosystems using eco-hydrological principals? Hydrology and Earth System Science Discussions 8: 8231–8253.

van Dijk, A.I.J.M., Warren, G., Van Neil, T., Byrne, G., Pollock, D. and Doody, T.M. (2015). Derivation of datalayers for medium resolution remote sensing to support mapping of groundwater dependent ecosystems.CSIRO Land and Water https://www.researchgate.net/profile/Tanya Doody/publications.

Wood C., Plush B. and Riches V. (2016). Evapotranspiration Study of Pike Floodplain Using CMRSET Data,DEWNR Technical note 2016/10, Government of South Australia, through the Department of Environment,Water and Natural Resources, Adelaide.

MEMORANDUM - epa.wa.gov.au D6... · Lake DisapPointment: NDVI, NDWI AND ET calculations p3...

Documents

Transcript of MEMORANDUM - epa.wa.gov.au D6... · Lake DisapPointment: NDVI, NDWI AND ET calculations p3...