Conference PaperPDF Available

FRACTIONAL GROUND COVER MONITORING OF PASTURES AND AGRICULTURAL AREAS IN QUEENSLAND

September 2010

Conference: Proceedings of the 15th Australasian Remote Sensing and Photogrammetry Conference

Authors:

Michael Schmidt

German Aerospace Center (DLR)

Peter Scarth

University of Queensland

Vegetation cover is a critical attribute of the landscape, affecting infiltration, runoff, water erosion and wind erosion. Ground cover relates to living and non-living materials on the soil surface of the lower strata (<2m). Besides the direct economical value for graziers, varying levels of ground cover have indirect economical and major ecological implications on productivity, land condition and biodiversity. Remote sensing offers one of the few cost effective approaches to monitoring long term ground cover change over large spatial extents. In Queensland's dryland ecosystems, changes in ground cover are generally related to management practices and climate influences. The spatial resolution of Landsat TM and ETM+ (Thematic Mapper and Enhanced TM Plus) imagery provides relevant information to monitor ground cover and to inform natural resource management decision makers. Three ground cover prediction algorithms are evaluated based on 692 field sites: (i) a regression based bare ground prediction, (ii) multiple logistic regression algorithms, and (iii) a spectral unmixing approach. The estimates of (ii) and (iii) include fractions of bare ground, green vegetation and senescent vegetation. The three models estimate bare ground with a root mean square error of 16.9%, 14.6% and 17.4% and a median average error of 10.1, 10.6 and 11.5, respectively. Although the algorithms were trained in pasture field sites across the State of Queensland, the three approaches show correlations with ground cover field observations in an agricultural area of r 2 =0.40, r 2 =0.68 and r 2 =0.56, respectively. The utility of ground cover production models across pastoral and agricultural land uses requires further consideration for accuracy and operational efficiencies. Introduction Dynamic ground cover information is important for soil erosion and nutrient flux estimates into the stream network. These are of particular interest in catchments adjacent to the Great Barrier Reef (Karfs et al. 2009, Queensland Government 2009). Episodic interactions of low ground cover with heavy rainfall, often arising after overgrazing and during periods of droughts or low rainfall, have led to eventual catchment degradation in many Australian

Comparison for three different ground cover models with field observations. Line 1 compares the ground cover estimate of all three models, while in lines 2 and 3 the GV and NGV predictions of Model 2 and 3 are shown. The solid line is a 1:1 line, the dotted line a regression.

…

Comparison of three different ground cover prediction models and field observations as box and whisker plots of deciles. In row 1 the ground cover estimates of all three models are compared, while rows 2 and 3 show the GV and NGV

…

Spatial representation of fieldwork in May 2009 on an agricultural site near Goondiwindi: a) Landsat true colour composite showing field site and field photo locations; b) Model 1-Ground Cover Index (GCI) as described in Scarth et al. (2006); c) Model 2-fractional cover descriptions of bare, GV and NGV as described in Schmidt, Denham & Scarth, (in prep); and d) Model 3-fractional cover descriptions of bare, GV and NGV as described in Scarth et al. (in prep.). Landsat TM image observation date: 2009/04/30.

…

Spatial representation of fieldwork in October 2009 on an agricultural site near Goondiwindi: a) Landsat true colour composite, indicating field site and field photo locations; b) Model 1-Ground Cover Index (GCI) as described in Scarth et al. (2006); c) Model 2-fractional cover descriptions of bare, GV and NGV as described in Schmidt, Denham & Scarth (in prep); and d) Model 2-fractional cover descriptions of bare, GV and NGV as described in Scarth et al. (in prep.). Landsat TM image observation date: 2009/10/23.

…

Figures - uploaded by Michael Schmidt

Content may be subject to copyright.

Content uploaded by Michael Schmidt

Content may be subject to copyright.

FRACTIONAL GROUND COVER MONITORING OF PASTURES

AND AGRICULTURAL AREAS IN QUEENSLAND

Michael Schmidt, Robert Denham and Peter Scarth

Department of Environment and Resource Management

Climate Building

80 Meiers Rd

Indooroopilly QLD 4068

Phone: 07 3896 9297 Facsimile 07 3896 9843

michael.schmidt@nrw.qld.gov.au

Abstract

Vegetation cover is a critical attribute of the landscape, affecting infiltration,

runoff, water erosion and wind erosion. Ground cover relates to living and non-

living materials on the soil surface of the lower strata (<2m). Besides the direct

economical value for graziers, varying levels of ground cover have indirect

economical and major ecological implications on productivity, land condition and

biodiversity. Remote sensing offers one of the few cost effective approaches to

monitoring long term ground cover change over large spatial extents. In

Queensland’s dryland ecosystems, changes in ground cover are generally

related to management practices and climate influences. The spatial resolution

of Landsat TM and ETM+ (Thematic Mapper and Enhanced TM Plus) imagery

provides relevant information to monitor ground cover and to inform natural

resource management decision makers.

Three ground cover prediction algorithms are evaluated based on 692 field

sites: (i) a regression based bare ground prediction, (ii) multiple logistic

regression algorithms, and (iii) a spectral unmixing approach. The estimates of

(ii) and (iii) include fractions of bare ground, green vegetation and senescent

vegetation. The three models estimate bare ground with a root mean square

error of 16.9%, 14.6% and 17.4% and a median average error of 10.1, 10.6 and

11.5, respectively. Although the algorithms were trained in pasture field sites

across the State of Queensland, the three approaches show correlations with

ground cover field observations in an agricultural area of r2=0.40, r2=0.68 and

r2=0.56, respectively. The utility of ground cover production models across

pastoral and agricultural land uses requires further consideration for accuracy

and operational efficiencies.

Introduction

Dynamic ground cover information is important for soil erosion and nutrient flux

estimates into the stream network. These are of particular interest in

catchments adjacent to the Great Barrier Reef (Karfs et al. 2009, Queensland

Government 2009). Episodic interactions of low ground cover with heavy

rainfall, often arising after overgrazing and during periods of droughts or low

rainfall, have led to eventual catchment degradation in many Australian

rangelands (Bastin et al. 2009). Ground cover levels may vary due to

anthropogenic management of grazing enterprises and agricultural land

management practices, or natural changes in seasonal rainfall. Many extensive

grazing areas and rangelands in the reef catchments are subject to high climate

variability on seasonal, annual, decadal and longer timescales, making

management for economic and environmental sustainability difficult (Ludwig et

al. 2007).

A recent report (Leys et al. 2009) identified ground cover as a key indicator of

land management practices. At both the national and regional levels, there

remains a lack of comprehensive and consistent ground cover data at a

temporal and spatial scale adequate for monitoring and assessing

environmental targets related to soil erosion and land management.

To date, the Department of Environment and Resource Management (DERM)

ground cover monitoring program has reported annually on the percentage of

ground cover in Queensland based on Landsat imagery using a model

described in Scarth et al. (2006). Recent research has resulted in an additional

two improved ground cover models. One is based on a linear spectral unmixing

approach (Scarth et al. in prep.) and the other is based on a multinominal

regression (Schmidt, Denham & Scarth in prep). Both models predict three

fractions of surface cover for the ground stratum: bare ground (bare), green

vegetation (GV) and dry (or non-green) vegetation (NGV).

A comparison between these three ground cover models is described here

using historic field observations in rangelands and field data from a pilot study in

an agricultural area.

Data

Remote sensing data

Landsat TM and ETM+ imagery were historically obtained once per annum for

Queensland from ACRES (Australian Centre for Remote Sensing). Additional

imagery was acquired to coincide with field observations. A large archive of

more than 2500 Landsat images is held by DERM (Trevithick & Gillingham

2010).

All freely available (largely cloud free) Landsat TM and Landsat ETM+ imagery

from the United States Geologic Survey (USGS) EarthExplorer website

(http://edcsns17.cr.usgs.gov/EarthExplorer/) covering Queensland have recently been

downloaded and added to the DERM image archive. This includes more than

15,000 Landsat images and means that on average about 20 image dates per

year from 1999 to 2010 are available for each scene (see Figure 1). This

includes Landsat ETM+ with scan line corrector “off” imagery

(http://landsat.usgs.gov/products_slcoffbackground.php).

The Landsat TM (Thematic Mapper) and ETM+ (Enhanced TM Plus) images (of

both sources) were stored in a raw stage, but the higher level products have

been corrected using methods developed and implemented by the SLATS

project (http://www.derm.qld.gov.au/slats/). All images were pre-processed and

subjected to rigorous radiometric corrections as per standard DERM remote

sensing methods (de Vries et al. 2007) and were standardised for surface

reflectance and atmospheric effects. Standard masks for cloud, cloud shadow

and water were applied to the ACRES imagery, while preliminary cloud, cloud

shadow and water body masking (Muir, Danaher 2008) processes were

implemented for USGS data (Schmidt & Trevithick 2010).

Field data

Ground cover field data were collected in across a broad range of land types in

Queensland over more than a decade. A transect point-intercept method was

used in grasslands and improved pastures with the same field methodology.

Sample sites were chosen to be homogeneous and to represent a 100m x

100m area. Descriptions of general site details are performed according to the

method described by Tongway and Hindley (1995).

All field data were collected using a modified discrete point sampling method of

300 individual measurements at each metre along three 100m tapes. For this

study, the site measurements were summarised in percentages of the ground

cover components: bare ground (including rock and cryptogam), green

vegetation and dry vegetation (including litter and dead plant material). In the

midstorey (woody plants <2m) and overstorey (woody plants >2m) stratas,

components of dry/dead leaf, branch and green leaf were recorded also. For a

detailed description of the field methodology see Schmidt, Muir and Scarth

(Draft). For ground cover studies we selected field data which was collected

from sites with low Foliage Projective Cover (FPC) in the overstorey, typically

less than 15%.

Figure 1 gives an overview of the location and spatial distribution of field data

sites, and the USGS data coverage for Queensland according to the World

Reference System II path/row for Landsat imagery

(http://landsathandbook.gsfc.nasa.gov/handbook.html).

(

((!

(

((

(

92/

97/

80 89/

90/

91/

94/

95/

96/

80 93/

89/

91/

94/

79 93/

95/

96/

79 92/

79 90/

97/

90/

78 89/

91/

99/

78 95/

78 92/

94/

78 93/

97/

98/

78 96/

92/

94/

97/

77 95/

96/98/

77 93/

77 91/

99/

77 89/

90/

91/

93/

76 92/

94/

95/

76 90/

99/

76 97/

76 96/

98/

95/

97/

99/

75 96/

98/

75 93/

94/

75 92/

75 91/

98/

99/

74 97/

74 94/

95/

96/

74 93/

74 92/

99/

73 98/

73 94/

97/

73 96/

73 95/

99/

72 97/

98/

72 95/

96/

97/

71 96/

98/

99/

70 97/

98/

70 96/

97/

99/

69 98/

98/

99/

98/

99/

100/

0 250 500125

Kilometers

Field Sites

USGS Scene Boundaries

Figure 1: Location of ground cover field observations in Queensland and USGS data

coverage in path/row; some sites were repeatedly visited.

Three field work campaigns were undertaken at different crop growth stages in

an agricultural area near Goondiwindi (data point in scene 91/80) in 2009 with a

modified sampling strategy of two 100m tapes in a 45-degree angled tape

layout across linearly sown crops, as suggested by Schmidt et al. (2010).

Methods

Ground cover predictions of three algorithms were compared independently

against field observations. For each of the field observations the Landsat image

observation with the closest date was used, with a time difference of no more

than 60 days. A brief description of the three methods and previously derived

error statements follows.

(i) Model 1 is based on a linear regression approach as described by Scarth et

al. (2006). The approach estimates the amount of vegetative ground cover as

Ground Cover Index (GCI) as a percentage, or its inverse bare ground as Bare

Ground Index (BGI) using data from 431 field sites in Queensland with a

reported RMSE of 13.8%;

(ii) Model 2 uses a multinominal logistic regression as described in Schmidt,

Denham & Scarth (in prep). Three fractions are predicted as percentages: bare

ground, GV and NGV – which sum up to 100%. 512 field sites in Queensland

were used with a reported RMSE of 14.9 % in this approach;

(iii) Model 3 estimated ground cover fractions via a (linear) spectral mixture

analysis (SMA) as described by Scarth et al. (in prep.). Three fractions of

ground cover are predicted as: bare ground, GV and NGV plus an error term of

the SMA, summing up to 100%. This approach used 577 field sites in

Queensland and additional 83 sites in New South Wales with a reported RMSE

of 11.8%.

For this analysis the root mean squared error (RMSE) of the observed and

predicted values was reported as well as the median average error (MAE) as a

more robust (less sensitive to outliers) description. This was performed on 692

field sites which are now available for grasslands and improved pastures in

Queensland. Data from more than 110 additional site data, which were not part

of any model calibration, were used for validation. All available field data were

used without the removal of outliers or bad data.

A second analysis step was performed in an agricultural area based on 25 field

observations during three field campaigns.

All three ground cover models were developed on the basis of Landsat data

from the ACRES which have a slightly different data calibration compared to the

USGS data. A data inter-comparison between the two data sources for five

different land-use types was performed for Landsat bands 3, 5 and 7 which

were used in Model 1.

Results

Rangeland sites

The first row in Figure 2 shows scatter-plots of field observations and bare

ground predictions of the three ground cover models. The plots show regression

lines as well as the RMSE and MAE per model. Only data from the ACRES

archive were used here for all three model comparisons.

0.0 0.2 0.4 0.6 0.8 1.0

Figure 2: Comparison for three different ground cover models with field observations.

Line 1 compares the ground cover estimate of all three models, while in lines 2 and 3

the GV and NGV predictions of Model 2 and 3 are shown. The solid line is a 1:1 line,

the dotted line a regression.

The error statistics as well as the data scattering around the plotted regression

line show that all three models give useful estimates of bare ground with MAE

of 10.1% in Model 1, 10.6% in Model 2 and 11.5% in Model 3. The RMSE of

14.6% is lowest in Model 2, with less data scatter occurring. The regression line

in Model 2 is also very close to the 1:1 of field observations and predicted bare

ground, while the regression lines in Model 1 and Model 3 have a slightly

different slope and bias.

predicted

observed

0.0

0.2

0.4

0.6

0.8

1.0

MAE=0.101

RMSE=0.169

Model I (linear)

Bare

MAE=0.106

RMSE=0.146

Mod el II (multinomia l)

Bare

MAE=0.115

RMSE=0.174

Mo del III (SM A)

Bare

Model I

(

linear

)

MAE=0.043

RMSE=0.121

Model II

(

multinomial

)

0.0

0.2

0.4

0.6

0.8

1.0

MAE=0.05

RMSE=0.123

Mo del III

(

SMA

)

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.2 0.4 0.6 0.8 1.0

Model I

(

linear

)

NGV

MAE=0.117

RMSE=0.173

Model II

(

multinomial

)

NGV

0.0 0.2 0.4 0.6 0.8 1.0

MAE=0.135

RMSE=0.209

Mo del III

(

SMA

)

NGV

Row 2 shows the observations and predictions for GV, both Model 1 and 2 have

similar errors with an RMSE of 12.1% and 12.3% RMSE respectively and 4.3%

and 5% for the MAE. Model 2 seems to have a slightly greater deviation from

the 1:1 line. The predictions for NGV seem to work better in Model 2 than in

Model 3 with an RMSE of 17.3% compared to 20.9 % in Model 3, also the MAE

(11.7%) is lower in Model 2 than in Model 3 (13.5%). The point cloud seems to

represent the data in Model 2 well with a regression very close to the 1:1 line,

whereas there is a different slope and bias for the predictions of Model 3. Figure

3 shows the residuals of Figure 2 summarised in deciles.

Figure 3: Comparison of three different ground cover prediction models and field

observations as box and whisker plots of deciles. In row 1 the ground cover estimates

of all three models are compared, while rows 2 and 3 show the GV and NGV

Decile of prediction

-0.5

0.0

0.5

Model I

(

linear

)

Bare Mod el II

(

multinomial

)

Bare Mode l III

(

SMA

)

Bare

Model I

(

linear

)

GV Mo del II

(

multinomial

)

GV Mo del III

(

-0.5

0.0

0.5

SMA

)

-0.5

0.0

0.5

12345678910

Model I

(

linear

)

NGV

12345678910

Mo del II

(

multinomial

)

NGV Mo del III

(

SMA

12345678910

)

NGV

predictions of Model 2 and 3 are shown (the doted lines are plotted at the 0.1 and 0.25

values)

Figure 3 shows box and whisker plots of the three model predictions. The

residuals of the bare ground predictions in Model 1 and 2 are very evenly

distributed and close to zero, while Model 3 shows some systematic deviation

from zero with increasing standard deviation in the deciles 4 to 7 with the

highest deviations. The range of the residuals appears lowest in Model 2. Some

data outliers are plotted, which might be due to artefacts of cloud and cloud

shadow masking.

The GV estimation seems to show a systematic behaviour of the residuals per

decile in Model 2 and to a lesser extent in Model 3. The residuals per decile for

NGV in both Model 2 and 3 do not appear to have a systematic behaviour, with

a lower range and standard deviation in Model 2.

Agricultural areas

An ACRES and USGS Landsat data comparison for different land cover types

for the Landsat spectral band used for the BGI in model 1 are shown in Table 1.

Table 1: Differences in BGI for ACRES and USGS data in Landsat bands 3, 5 and 7

top of the atmosphere reflectance (used in Model 1), as well as the BGI.

Description b3 b5 b7 BGI b3 b5 b7 BGI Difference

openpasture 69 149 86 179 66 145 87 183 4

pasture(gold) 31 117 39 101 28 113 37 101 0

pasture2(adav) 69 148 83 173 67 143 82 175 2

openwoodland 42 104 55 122 39 100 56 123 1

openwoodland2 76 143 56 111 72 137 56 116 5

grass/riverbank 73 142 84 186 74 146 88 189 3

shrubs/riverbed 33 118 52 103 31 114 51 102 1

agriculture 30 56 34 186 29 56 33 179 7

claypan 103 193 108 194 97 187 108 196 2

ACRES USGS

Pasture

Open woodland

Riverbed (grassy)

Shrubs

Claypan

Agricultural field

BGI

[%]

The differences between the two Landsat data sources used for BGI calculation

in Model 1 were 5% or less in natural environments and had a maximum

difference of 7% in an agricultural site. The differences in BGI were below the

model accuracy (Scarth et al. 2006) and the data was thus deemed adequate

for further model comparison. For this analysis only cloud free USGS data were

utilised.

Field work data from three campaigns near Goondiwindi (QLD) were collated at

three different growth stages of a wheat field and a fallow field of Sorghum and

a fallow wheat field. The utilised wheat field (Site 2 in Figures 4 to 6) was visited

firstly after ploughing (May 2009), secondly during greening up (August 2009)

and thirdly after harvesting (October 2009).

Figure 4 shows a spatial representation of the utilised fields of the property

‘Monte Christo’ in a true colour Landsat TM 5 image, the three ground cover

models and field photos. Some of the field sites from May 2009 (near photo 4)

where ‘short’ transects (10m) for test purposes and excluded from this analysis

(Schmidt et al. 2010). So that 13 full transect field sites were used in the

following, including two data points from the site near photo point 1 in Figures 4

and 5 from an improved pasture site.

Figure 4: Spatial representation of fieldwork in May 2009 on an agricultural site near

Goondiwindi: a) Landsat true colour composite showing field site and field photo

locations; b) Model 1 - Ground Cover Index (GCI) as described in Scarth et al. (2006);

c) Model 2 - fractional cover descriptions of bare, GV and NGV as described in

Schmidt, Denham & Scarth, (in prep); and d) Model 3 - fractional cover descriptions of

bare, GV and NGV as described in Scarth et al. (in prep.). Landsat TM image

observation date: 2009/04/30.

Figure 4 shows a better differentiation within and across field in Models 2 and 3

than in Model 1. Models 2 and 3 estimate noticeably different greenness

intensities in one of the fields and also in the riparian vegetation. The freshly

ploughed field in photo 2, with 94% observed bareness is predicted as being

62% bare by Model 3, 70% by Model 1 and 76% by Model 2.

Figures 5 and 6 show the same area at a different crop growth stage and time

of year.

Figure 5: Spatial representation of fieldwork in August 2009, on an agricultural site

near Goondiwindi: a) Landsat true colour composite, indicating field site and field photo

locations; b) Model 1 - Ground Cover Index (GCI) as described in Scarth et al. (2006);

c) Model 2 - fractional cover descriptions of bare, GV and NGV as described in

Schmidt, Denham & Scarth (in prep); and d) Model 3 - fractional cover descriptions of

bare, GV and NGV as described in Scarth et al. (in prep.). Landsat TM image

observation date: 2009/08/04.

Some of the fields in Figure 5 show active cropping (near Photo 2), while other

fields (near Photos 3 and 4) remained unutilised with very little change in

ground cover.

Figure 6 shows the same area after harvesting. It is noticeable that the

harvested wheat stubble (near Photo 2) shows a different cover than the older

wheat stubble (near Photo 4). Observations: 65.5% cover (65% NGV plus 0.5%

GV), – Model 1: 60% cover, Model 2: 63% cover (56% NGV plus 7% GV),

Model 3: 72% cover (65% NGV plus 8% GV).

Figure 6: Spatial representation of fieldwork in October 2009 on an agricultural site

near Goondiwindi: a) Landsat true colour composite, indicating field site and field photo

locations; b) Model 1 - Ground Cover Index (GCI) as described in Scarth et al. (2006);

c) Model 2 - fractional cover descriptions of bare, GV and NGV as described in

Schmidt, Denham & Scarth (in prep); and d) Model 2 - fractional cover descriptions of

bare, GV and NGV as described in Scarth et al. (in prep.). Landsat TM image

observation date: 2009/10/23.

A regression plot of all 13 observed field sites for the agricultural area is shown

in Figure 7. The three cover predictions of bare, GV and NGV are plotted

separately for Models 1, 2 and 3.

R2 = 0.2483

R2 = 0.4511

R2 = 0.3916

100

120

140

160

180

200

0 20406080100

observed

predicted

Model 1

Model 2

Model 3

Linear (Model 1)

Linear (Model 2)

Linear (Model 3)

100

= 0.4848

= 0.5582

100

120

140

160

180

200

0 20406080100

observed

predicted

=0.24

=0.39

bare

=0.45

= 0.7942

= 0.6979

100

120

140

160

180

200

0 20406080100

observed

predicted

Model 2

Model 3

Linear (Model 2)

Linear (Model 3)

100

=0.79

=0.70

NGV

= 0.7942

= 0.6979

100

120

140

160

180

200

0 20406080100

observed

predicted

Model 2

Model 3

Linear (Model 2)

Linear (Model 3)

100

=0.79

=0.70

NGV

= 0.9605

= 0.9297

100

120

140

160

180

200

020406080100

observed

predicted

Model 2

Model 3

Linear (Mod el 2)

Linear (Mod el 3)

100

=0.96

=0.93

= 0.9605

= 0.9297

100

120

140

160

180

200

020406080100

observed

predicted

Model 2

Model 3

Linear (Mod el 2)

Linear (Mod el 3)

100

=0.96

=0.93

= 0.6817

Model 1

Model 2

Model 3

Linear (Model 1)

Linear (Model 2)

Linear (Model 3)

100100

8080

6060

4040

2020

bare

=0.49

=0.56

a) b)

c) d)

=0.68

Figure 7: Model comparison for the agricultural field sites near Goondiwindi for three

different ground cover prediction models for the fractions bare (a), NGV (c) and GV (d);

(b) shows the bare fractions of all models without the two sites with entirely green

vegetation coverage.

The bare component shows relatively low R2 values in all three models: 0.24 in

Model 1, 0.45 in Model 2 and 0.39 in Model 3. This might be because the

training sites for all three models were chosen in pastoral environments. In fact,

two field sites were measured during the greening up phase in August 2009

with no NGV components, which is very different from natural or pastoral

environments. The observed cover components were entirely composed of a

lush green wheat crop, growing in a black, ploughed soil. Not only has it been

proven difficult in estimating cover in black soils, but also the greenness is

outside the range of our training data. When these two sites were taken out of

the field observations (Figure 7 b) the R2 values for observed and predicted

data increased to 0.49, 0.68 and 0.56 for Models 1, 2 and 3 respectively.

The plot with the NGV data (Figure 7 c) shows a much better fit with R2 values

of 0.79 and 0.70 for Model 2 and 3, respectively, which both seem to appear

with an offset value in the linear regression.

The R2 values of the GV components in Models 2 and 3 (Figure 7 d) are with

0.96 and 0.93 high although a clustering of low data points limits the

interpretability of this plot.

Discussion and conclusions

In grasslands and improved pastures all three models seem to predict bare

ground (or its inverse ground cover) well, with a MAE of 10.1 in Model 1, 10.6 in

Model 2 and 11.5 in Model 3. The scatter in the data in Model 2 is slightly lower

than in Models 1 and 3. The bare ground predictions in Model 2 are less biased.

Model 2 represents GV slightly better than Model 3 while Model 3 seems to

have lower bias in the predictions. NGV is better represented in Model 2 than in

Model 3 with almost no bias.

The RMS errors in all three models are higher than the reported values in the

model calibration; this can be seen as a result of several factors:

- a large time difference of up to 60 days between field and image

observations was allowed;

- no weighting function for the time difference was applied;

- no outlier removal was applied;

- additional field sites were used as compared to the model calibration;

- some of the additional field data may have (depending on the field

operator) differences in the separation between green and non green

vegetation in the field observations which can lead to false estimated in

these components.

The error estimates presented here should be seen as a (very) conservative

error estimate. Figure 8 shows an improved plot with a maximum time

difference of 15 days between field and image observations (351 data points).

The scatter is much reduced with improved error statistics particularly in Model

3 (14.6%). Further analysis of the Model sensitivities is underway.

0.0

0.2

0.4

0.6

0.8

1.0

MAE=0.1

RMSE=0.162

Model I (linear)

Bare

0.0 0.2 0.4 0.6 0.8 1.0

MAE=0.11

RMSE=0.149

Model II (multinomial)

Bare

MAE=0.095

RMSE=0.146

Model III (S MA)

Bare

predicted

0.0 0.2 0.4 0.6 0.8 1.0 0.00.20.40.60.81.0

observed

Figure 8: Comparison for the bare ground component with field observations with a

maximum of 15 days time difference between field and image observation (see Figure

2).

The residuals per decile of all three models (Figure 3) show slight systematic

errors in some fractions, which could potentially be improved with better model

fits. The errors in the estimates can potentially be reduced by using USGS

imagery in the analysis, as the time difference between image observation and

field observation for a fast changing surface component, such as ground cover,

is important. With the United States Geologic Survey (USGS) opening their

Landsat image archive freely to the public, all archived images (after 2000) are

now available. This coverage of up to 20 images per year creates new

opportunities for Landsat based monitoring approaches and will potentially

reduce the errors in the ground cover model.

The differences between the two Landsat data sources (ACRES and USGS) in

Model 1 were 5% or less in natural environments and in an agricultural site, with

the maximum difference of 7%, found to be acceptable for a ‘fair’ model

comparison, as differences are all below the ground cover model accuracy.

The under-estimation of the bare ground in the freshly ploughed field in the

agricultural example might be due to the fact that the surface is disturbed by the

ploughing, but also dark soils have historically been the most difficult soil

backgrounds for estimating ground cover. The field observations with high

values of greenness in the wheat field also posed some difficulties for all three

model predictions. The cases of absolutely bare and entirely lush green cover

are extreme cases that were historically not captured in the rangelands and

thus are outside of the data range used in the calibration of the data models.

This leads to the conclusion that estimates at these extremes need to be

interpreted carefully and that more data points in agricultural areas might be

needed to account for these situations. However, the three models perform well

in agricultural areas that hold a ground cover similar to rangeland conditions, for

example with covers different from nearly 100% bare or 100% green. The

predictions for GV and NGV appear useful in the agricultural area shown here,

despite a small bias in the predictions.

Acknowledgements

Thanks to everyone involved in the Goondiwindi fieldwork and to the property

owner Peter Russell who allowed the fieldwork. The authors would like to thank

the reviewers who helped to improve this manuscript.

References

Bastin, G.N., Smith, D.M.S., Watson, I.W. & Fisher, A. 2009, "The Australian

Collaborative Rangelands Information System: preparing for a climate of

change", The Rangeland Journal, vol. 31, no. 1, pp. 111-125.

de Vries, C., Danaher, T., Denham, R., Scarth, P. & Phinn, S. 2007, "An

operational radiometric calibration procedure for the Landsat sensors based

on pseudo-invariant target sites", Remote Sensing of Environment, vol.

107, no. 3, pp. 414-429.

Karfs, R.A., Abbott, B.N., Scarth, P.F. & Wallace, J.F. 2009, "Land condition

monitoring information for Reef catchments: A new era", Rangeland

Journal, vol. 31, no. 1, pp. 69-86.

Leys, J., Smith, J., MacRae, C., Xihua, J., Randall, L., Hairsine, P., Dixon, J. &

McTanish, G. 2009, Improving the capacity to monitor wind and water

erosion: A review., Brureau of Rural Sciences, Canberra.

Ludwig, J.A., Bastin, G.N., Wallace, J.F. & McVicar, T.R. 2007, "Assessing

landscape health by scaling with remote sensing: When is it not enough?",

Landscape Ecology, vol. 22, no. 2, pp. 163-169.

Muir, J.S. & Danaher, T. 2008, "Mapping water body extent in Queensland

through time series analysis of Landsat imagery", Proceedings of the 14th

ARSPC conference, Darwin. ARSPC.

Queensland Government 2009, Reef Water Quality Protection Plan 2009 for the

Great Barrier Reef World Heritage Area and Adjacent Catchments,

Australian Government; Queensland Government.

Scarth, P., Byrne, M., Danaher, T., Henry, B., Hassett, R., Carter, J. & Timmers,

P. 2006, "State of the paddock: monitoring condition and trend in

groundcover across Queensland", Proceedings of the 13th ARSPC

conference, Canberra. ARSPC, .

Scarth, P., Roder, A., Schmidt, M. & Denham, R. in prep., "Fractional

groundcover using constrained spectral mixture analysis".

Schmidt, M., Muir, J.S. & Scarth, P. Draft, Handbook for Surface Cover Field

Observation, Bureau of Rural Sciences, Canberra.

Schmidt, M., Tindall, D., Speller, K., Scarth, P. & Dougall, C. 2010, Ground

cover management practices in cropping and improved pasture grazing

systems: ground cover monitoring using remote sensing, Bureau of Rural

Sciences, Canberra.

Schmidt, M. & Trevithick, R. 2010, " Seasonal ground cover monitoring in the

grazing lands of the Great Barrier Reef catchments with Landsat time

series data", Proceedings of the 15th ARSPC conference, Alice Springs.

ARSPC.

Schmidt, M., Denham, R.J. & Scarth, P. in prep, "Large scale fractional

vegetative ground cover monitoring based on long term field observations

and LANDSAT satellite imagery".

Tongway, D.J. & Hindley, N. 1995, Manual for soil condition assessment of

tropical grasslands, CSIRO Division of Wildlife and Ecology, Lyneham,

A.C.T.

Trevithick, R. & Gillingham, S. 2010, "The use of open source geospatial

software within the remote sensing centre, QLD", Proceedings of the 15th

ARSPC conference, Alice Springs. ARSPC.

Green space in health research: an overview of common indicators of greenness

Article

Full-text available

Nov 2022
Rev Environ Health

Human environments influence human health in both positive and negative ways. Green space is considered an environmental exposure that confers benefits to human health and has attracted a high level of interest from researchers, policy makers, and increasingly clinicians. Green space has been associated with a range of health benefits, such as improvements in physical, mental, and social wellbeing. There are different sources, metrics and indicators of green space used in research, all of which measure different aspects of the environment. It is important that readers of green space research understand the terminology used in this field, and what the green space indicators used in the studies represent in the real world. This paper provides an overview of the major definitions of green space and the indicators used to assess exposure for health practitioners, public health researchers, and health policy experts who may be interested in understanding this field more clearly, either in the provision of public health-promoting services or to undertake research.

Examining Spatiotemporal Photosynthetic Vegetation Trends in Djibouti Using Fractional Cover Metrics in the Digital Earth Africa Open Data Cube

Article

Full-text available

Mar 2024

The Horn of Africa has sensitive, arid ecosystems, with its vegetation commonly distressed by factors such as climate change, population increase, unstable water resources, and rarely enforced land use management practices. These factors make countries such as Djibouti highly variable locations for the growth of vegetation and agricultural products, and these countries are becoming more vulnerable to food insecurity as the climate warms. The rapid growth of satellite and digital image processing technology over the last five decades has improved our ability to track long-term agricultural and vegetation changes. Data cubes are a newer approach to managing satellite imagery and studying temporal patterns. Here, we use the cloud-based Digital Earth Africa, Open Data Cube to analyze 30 years of Landsat imagery and orthomosaics. We analyze long-term trends in vegetation dynamics by comparing annual fractional cover metrics (photosynthetic vegetation, non-photosynthetic vegetation, and bare ground) to the Normalized Difference Vegetation Index. Investigating Djibouti-wide and regional vegetation trends, we provide a comparison of trends between districts and highlight a primary agricultural region in the southeast as a detailed example of vegetation change. The results of the Sen’s slope and Mann–Kendall regression analyses of the data cube suggest a significant decline in vegetation (p = 0.00002), equating to a loss of ~0.09 km² of arable land per year (roughly 2.7 km² over the 30-year period). Overall, decreases in photosynthetic vegetation and increases in both non-photosynthetic vegetation and bare soil areas indicate that the region is becoming more arid and that land cover is responding to this trend.

Is ground cover a useful indicator of grazing land condition?

Article

Full-text available

Sep 2021

Remotely sensed ground cover data play an important role in Australian rangelands research development and extension, reflecting broader global trends in the use of remotely sensed data. We tested the relationship between remotely sensed ground cover data and field-based assessments of grazing land condition in the largest quantitative analysis of its type to date. We collated land condition data from 2282 sites evaluated between 2004 and 2018 in the Burdekin and Fitzroy regions of Queensland. Condition was defined using the Grazing Land Management land condition framework that ranks grazing land condition on a four point ordinal scale based on dimensions of vegetation composition, ground cover level and erosion severity. Nine separate ground cover derived indices were then calculated for each site. We found that all indices significantly correlated with grazing land condition on corresponding sites. Highest correlations occurred with indices that benchmarked ground cover at the site against regional ground cover assessed over several years. These findings provide some validation for the general use of ground cover data as an indicator of rangeland health/productivity. We also constructed univariate land condition models with a subset of these indices. Our models predicted land condition significantly better than random assignment though only moderately well; no model correctly predicted land condition class on >40% of sites. While the best models predicted condition correctly at >60% of A and D condition sites, condition at sites in B and C condition sites was poorly predicted. Several factors limit how well ground cover levels predict land condition. The main challenge is modelling a multidimensional value (grazing land condition) with a unidimensional ground cover measurement. We suggest that better land condition models require a range of predictors to address this multidimensionality but cover indices can make a substantial contribution in this context.

SEASONAL GROUND COVER MONITORING WITH LANDSAT TIME SERIES DATA IN THE GRAZING LANDS OF THE GREAT BARRIER REEF CATCHMENTS

Conference Paper

Full-text available

Sep 2010

Ground cover is a critical attribute of the landscape affecting infiltration and can functionally be related to surface runoff, nutrient fluxes and soil erosion. Ground cover relates to living and non-living materials on the soil surface. Besides the direct economic value for graziers, varying levels of ground cover have indirect economic and major ecological implications on productivity, land condition and biodiversity. Remote sensing offers one of the few ways to monitor long term ground cover change over large spatial extents. This work is supporting the recently initiated Great Barrier Reef Water Quality Protection Plan to limit negative environmental impacts to the Great Barrier Reef. Landsat Thematic Mapper and Enhanced Thematic Mapper+ imagery have recently become freely available and thus make possible time series applications on medium-high spatial resolution data at a temporal resolution greater than annual. This contribution outlines the operational product development of a seasonal ground cover product based on different temporal compositing methods. Introduction Queensland's rangelands are subject to high climate variability on seasonal, annual, decadal and longer timescales, making management for economic and environmental sustainability difficult. Ground cover levels may vary due to anthropogenic management of grazing enterprises and agricultural land management practices, or natural seasonal changes in rainfall. Dynamic ground cover information is important for reliably estimating soil erosion and nutrient flux into the stream network and the adjacent Great Barrier Reef (GBR). With the United States Geologic Survey (USGS) opening their Landsat image archive freely to the public, all archived images from 1999 onwards are now available with next to no budget constraints. This coverage of up to 20 images per year creates new opportunities for Landsat based time series analysis and environmental monitoring approaches.

The association of fractional cover, foliage projective cover and biodiversity with birthweight

Article

Oct 2020
SCI TOTAL ENVIRON

Introduction Environmental exposures can contribute both benefits and risks to human health. Maternal exposure to green space has been associated with improvements in birthweight, among other birth outcomes. Newer measures of green space have been developed, which allows for an exploration of the effect of different ground covers (green, dry and bare earth), as well as measures of biodiversity. This study explores the association of these novel green space measures with birthweight in a large birth cohort in Queensland, Australia. Methods Birthweight was acquired from the routine health records. Records were allocated green space values for fractional cover, biodiversity and foliage projective cover. Directed acyclic graphs were developed to guide variable selection. Mixed-effects linear regression and generalised linear mixed-effects models were developed, with random intercepts for maternal residential locality and year of birth. Results are presented as standardised beta coefficients or odds ratios, with 95% confidence intervals. Results An IQR increase of green cover (29.6 g, 95% CI 13.8–45.5) and foliage projective cover (26.0 g, 95% CI 10.8–41.3) are associated with birthweight in urban areas. An IQR increase in dry cover −34.4 g, 95% CI -60.4 to −8.4) and bare earth (−17.7 g, 95% CI -32.8 to −2.6) are associated with lower birthweight. Mothers living in rural areas had similar results, with an IQR increase in green cover (17.8 g, 95% CI 2.9–32.7) associated with higher birthweight, and bare earth (−27.7 g, 95% CI -45.7 to −9.7) was associated with lower birthweight. The biodiversity measure used in this study was not associated with any birthweight outcomes. Conclusion This study finds that the types of ground cover within the maternal residential locality are associated with small, but significant, changes in estimated birthweight, and these effects are not limited to urban areas.

Mapping and Analysis of Nature and Land Asset Health for the Condamine Catchment, Final report to the Condamine Alliance Natural Resource Management Body. Toowoomba, Queensland, Australia.

Technical Report

Nov 2012

Disclaimer The authors of this report and their respective organisations are not responsible for the outcomes of any actions taken on the basis of information in this research report, nor for any errors and omissions. The information contained in this publication is intended for general use, to assist public knowledge and discussion. Readers should seek expert advice before taking any action or decision based on the information in this publication. To the extent permitted by law, the authors of this publication and their employers do not assume liability of any kind whatsoever resulting from any person's use or reliance upon the content of this publication

THE QUEENSLAND GROUND COVER MONITORING PROGRAM

Conference Paper

Full-text available

Aug 2012

Ground cover plays an important role in Australia's environmental, agricultural and economic sustainability. It is effective in reducing the loss of sediments through wind and water erosion and supports a diverse range of biodiversity. The Queensland Remote Sensing Centre (RSC) in the Queensland Government has an established ground cover monitoring program. The Queensland Ground Cover Monitoring Program (QGCMP) uses satellite imagery and field data to research and develop state-wide, regional and local ground cover mapping and monitoring products. The products support government legislation, conservation efforts and land management initiatives. The QGCMP is based on the extensive archives and processing streams for Landsat and other imagery. The Program developed the Ground Cover Index (GCI) which uses ~25 years of Landsat TM/ETM+ imagery and field data to quantify ground cover levels. The resultant index was applied across Queensland with an RMSE of 12.9%. Subsequent additions and improvements to the QGCMP have included: establishment of a supporting field monitoring program including research of new field-based technologies; development of new unmixing algorithms for fractional vegetation cover; research and development of systems and methods for time-series analysis for more regular monitoring; and, cross-calibrations between sensors for deriving and modelling biomass and other ground cover metrics. This paper outlines some of the research undertaken and products developed for the QGCMP and discusses the application of the products in key state and federal government policies and initiatives.

Analysing and quantifying vegetation responses to rainfall with high resolution spatio-temporal time series data for different ecosystems and ecotones in Queensland

Conference Paper

Full-text available

Sep 2012

Vegetation responses and ecosystem function are spatially variable and influenced by climate variability. The Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) was used to combine MODIS (Moderate Resolution Imaging Spectrometer) and Landsat TM/ETM+ (Thematic Mapper/ Enhanced Thematic Mapper plus) imagery for an 8 year dataset (2000-2007) at 30m spatial resolution with 8 day intervals. This dataset allows for a functional analysis of ecosystem responses, suitable for heterogeneous landscapes. Derived vegetation index information in form of the NDVI (Normalised Difference Vegetation Index) was used to investigate the relationship between vegetation responses and gridded rainfall data for regional ecosystems. A hierarchical decomposition of the time series has been carried out in which relationships among the time-series were individually assessed for deterministic time-series components (trend component and seasonality) as well as for the stochastic seasonal anomalies. While no common long-term trends in NDVI and rainfall data in the time period considered exist, there is however, a strong concurrence in the seasonally of NDVI and rainfall data. This component accounts for the majority of variability in the time-series. On the level of seasonal anomalies, these relationships are more subtle. The statistical analysis required, among others, the removal of temporal autocorrelation for an unbiased assessment of significance. Significant lagged correlations between rainfall and NDVI were found in complex Queensland savannah vegetation communities. For grasslands and open woodlands, significant relationships with lag times between 8 and 16 days were found. For denser, evergreen vegetation communities greater lag times of up to 2.5 months were found. The derived distributed lag models may be used for short-term NDVI and biomass predictions on the spatial resolution scale of Landsat (30m).

Use of open source geospatial software within the Remote Sensing Centre, QLD, Australia

Conference Paper

Full-text available

Oct 2009

The Australian Collaborative Rangelands Information System: Preparing for a climate of change

Article

Full-text available

Apr 2009

Change is a constant in Australia’s rangelands. Appropriate management,of this change,requires a sound knowledge of drivers (e.g. climate variability, livestock grazing), their impacts on natural resources (state and trend), socio-economic outcomes, and how these feed back through learning and adaptive management to affect drivers and their impacts.Informationisrequiredatscalesfromenterprisetonational,withregionalandbroaderlevelinformationservingto influence rangelands governance through institutional arrangements, policy and funding programs. The Australian Collaborative Rangelands Information System (ACRIS) collates and analyses datafrom,national sources andfrom,its State and Territory jurisdictional partners to track and understand,change,at regional to national scales. ACRIS has recently reportedchangesbetween1992and 2005inseveralbiophysicalandsocio-economic themesatbioregionalresolution. This paperdescribestheprocessesusedtocollateandanalysetheoftendisparatedata,tosynthesiseinformationacrossdatatypes and to integrate emergent higher order information across drivers, impacts and outcomes to provide more complete understanding ofchange. Data gaps and inconsistencies were a major challenge, and we illustrate how some of these issues were addressed,by using indicators to report changes,in biodiversity. ACRIS now,needs to foster increased coordinated monitoring activity and develop its reporting capacity to become,the valued information system for Australia’s rangelands. We propose,that future improvements,will be best structured within a hierarchically nested framework,that provides consistent overarching,data at national scale relevant to the variety of rangeland values (e.g. change,in ground,cover) but focusesonregionally-relevantecosystemservices,andtheirappropriatemeasures,attheregionalscale.Akeychallengeisto implementconsistentandsystematicmethodsformonitoringbiodiversitywithinthishierarchicalframework,givenlimited institutionalresources.Finally,ACRISneedstodevelopadynamicweb-baseddeliverysystemtoenablemorefrequentand flexible reporting of interpreted change,than is possible through periodic published reports. Additional keywords: biodiversity, institutions, landscape function, livestock, rangeland monitoring, seasonal quality,

Ground cover management practices in cropping and improved pasture grazing systems: ground cover monitoring using remote sensing. FINAL REPORT

Article

Full-text available

May 2010

1. Introduction 1.1. Ground cover and erosion Ground cover is the proportion of the underlying soil material that is covered by the vegetative ground layer or superficial rock material. The vegetative ground layer consists of living or dead plant material and includes live grasses and forbs, leaves and branches and cryptogams such as mosses and lichens. Ground cover changes in response to climate, vegetation dynamics and land management. The quantity of ground cover affects water infiltration, runoff and erosion. It is also linked to the key land condition indicators of pasture production and biodiversity. Estimates of ground cover and changes to the quantities and spatial arrangement of ground cover over time provide land managers, policy-makers and scientists with valuable information. This helps inform decisions in relation to soil erosion, soil carbon, water quality, pasture production, land condition, climate change, fire and regional management targets. For the purposes of this report, we refer to ground cover as the vegetative ground cover component, which is detectable through the use of satellite imagery. A recent report by Leys et al. (2009) identified ground cover as a key indicator of land management practices. Importantly, ground cover can be used to infer and monitor wind and water erosion risk. The report noted that many catchment action plans of regional Natural Resource Management (NRM) bodies are using ground cover as a surrogate for erosion risk. However, at both the national and regional levels, there remains a lack of comprehensive, consistent ground cover data at a temporal and spatial scale adequate for monitoring and assessing environmental targets related to soil erosion and land management. 1.2. Remote sensing of ground cover Ground cover estimates, derived from remotely sensed imagery are important land condition indicators and become increasingly useful when a time-series of ground cover information is available. A number of techniques have been developed for estimating ground cover quantities and crop residues using remotely sensed imagery. For example, Daughtry (2001) and Daughtry et al. (2005, 2006) used the spectral libraries of crop residues and crop types, combined with multispectral (Landsat TM) and hyperspectral imagery (AVIRIS, Hyperion) to compare spectral residue and soil indices, including the cellulose absorption index (CAI), to assess crop residue levels and types in the United States. In Australia, a comprehensive review of techniques was undertaken by Leys et al. (2009) (refer to appendix 3 in their report). In their report, and following an expert workshop, they summarised four ground cover estimation techniques that may be most suitable for erosion modelling. These techniques included: i. Annual estimates of woody fractional cover using the Queensland statewide landcover and trees study (SLATS) method based on Landsat TM data (Danaher et al., 1998). ii. Monthly bare ground index (BGI) methods for Landsat (Scarth et al., 2006) and MODIS (Milne et al., 2007). iii. Monthly fractional cover using the CSIRO MODIS non-woody fractional cover method (Guerschman et al., 2009). iv. Monthly fractional cover using the CSIRO AVHRR data archive (Donohue et al., 2008). There are, in fact, five products listed above as ii. contains both Landsat and MODIS scale cover products. Each of these products derives slightly different estimates of cover due to what they measure and the temporal and spatial scale at which they operate. The original estimates of woody fractional cover (Danaher et al., 1998) undertaken by the Queensland Remote Sensing Centre (QRSC) are presently generated as annual estimates of woody Foliage Projective Cover (FPC) following the methods described by Danaher et al. (2004). The BGI methods of Scarth et al. (2006) are presently annual estimates of ground cover as derived from a Ground Cover Index (GCI) using Landsat imagery. The GCI is calibrated using ground-based measurements of ground cover fractions, primarily from rangeland systems. The BGI methods of Milne et al. (2007) use the MODIS 16-day and 8-day composites at 1km and 500m resolution respectively, and were calibrated using the Landsat GCI product. The products developed by Guerschman et al. (2009) use field-derived spectral libraries of fractional ground cover in Australian tropical savannas. The spectral libraries in this method were compiled using EO-1 Hyperion hyperspectral imagery to explore the spectral response space to develop fractional cover endmembers. These were used along with the Normalised Difference Vegetation Index (NDVI) and the CAI (Cellulose Absorption Index) in a linear mixture approach. The approach was then applied to daily MODIS imagery at 500m resolution (coincident with the Hyperion imagery) and then to six years of the MODIS 16-day composite imagery at 1km resolution to resolve quantitative estimates of photosynthetic vegetation (PV), non-photosynthetic vegetation (NPV) and bare soil for ~2 million km2 of the Australian tropical savanna zone. More recently, this approach has been applied to the MODIS 8-day composite at 500m resolution for the entire Australian continent. The monthly fractional cover estimates produced by Donohue et al. (2008) are a product of a relative reflectance calibration technique that assumes that the position of the vegetation cover triangle is invariant in reflectance space. This technique has been applied nationally to AVHRR data at approximately 1km resolution. Of the ground cover and crop residue measurement techniques described above, none of the methods developed have been designed or calibrated for the range of intensive agricultural systems (i.e. cropping and improved pasture) in Australia. Some studies have been undertaken using remote sensing time-series techniques to monitor Australian crop types and seasonal patterns in cropping rotations (e.g. Potgeiter et al., 2007; Pringle et al., 2008). However, these studies do not provide estimates of ground cover and crop residue levels at different stages of the cropping cycle and have used imagery (e.g. MODIS) that is at a resolution too coarse to accurately monitor many cropping systems in Australia.

Improving the Capacity to Monitor Wind and Water Erosion: A Review

Article

Full-text available

Jan 2009

SEASONAL GROUND COVER MONITORING WITH LANDSAT TIME SERIES DATA IN THE GRAZING LANDS OF THE GREAT BARRIER REEF CATCHMENTS

Conference Paper

Full-text available

Sep 2010

Assessing landscape health by scaling with remote sensing: When is it not enough?

Article

Full-text available

Feb 2007

Assessment of the health of landscapes, by monitoring their condition over space and time, is needed to better understand the processes for sustaining or, in many cases, repairing them. Remote sensing is a tool that can efficiently identify and assess areas of landscape damage at different scales and help land managers solve specific problems. Remote sensing may appear to be a panacea for all monitoring situations but sometimes the information it provides is not enough by itself. In this paper we give examples of both scenarios—when remote sensing alone is adequate and when it is not. When remotely sensed data alone is not sufficient, monitoring problems can be solved by incorporating additional finer scale data. We use a five-step procedure based on scaling to help land managers answer the question: when is remote sensing data alone not sufficient to underpin the information needs required to achieve a specific management goal?

Land condition monitoring information for Reef catchments: A new era

Article

Full-text available

Apr 2009

Land condition monitoring information is required for the strategic management of grazing land and for a better understanding of ecosystem processes. Yet, for policy makers and those land managers whose properties are situated within north-eastern Australia's vast Great Barrier Reef catchments, there has been a general lack of geospatial land condition monitoring information. This paper provides an overview of integrated land monitoring activity in rangeland areas of two major Reef catchments in Queensland: the Burdekin and Fitzroy regions. The project aims were to assemble land condition monitoring datasets that would assist grazing land management and support decision-makers investing public funds; and deliver these data to natural resource management(NRM) community groups, which had been given increased responsibility for delivering local environmental outcomes. We describe the rationale and processes used to produce new land condition monitoring datasets derived from remotely sensed Landsat thematic mapper (TM) and high resolution SPOT 5 satellite imagery and from rapid land condition ground assessment. Specific products include subcatchment groundcover change maps, regional mapping of indicative very poor land condition, and stratified land condition site summaries. Their application, integration, and limitations are discussed. The major innovation is a better understanding of NRM issues with respect to land condition across vast regional areas, and the effective transfer of decision-making capacity to the local level. Likewise, with an increased ability to address policy questions from an evidence-based position, combined with increased cooperation between community, industry and all levels of government, a new era has emerged for decision-makers in rangeland management.

An operational radiometric calibration procedure for the Landsat sensors based on pseudo-invariant target sites

Article

Apr 2007
REMOTE SENS ENVIRON

The Statewide Landcover and Trees Study (SLATS) use both Landsat-7 ETM+ and Landsat-5 TM imagery to monitor short-term woody vegetation changes throughout Queensland, Australia. In order to analyse more subtle long-term vegetation change, time-based trends resulting from artefacts introduced by the sensor system must be removed. In this study, a reflectance-based vicarious calibration approach using high-reflectance, pseudo-invariant targets in western Queensland was developed. This calibration procedure was used to test the existing calibration models for ETM+ and TM, and develop a consistent operational calibration procedure which provides calibration information for the MSS sensors. Ground based data, sensor spectral response functions and atmospheric variables were used as input to MODTRAN radiative transfer code to estimate top-of-atmosphere radiance. The estimated gains for Landsat-7 ETM+ (1999–2003), -5 TM (1987–2004), -5 MSS (1984–1993) and -2 MSS (1979–1982) are presented. Results confirm the stability and accuracy of the ETM+ calibration, and the suitability of this data as a radiometric standard for cross-calibration with TM. Vicarious data support the use of the existing TM calibration model for the red and two shortwave-infrared bands. However, alternative models for blue, green and near-infrared bands are presented. The models proposed differ most noticeably at dates prior to 1995, with differences in estimated gains of up to 9.7%, 10.8% and 6.9% for the blue, green and near-infrared bands respectively. Vicarious gains for Landsat-2 MSS and Landsat-5 MSS are presented and are compared with those applied by the on-board calibration system. Updated calibration coefficients to scale MSS data to the SLATS vicarious measurements are given. The removal of time based calibration trends in the SLATS data archive will enable the measurement of vegetation changes over the 26 year period covered by Landsat -2, -5 and -7.

Mapping water body extent in Queensland through time series analysis of Landsat imagery

Jan 2008

J S Muir
T Danaher

Muir, J.S. & Danaher, T. 2008, "Mapping water body extent in Queensland through time series analysis of Landsat imagery", Proceedings of the 14th ARSPC conference, Darwin. ARSPC.

Reef Water Quality Protection Plan 2009 for the Great Barrier Reef World Heritage Area and Adjacent Catchments, Australian Government

Jan 2009

Queensland Government

Queensland Government 2009, Reef Water Quality Protection Plan 2009 for the Great Barrier Reef World Heritage Area and Adjacent Catchments, Australian Government; Queensland Government.

FRACTIONAL GROUND COVER MONITORING OF PASTURES AND AGRICULTURAL AREAS IN QUEENSLAND

Abstract and Figures

Recommended publications

My knowledge is shaping the future.

DLR research in and from space

DLR research for the energy system of the future

Get to know the German Aerospace Center as an Employer

Assessment of changes in the area of the water conservation forest in the Qilian Mountains of China’...

Participatory Monitoring and Evaluation of a Project to Combat Desertification in Drylands ( Case St...

Participatory Monitoring and Evaluation of a Project to Combat Desertification in Drylands (Case Stu...

Soil Erosion in Europe