Management of Environmental Quality: An International Journal
Emerald Article: A remote sensing based monitoring system for 
discrimination between climate and human-induced vegetation change in 
Central Asia
P.A. Propastin, M. Kappas, N.R. Muratova
Article information:
To cite this document: P.A. Propastin, M. Kappas, N.R. Muratova, (2008),"A remote sensing based monitoring system for 
discrimination between climate and human-induced vegetation change in Central Asia", Management of Environmental Quality: An 
International Journal, Vol. 19 Iss: 5 pp. 579 - 596
Permanent link to this document: 
http://dx.doi.org/10.1108/14777830810894256
Downloaded on: 27-11-2012
References: This document contains references to 23 other documents
Citations: This document has been cited by 1 other documents
To copy this document: permissions@emeraldinsight.com
Access to this document was granted through an Emerald subscription provided by NIEDERSAECHSISCHE STAATS UNIVERSITAETSBIBLIOTHEK 
GOETTINGEN                                         
For Authors: 
If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service. 
Information about how to choose which publication to write for and submission guidelines are available for all. Please visit 
www.emeraldinsight.com/authors for more information.
About Emerald  www.emeraldinsight.com
With over forty years' experience, Emerald Group Publishing is a leading independent publisher of global research with impact in 
business, society, public policy and education. In total, Emerald publishes over 275 journals and more than 130 book series, as 
well as an extensive range of online products and services. Emerald is both COUNTER 3 and TRANSFER compliant. The organization is 
a partner of the Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive 
preservation.
*Related content and download information correct at time of download.
A remote sensing based
monitoring system for
discrimination between climate
and human-induced vegetation
change in Central Asia
P.A. Propastin
Department of Geography, Georg-August University Go¨ttingen, Go¨ttingen,
Germany and Laboratory of Remote Sensing and Image Analysis,
Kazakh Academy of Science, Almaty, Kazakhstan
M. Kappas
Department of Geography, Georg-August University Go¨ttingen, Go¨ttingen,
Germany, and
N.R. Muratova
Laboratory of Remote Sensing and Image Analysis,
Kazakh Academy of Science, Almaty, Kazakhstan
Abstract
Purpose – This paper aims to demonstrate the importance of taking into account precipitation and
the vegetation response to it when trying to analyse changes of vegetation cover in drylands with high
inter-annual rainfall variability.
Design/methodology/approach – Linear regression models were used to determine trends in
NDVI and precipitation and their interrelations for each pixel. Trends in NDVI that were entirely
supported by precipitation trends were considered to impose climate-induced vegetation change.
Trends in NDVI that were not explained by trends in precipitation were considered to mark
human-induced vegetation change. Modelling results were validated by test of statistical significance
and by comparison with the data from higher resolution satellites and fieldtrips to key test sites.
Findings – More than 26 percent of all vegetated area in Central Asia experienced significant
changes during 1981-2000. Rainfall has been proved to enforce most of these changes (21 percent of the
entire vegetated area). The trends in vegetation activity driven by anthropogenic factor are much
scarcer and occupy about 5.75 percent of the studied area.
Practical implications – Planners, decision makers and other interest groups can use the findings
of the study for assessment and monitoring land performance/land degradation over dry regions.
Originality/value – The study demonstrates the importance of taking into account precipitation and
the vegetation response to it when trying to analyse changes of vegetation cover in drylands with high
inter-annual rainfall variability.
Keywords Deserts, Environmental management, Climatology, Central Asia, Landforms
Paper type Research paper
Introduction
Desertification refers to land degradation at arid, semi-arid and dry sub-humid areas
and has been seen as one of the major environmental problems in large parts of the
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1477-7835.htm
Remote
monitoring
system
579
Received 25 August 2007
Revised 29 February 2008
Accepted 31 March 2008
Management of Environmental
Quality: An International Journal
Vol. 19 No. 5, 2008
pp. 579-596
q Emerald Group Publishing Limited
1477-7835
DOI 10.1108/14777830810894256
land surface. Desertification is considered as result of a series of complex natural,
mainly climatic, and anthropogenic processes that leads to gradual environmental
degradation or loss of the lands biological and economical productivity. The processes
leading to desertification manifest in all compartments of ecosystems and are divided
into the following groups: vegetative, biological, chemical, hydrological, and
morpho-dynamical (UNCOD, 1977). Drylands occupy about 40 percent of the land
surface, contain about 30 percent of the world’s carbon and provide habitat to more
than 1 billion humans. Their rangeland supports approximately 50 percent of the
world’s livestock. Taking into account the importance of these territories for the
survival of the mankind it is not surprising that the international community is making
intensive efforts to combat desertification in drylands and to preserve them from
further degradation.
The first step on the way to stop and combat desertification is to detect areas
already degraded or undergoing degradation and to assess the desertification degree.
International institutions like FAO or UNEP have developed multidimensional
methodologies for desertification assessment and promoted the monitoring of
desertification with these methodologies. Common approaches to assess desertification
demand for measurements of different indicators, which usually describe one or more
aspects of desertification and provide data on threshold levels, status and evolution of
relevant processes (FAO/UNEP, 1984; Veron et al., 2006). Usually, these indicator
measurements require both extensive fieldworks and analysis of remotely sensed data.
Degradation of vegetation cover is one of the most important and the most effective
desertification indicator and can be rather easily monitored over broad areas using
satellite imagery. Satellite derived Normalized Difference Vegetation Index (NDVI) has
been commonly used as a general proxy of vegetation conditions and is a convenient
tool for monitoring of vegetation cover at all scales from global to local. The vegetation
absorbs a great part of incoming radiation in the visible portion of the electro-magnetic
spectrum (VIS ¼ 380-730 nm) and reaches maximum reflectance in the near-infrared
channel (NIR ¼ 730-1100 nm). The NDVI, defined as ratio (NIR-VIS)/(NIR þ VIS),
represents the absorption of photosynthetic active radiation and hence is a
measurement of the photosynthetic capacity of the canopy. The NDVI is established
to be highly correlated to green-leaf density, absorbed fraction of photosynthetically
active radiation and above-ground biomass and can be viewed as a surrogate for
photosynthetic capacity.
The NDVI has successfully served as a vegetation indicator in many studies on
drought in watching desertification and land degradation (Kogan, 1997; Weiss et al.,
2001; Wessels et al., 2004; Symeonakis and Drake, 2004). However, in many cases, the
only use of NDVI for desertification monitoring can be problematic, for the reason that
vegetation cover performance is strongly predicated on macro- and micro-climatic
factors, such as temperature and rainfall distribution change, local topography
characteristics etc. (Nicholson and Farrar, 1994; Yang et al., 1998; Li et al., 2002; Wang
et al., 2003). Therefore, discrimination between different causes of change in vegetation
cover, climate and human activity is rather difficult. The neglecting of this aspect can
lead to mistakes by evaluation of land conditions. For example, the contemporary
greening patterns in the Sahel are proofed to be driven by an increasing trend in
rainfall (Olsson et al., 2005; Anyamba and Tucker, 2005). A methodology of
discrimination between human and climate influence on vegetation cover can be based
MEQ
19,5
580
on identification of the climate signal in the inter-annual dynamics of vegetation
activity. Once the climate signal is identified, it can be removed from the trends in
vegetation activity. The remaining vegetation changes can be attributed to human
influence. A number of recent studies have already developed monitoring systems for
land degradation assessment that separate the dynamics of vegetation cover driven by
climate and by human activity (Evans and Geerken, 2004; Li et al., 2004). However,
these and other similar monitoring systems have been designed for the use in a certain
geographical region and are rather difficult transferable in other regions. Besides, all
existing monitoring systems for desertification assessment based on the use of
remotely sensed data are not perfect and the remote sensing community is making a
good deal of effort to improve these systems using the most modern understanding of
desertification and its causes.
In this study the authors have developed a simple and effective monitoring system
for separation of driving forces for degradation and rehabilitation of vegetation cover
in the former Soviet Republics of Central Asia (Figure 1). This separation is based on
synchronous monitoring of two desertification indicators: vegetative and climatic. The
monitoring system utilized a remotely sensed derived NDVI a dataset and dataset of
precipitation amounts over the whole area of Central Asia during the period of
1981-2000. We examined inter-annual trends in vegetation activity and precipitation
over the study period and separated climate-induced trends from human-induced.
Furthermore, the monitoring system developed in this study enabled us to detect
degradation/rehabilitation of vegetation cover in areas where no significant trends
could be defined. Areas of degradation and rehabilitation of vegetation cover were
mapped and measured.
Data used in the study
GIMMS NDVI dataset
Global inventory monitoring and modelling studies (GIMMS) NDVI dataset is a
derivate from the Advanced Very High Resolution Radiometer (AVHRR) data and is
freely available on the internet. The National Oceanic and Atmospheric Administration
Figure 1.
Map displays the region of
the former Soviet Central
Asia. The total area is
about 6.5 million km2
Remote
monitoring
system
581
(NOAA) launched the AVHRR in 1978. AVHRR NDVI product has been broadly used
for monitoring vegetation activity and environmental changes at regional and global
scales (Xiao and Moody, 2004).
The authors used a dataset of GIMMS NDVI over the region of Central Asia
associated with the period of 1981-2000. The GIMMS NDVI data, at 8 km spatial
resolution, are originally processed as 15-day composites using the maximum value
procedure to minimize effects of cloud contamination (Holben, 1986). Initially, the
GIMMS dataset was designed to remove non-vegetated signals, which are effects of
sensor degradation and stratosphere aerosols derived from the eruption of Mt.
El-Chichon in 1982 and Mt. Pinatubo in 1991. The GIMMS dataset is already calibrated
for sensor differences and orbital drift. However, before the GIMMS NDVI dataset was
used in our monitoring system, we had thoroughly investigated it for occurrence of any
non-vegetated noise. Various little noises remaining in the dataset were eliminated by
calibrating the data against three time-invariant desert targets using a method
described by Los (1993). After that, the data were averaged to generate a mean
growing-season NDVI, mean summer NDVI and mean spring-summer NDVI for each
year.
Precipitation dataset
A monthly precipitation dataset used in this work originated from the global
climatologic dataset produced by the School of Environmental Science, University of
East Anglia (www.cru.uea.ac.uk/). New et al. (2000) gives a full description of the
dataset and explains methods used for its derivation. This grid raster data are
interpolated from climate station records and have a basic spatial resolution of 0.5.
For this work, the monthly precipitation data were resized to 8 km resolution to get
a regional vision and to match the 8 km NDVI dataset. In order to minimize information
distortion and loss during the conversion process, we used an interpolation method
known as kriging with external drift. The use of an additional variable, altitude
significantly decreased all negative consequences on the resizing of the source 0.58
dataset.
Methods
General purpose of the methodology
Satellite derived NDVI has proven to be an effective tool for monitoring vegetation
cover and its conditions both at inter-annual and intra-annual time-scale. Previous
studies have also shown a strong relationship between inter-annual changes in
vegetation activity and precipitation, particularly in dry regions (Nicholson and Farrar,
1994; Richard and Poccard, 1998; Yang et al., 1998; Li et al., 2002; Wang et al., 2003).
Precipitation has a substantial control on NDVI through annual precipitation also in
Central Asia. Examples demonstrating strong relationships between inter-annual
dynamics of NDVI and precipitation in Central Asia are presented in Figure 2. The
graphs display the time-series of annual precipitation and growing season NDVI over
Kazakhstan and Uzbekistan during the period 1982-2000. To produce these graphs,
data were averaged over the country’s territory. This control, however, should be
predictable in every point of the study area where the relationship between NDVI and
precipitation are statistically significant. Since in drylands NDVI is usually stronger
affected by rainfall than any other factors including human activities, assessment of
MEQ
19,5
582
desertification may not be based on the use of NDVI time-series alone, it should also
take into account the rainfall.
Generally, speaking in terms of climatic effects on vegetation cover, a permanent or
contemporary decrease of annual precipitation causes a decrease of photosynthetic
activity of vegetation and diminishes the biological productivity potential of
ecosystem. In this case one can speak only about desertification caused by climatic
forces. However, if the human actions in the ecosystem does not adapt to the new
conditions of the rainfall scarcity, the climate-induced degradation processes may be
reinforced by human influence. Both driving forces of desertification, climate change
and human impact, are closely interrelated in a compound of desertification processes.
A separation of these driving forces is very difficult.
Nevertheless, a few recent studies tried to find solution of this problem by
identification and removing the climatic signal from inter-annual NDVI dynamics (Li
et al., 2004; Evans and Geerken, 2004). The authors developed a simple monitoring
system based on a concept of synchronism/non-synchronism between trends in rainfall
and NDVI. This concept and the developed model will be thoroughly addressed in the
next sections.
Concept of synchrony/asynchrony between long-time trends in precipitation and NDVI
Since both climate conditions and climate variations have a substantial control on
vegetation cover, we based our concept on the assumption that the response of
vegetation to precipitation is relatively constant over time if there is no change in
external factors, particularly human activity. A changed response to rainfall in an
otherwise responsive area could be the result of change in human activity and
influence.
This concept is explained on an example framework (Figure 3). In panel (a), the
upward trends in NDVI and precipitation are synchronous. Obviously, here we observe
improving vegetation cover due to increasing precipitation amount alone. In panel (b),
the downward trends in NDVI and precipitation are also synchronous. In this case,
decreasing NDVI is only driven by a decrease of precipitation without any
human-induced worsening of vegetation cover, because human impact is not evident in
the vegetation trend. In (c), the trends are asynchronous. NDVI increases even as
Figure 2.
Time-series of annual
precipitation and the mean
growing season NDVI for
Kazakhstan and
Uzbekistan
Remote
monitoring
system
583
precipitation decreases. This would indicate a case when vegetation cover is recovering
due to diminishing human impact. In (d), the trends are once more asynchronous, but
an increase of precipitation did not cause an improving of vegetation cover. On the
contrary, the NDVI trend is negative. Here, we can suppose human-induced
degradation of the vegetation cover.
However, degradation or improvement of vegetation cover is not always clearly
manifested in NDVI trends. Despite the fact that, at some places, vegetation might not
exhibit any significant changes, processes of degradation or rehabilitation of
vegetation cover may go on. These processes may be revealed through observation of
corresponding precipitation dynamics. If at these places precipitation amount shows
any significant upward or downward trend, the response of vegetation to precipitation
is changing over time. As the authors suggested that the response of vegetation to
precipitation is an indicator for performance of vegetation cover, the authors can
consider in such cases undergoing degradation or improvement caused by
anthropogenic factor. These cases are presented in Figure 4. Both panels
demonstrate vegetation cover being constant over the time of observation. However,
in case (a) a significant increase of precipitation occurs during the time. This rainfall
Figure 3.
Scenarios illustrating how
the combined use of NDVI
and precipitation
time-series may help to
detect vegetation cover
being improved or
degraded
Figure 4.
The same as in Figure 3
but for the areas without
significant trends in NDVI
MEQ
19,5
584
increase does not force any improvement of vegetation activity. It means a drop of
vegetation response to precipitation what could be a consequence of a degradation
process driven by non-climatic factors, particularly human influence. In Panel (b) is
shown an opposite case.
Figure 5 illustrates a theoretical and conceptual framework for discrimination
between climate-induced and human-induced trends in vegetation cover. The
framework aims to help by understanding the model that will be described in the
following sections.
Model: Identification of climate signal in vegetation change
The monitoring system based on synchrony/asynchrony concept would work only in
areas where relationship between trends in vegetation and trends in climate factors are
statistically significant. The first step of monitoring was to detect areas with
statistically significant relationship between precipitation and NDVI. These areas were
detected by regressing time-series of NDVI against precipitation amount at the
per-pixel scale. On this way we derived a map of areas where the control of vegetation
dynamics by precipitation is statistically significant and strong enough.
The second step was to find out pixels located within these areas, which exhibit any
statistically significant trend in NDVI over the period of 1981-2000. These pixels
represent territories of climate-driven change in vegetation cover. If the trend is
negative, the territory is considered to undergo degradation of vegetation cover driven
by a precipitation decrease. If the trend is positive, the territory undergoes
improvement of vegetation cover. For example, if an area reveals statistically
significant correlation with precipitation and NDVI increases throughout the study
period, it considers to indicate a precipitation-driven improvement in vegetation cover
(Figure 3(a)). If a trend in NDVI is negative and an area reveals strong relationship with
precipitation, it considers indicating degradation of vegetation cover due to decrease in
precipitation (Figure 3(b)).
Model: Identification of anthropogenic signal in vegetation change
In general, vegetation cover reacts very sensitively on changes of precipitation
particularly in dry regions (Yang et al., 1998; Richard and Poccard, 1998; Li et al., 2002;
Figure 5.
The same as in Figure 6
but for the case of a
human-driven
improvement of
vegetation cover
Remote
monitoring
system
585
Wang et al., 2003). If the response of vegetation to precipitation remains constant,
increase/decrease of vegetation activity should exactly reflect changes in precipitation.
That is the rule for undisturbed vegetation cover. Disturbance of vegetation cover is
manifested through changing its response to climate (Wessels et al., 2004; Li et al.,
2004). Generally, a worse response of vegetation to climate means a degradation of
vegetation cover caused by non-climatic factors, particularly human impact, while a
better response means an improvement of vegetation conditions. Figure 3 ((c) (d))
presents examples of human-induced changes in vegetation cover. In both cases the
trends in NDVI are opposite to trends in precipitation.
However, degradation or rehabilitation of vegetation cover may also occur without
any significant change in corresponding NDVI values (Figure 4). Figure 4 (a)
demonstrates a case where there is no trend in NDVI over time but vegetation response
to precipitation is getting worse. This would indicate a case of human-induced
degradation of vegetation cover. An opposite case is shown in Figure 4 (b). Here,
vegetation cover demonstrates increasing response to precipitation.
To detect human-induced trends in vegetation cover, the authors analysed linear
regressions between NDVI and rainfall time-series over the period 1981-2000. For a
given value of rainfall, a value of NDVI predicted by the regression, abbreviated
asNDVIpred , was obtained for every pixel and for each year, this value was considered
to reveal the climatic component. The observed NDVI, abbreviated asNDVIobs, may
show deviations from the regression line (Figure 6 (a)). Positive deviations indicate a
better response of vegetation to precipitation while negative deviations indicate a
worse response. Deviations in NDVIobs from NDVIpred expressed in the regression
residuals were computed at pixel-by-pixel basis for every year. On this way, the
authors derived the response strength for every year and every pixel. In order to detect
change in response strength, the authors looked into time-series of regression residuals.
The authors suggested that any trend through time presented in the residuals would
indicate changes in NDVI response not due to the climatic variable. A negative trend
would mean diminishing response of vegetation cover to climate. This reduce can be
caused either by a decrease of vegetation cover or by a change in plant species
composition. According to this suggestion, this negative trend, if it was statistically
Figure 6.
Potential effects of
degradation on the
observed rainfall-NDVI
relationships
MEQ
19,5
586
significant, would indicate an area experiencing human induced degradation.
Examples illustrating how the trends of regression residuals reflect changes in
vegetation response to precipitation are shown in Figure 6 and Figure 5. Panel (a) in
Figure 6 shows steady decrease of deviation of NDVIobs from NDVIpredover the time.
The deviation value turned from positive into negative during the observation period.
That means a decreasing response of vegetation cover to precipitation. Even though
precipitation amount rose, NDVI value remained at the same level. Panel (b) shows a
regression between NDVI and precipitation for an area with similar situation and panel
(c) shows residuals from the regression arranged in a time-sequence. An opposite case
would indicate a positive trend in residuals and is presented in Figure 5.
Results
Precipitation signal in the inter-annual dynamics of NDVI
From the previous studies it seems to be important to test several different analysis
periods in order to find an optimum correlation between time-series of NDVI and
precipitation. It is also apparent that over an area of about 6.5 million km2 dryland it is
unlikely that a single analysis period will provide the best correlation. Instead, the
authors would expect the optimum correlation period to vary with different vegetation
communities, with soil properties, with morphological characteristics (Nicholson and
Farrar, 1994; Yang et al., 1998; Wang et al., 2003; Li et al., 2002). Hence, correlations are
calculated for many different combinations of precipitation and NDVI, allowing
identification of its distinct optimum correlation (growing season, summer,
spring-summer etc.). The authors also tried to generate time series of precipitation
accumulated over two and more years and calculate correlation between them and time
series of NDVI.
Figure 7 shows the results of correlation calculations at the per-pixel level. The map
demonstrates the best correlation coefficient obtained for every pixel using different
analysis periods and precipitation accumulations. Our calculations revealed the
borders of the precipitation-dependent areas in Central Asia and Kazakhstan.
According to the results, about 75 percent of all vegetated pixels exhibited a
Figure 7.
Spatial distribution of the
best correlation coefficient
( p , 0.05) between
inter-annual NDVI and
precipitation derived by
using different
combinations of NDVI and
precipitation
Remote
monitoring
system
587
statistically significant correlation with precipitation over the 1981-2000. The value of
the correlation coefficient ranges from 0.47 to 0.92 indicating existence of a strong
relationship between NDVI and rainfall in these territories.
Long-time trends in NDVI
Spatial distribution of trends in growing season NDVI (April to October) from 1981 to
2000 over Central Asia is shown in Figure 8. In the north and south of the region
upwards trends spread broadly from the west border to the east. Especially high
increases in NDVI are associated with areas covered by forest vegetation in the
mountainous regions in the south as well as with areas covered by steppe vegetation.
Steppe occupies wide areas that stretch a broad band from the west to the east over the
whole study region. On the other hand, NDVI decreased in some areas located in the
central part, especially between the Caspian Sea and the Aral Sea, as well between the
Aral Sea and the Balkhash Lake. Taken over the entire study region, about 21 percent
of all vegetated pixels exhibited statistically significant upward trends of the growing
season NDVI throughout the study period, while 6.72 percent of all pixels exhibited
significant downward trends.
Explanation of NDVI trends by precipitation
In order to identify trends in NDVI driven by precipitation factor, the authors
compared the areas that exhibited significant correlation between NDVI and
precipitation with the areas of significant trends in NDVI. Intersection of these both
maps results in detecting pixels with NDVI trends caused by climate change. The
resulted map is shown in Figure 9. The map presents pixels that show both significant
trend in NDVI and significant correlation with inter-annual precipitation. The map
displays two categories of the climate-induced NDVI trends: positive trends which are
considered to represent an improvement of the vegetation cover, and negative trends
which are believed to indicate degradation of the vegetation cover. The entire area of
Figure 8.
Spatial distribution of
statistically significant
( p , 0.05) inter-annual
changes in mean growing
season NDVI over the
period 1981-2000
MEQ
19,5
588
precipitation driven trends is considerably less than that of all significant trends.
About 15 percent of all vegetated area is proved to exhibit upward trends in NDVI
driven by precipitation change. More than a half of significant downwards trends in
NDVI (4.7 percent of the entire vegetated area) showed a strong correlation with
precipitation.
Even a great deal of significant inter-annual changes in vegetation cover are proofed
to be dependent on trends in precipitation, a great part of the NDVI trends remained
unexplained by precipitation. One would expect that the unexplained part of the NDVI
trends could be driven by other internal and external factors. One of the most
important of them is human impact. The role of the human impact by explaining NDVI
trends will be investigated in the next section.
Detection of areas of human-induced NDVI change
The response of vegetation to climate is among other things determined by
non-climatic factors such as specific characteristics of plant physiology or structure of
vegetation canopy. In arid regions, annual and perennial species as well as grass and
shrub species differ in their response to precipitation. Changes in vegetation type may
also affect changes in the response of vegetation cover to rainfall. The man has a very
strong influence on vegetation cover, changing its biomass, structure, and mixture of
vegetation types. Therefore, in the precipitation dependent regions such as Central
Asia, human impact should be reflected in the response of vegetation cover to
precipitation factor, and the authors can measure this impact when detecting the
change of this response in time. On this way, it should be possible to identify not only
signals of vegetation degradation but also signals of vegetation improvement driven
by human impact.
The detection of change of the vegetation response was made through observation
of the deviations from the regression between the time-series of NDVI and
precipitation. The regression line was understood as the climatic signal. Deviation of
Figure 9.
Inter-annual changes in
NDVI explained by
precipitation
Remote
monitoring
system
589
the observed NDVI value from the value predicted by the regression was understood as
an indicator for the vegetation response to climate. Thus, any positive deviation
indicates better response while any negative deviation indicates worse response.
Having calculated these deviations for every year the authors derived time-series of
annual vegetation response for every pixel. These time-series contained important
information about evolution of the vegetation response to precipitation. Extraction and
interpretation of this information enabled identifying areas of vegetation change that
was caused by human impact. The authors supposed that any time-trend in vegetation
response would indicate a change of characteristics of the vegetation cover. A positive
trend would represent an enlargement of the vegetation response to precipitation. It
means that the vegetation improves its effectiveness of rain use. A consequence of that
is an increase of vegetation primary production per rainfall unit. It leads to a general
increase of aboveground biomass. A negative trend would represent an opposite
development.
For a given value of rainfall, a value of NDVI was obtained for each pixel and for
each year from individual regression of NDVI on rainfall. This obtained value of
NDVIpredis associated with the precipitation signal in the inter-annual changes of
NDVI. Deviations in NDVIobs from NDVIpred expressed in the regression residuals
were computed at pixel-by-pixel basis for every year. After that, the authors computed
trends in regression residuals over the study period. Pixels with statistically significant
trends within the precipitation-dependent zones were mapped as areas of
human-induced vegetation change. The map of human-induced change in vegetation
cover and results of area calculations is presented in Figure 10 (a). The map shows
areas of improvement and degradation of vegetation cover derived by the analysis of
regression residuals. Figure 10 (b) demonstrates the results of area measurements for
climate-induced and human-induced vegetation change in Central Asia and
Kazakhstan (in percent from the entire vegetated area). Generally, human impact
has a weak influence on the inter-annual trends of NDVI. The results show that, as a
whole, only about 5.75 percent of all vegetated territory is considered to undergo a
human-induced change of vegetation cover.
Validation of the model at the local scale
The results of the modelling have been tested at a number of sites exhibiting both
positive and negative trends of regression residuals. Example of an abandoned
Figure 10.
(a) Distribution of
vegetation change induced
by human impact and the
area of this impact
measured in percent from
the entire vegetated
area (b)
MEQ
19,5
590
agricultural area is presented in Figure 11. The corresponding regression between
NDVI and precipitation and time-series of residuals are displayed in Figure 12. After
arranging the regression residuals in accordance with the time, clear trend could be
observed. This area could be characterized by a positive trend of vegetation response
to precipitation and can be evaluated as area with improving vegetation cover caused
by diminishing human activity.
A number of the test sites revealed traces of human activities leading to degradation
of vegetation cover. It will be demonstrated on one example. This test site is located in
the southern part of the Moyunkum Sands in the southeast of Kazakhstan. Originally,
the area was covered by woody and grass vegetation, the Haloxylon-forest extends
from south-west to north-east as a broad stripe with a width from 10 to 25 km
(Figure 13, a). Because of a lower reflectance in blue spectrum, the areas occupied by
Haloxylon thickets occurs deep dark in the 1 channel of Landsat and can be easy
Figure 13.
Time series of Landsat
subsets (band 1) from a
test site in the Moyunkum
sands with clear signs of
land degradation
Figure 12.
(a) Regression between
mean growing season
NDVI and precipitation for
the test site shown in
Figure 11. (b) Time-series
of regression residuals
Figure 11.
Traces of land-use change
in satellite data at a test
site in the northern
Kazakshtan: (a) Landsat
MSS image from 1975, (b)
Landsat TM image from
1992, (c) Landsat ETM þ
image from 2001
Remote
monitoring
system
591
identified and mapped. On the Landsat image from 1990, many light areas, especially
in the western part, occurred within the contour of Haloxylon’s stripe (Figure 13, b). On
the third Landsat image shouted in year 2000, the light areas widespread over the
significant part of the image associated with Haloxylon occurrence (Figure 13, c). Field
surveys in this region in 2004 and 2005 as well as reports from local officials verified a
clear sign of massive destruction of vegetation cover caused by intensive wood felling.
Wide areas of the Haloxylon-forest are already completely cut out; at many places the
wood cutting continues intensive at the present time. As a result of wood cutting,
surfaces with only thin, rare vegetation cover composed by dwarf shrubs and
ephemerals were formed among overgrown sands on many places. When these areas
are grazed, they easy get trampled by livestock. After that, bare sand surfaces and
even mobile forms of sand such as barkhan chains and barkhan ridges appeared. The
destruction of vegetation cover, a decrease of biomass and higher rates of soil loss in
this region resulted in a drastic decrease of vegetation response to precipitation over
the period 1981-2000. That caused a negative trend of regression residuals (data not
shown).
Discussion
This study demonstrated a model for monitoring changes in vegetation activity and
detecting their driving forces. The theoretical background for the model separating the
driving forces of vegetation change was the concept of trends
synchronism/non-synchronism. This concept understands vegetation response to
precipitation as a constant if no change of external conditions occurs. The authors
considered that any processes of improvement or degradation in the areas of stable
response of vegetation cover are driven by climate factor alone. A change in the
response of vegetation to climate has been understood as an effective indicator for
human-induced change. Mathematically, this response is expressed in deviations of
observed values from the NDVI values predicted by the corresponding regression
model between NDVI and precipitation. These deviations were computed for every
year and each pixel. Any statistically significant trend of deviations found in a certain
pixel over the study period reflects a change in vegetation response and is considered
to be a representative sign for human-induced change of vegetation cover. Computed
trends of regression deviations help to detect areas with diminishing or increasing
vegetation response to rainfall. These areas were mapped as areas of human-induced
improvement or degradation of vegetation cover. Results of the modelling were
validated by test of statistical significance and by comparison with the data from the
remote sensing systems of fine resolution and fieldtrips to key field sites.
The model being applied to the study region found that most part of the trends in
NDVI observed in Central Asia during the period of 1982-2000 is explained by the
precipitation change. The inter-annual precipitation dynamics have driven about 75
percent and 82 percent of all positive and negative trends in NDVI respectively. That
amounts to 15.5 percent and 4.7 percent of the entire vegetated area of Central Asia.
The trends in vegetation activity driven by anthropogenic factor are much scarcer and
occupy only about 5.75 percent of the entire vegetated area. Thus, about 4.15 percent of
the entire vegetated area demonstrated the upward human-induced trend, while the
downward trend occurred in only 1.60 percent of the entire area. The results clearly
demonstrated that the area of the upward trend in vegetation activity significantly
MEQ
19,5
592
overwhelm the area of the corresponding downward trend. The reason for this seems
to be a rapid decrease of anthropogenic activity over the most area in Central Asia due
to the collapse of the Soviet Union in 1991 and a fundamental change in land-use
practice as well as favourable climate development over most part of the region. Some
relevant previous studies reported about significant increase in vegetation activity
over particular regions in Central Asia. These studies suspected diminishing
anthropogenic activity to be the major cause for it. De Beurs and Henebry (2004)
reported that the institutional changes in Central Asia positively affected land surface
phenology at spatial resolution relevant to meso-scale meteorological models.
Robinson et al. (2002) investigated land degradation and desertification in pasture
lands in the central Kazakhstan and found significant improvement of them associated
with the period from 1991 to 2000. Nonetheless, there also are contradict studies
reported about worsening of climate conditions over several parts of the formerly
soviet Central Asia that are unfavourable for vegetation growth.
No doubt, the ecosystems of Central Asia represent a gigantic reservoir of carbon,
and, on the one hand, they play a very considerable role in the global change, on the
other hand, they are very sensitive to the contemporary climate change. Both the
climatic conditions and physio-geographic patterns vary significantly within this
region. Climatic type, soil properties, vegetation type, and landforms show a high
variance within the region’s eco-geographic zones. Therefore, the authors have to
expect a high variance both in impact of the contemporary climate change and in the
change of human influence. The model presented in this paper revealed this variance
signifying spatial distribution of vegetation changes driven separately by climate and
by human influence. That is an advantage of this study over the previous related
studies from this region (De Beurs and Henebry, 2004).
However, each model is only a gross misspecification of reality and one or more
relevant variables are either omitted from the model or are represented by an incorrect
functional form. One of the improvements of the monitoring system presented in this
paper may be the incorporation of temperature into the model. Recent remote sensing
based studies on vegetation dynamics reported about significant role of the spring
temperature for increase in vegetation activity over high latitudes in North America
and Eurasia (Zhou et al., 2001). The region of Central Asia has experienced a warming
trend in order of 1-28C since the beginning of the twentieth century. This might have a
strong impact on the regional temperature and precipitation regimes and also on
vegetation cover. The authors based the monitoring system on the assumption that
Central Asia is a region where rainfall is the main climatic controlling factor (what has
been proved in the results of this study: about 75 percent of the entire territory
exhibited strong relationships between NDVI and rainfall). But the authors might
expect that in many areas temperature could also play an important role. It means that
some trends in NDVI could be driven either exclusively by temperature or by both
rainfall and temperature. An incorporation of temperature into the model and a
re-examination of climate and human-induced vegetation trends taking into account
this new variable can be objective of a further research. Another source of model’s
uncertainty is the use of 0.58 precipitation data. This spatial resolution of the origin
rainfall dataset is much larger than that of the used satellite data (8 km). The process of
conversion of the rainfall data leads to distortion of information and certainly
influences the end results of the model. The use of more progressive geostatistic’s
Remote
monitoring
system
593
techniques for data conversion or acquisition of a dataset with a spatial resolution
closer to 8 km would significantly reduce the uncertainty of the model.
Conclusions
The study developed a monitoring system for assessment degradation/improvement of
vegetation cover and discrimination of their two driving forces, climate change and
anthropogenic activity. The monitoring system bases on the concept of
synchrony/asynchrony in the trends of vegetation and climate, which has been
presented and thoroughly described in this paper. This monitoring system used the
temporal response of vegetation to precipitation as major indicator for degradation or
improvement of the vegetation cover. Trends in vegetation response over the study
period (1982-2003) have been considered to indicate positive or negative changes of
vegetation cover driven by human impact. Inter-annual time-series of the satellite
derived normalized difference vegetation index (NDVI) and precipitation served as
input data into the monitoring system, which has been successfully tested over the
region of formerly Russian Central Asia. The results obtained at the regional scale
were afterwards validated at the local scale using data from fieldwork and satellite
imagery of higher spatial resolution.
The authors suggest that the monitoring system developed in this study is a useful
tool for controlling the effects of rainfall in order to separate between human-induced
and climate-induced land degradation. The technique allows the monitoring trends in
vegetation cover driven by influences of other than climate. By doing so, it gives
valuable hints to potential areas submitted to human-induced changes of vegetation
cover. Once identified, they can be examined in more details with focus on their
vegetation cover status, the forces driving biomass production and the most suitable
rehabilitation measurements.
Examples from test sites showed that both negative and positive trends in
vegetation response can result from changes in human impact. Thus, a rapid
deforestation at the second test site results in a significant negative trend, while
abandonment of agricultural land at the first test site results in a strong increase of the
vegetation response. However, concerning human-induced change in vegetation cover,
the monitoring system can evidently identify areas where there has been a reduction or
increase of vegetation response to rainfall amount, but the exact reason of this change
can not be determined by this method alone. For identification of scrupulous reasons
(exact forms of human activities driving the change) for human-induced land
degradation or improvement of a certain area a closer investigation using
high-resolution remote sensing data and field work is needed. It is therefore
envisaged that the developed monitoring system would be a part of a multi-scale,
complex, monitoring program, where it can serve as a general tool to identify areas of
climate- and human-induced vegetation changes (hot spots). Field survey and analysis
of high-resolution satellite images can track aspects of anthropogenic land cover
change and are required, in order to provide validation data for the monitoring system
products.
The study demonstrates the importance of taking into account precipitation when
trying to analyse changes of vegetation cover in drylands with high inter-annual
rainfall variability. In precipitation-dependent regions such as Central Asia the
inter-annual change of vegetation cover is significantly predicted by the inter-annual
MEQ
19,5
594
rainfall dynamics. This signal could be used for discrimination between the
climate-induced vegetation change and the vegetation change triggered by other
factors, mainly by human impact.
References
Anyamba, A. and Tucker, C.J. (2005), “Analysis of Sahelian vegetation dynamics using NOAA
AVHRR NDVI data from 1981-2003”, Journal of Arid Environments, Vol. 63, pp. 596-614.
DeBeurs, K.M. and Henebry, G.M. (2004), “Land surface phenology, climatic variation, and
institutional change: analysing agricultural land cover change in Kazakhstan”, Remote
Sensing of Environment, Vol. 89, pp. 497-509.
Evans, J. and Geerken, R. (2004), “Discrimination between climate and humane-induced dryland
degradation”, Journal of Arid Environments, Vol. 57, pp. 535-54.
FAO/UNEP (1984), Provisional Methodology for Assessment and Mapping Desertification, United
Nations Environmental Programme, FAO/UNEP, Rome, p. 73.
Holben, B.N. (1986), “Characteristics of maximum-value composite images from temporal
AVHRR data”, International Journal of Remote Sensing, Vol. 7, pp. 1417-34.
Kogan, F.N. (1997), “Global drought watch from space”, Bulletin of the American Meteorological
Society, Vol. 78, pp. 621-36.
Li, B., Tao, S. and Dawson, R.W. (2002), “Relation between AVHRR NDVI and ecoclimatic
parameters in China”, International Journal of Remote Sensing, Vol. 23, pp. 989-99.
Li, J., Lewis, J., Rowland, J., Tappan, G. and Tieszen, L. (2004), “Evaluation of land performance in
Senegal using multi-temporal NDVI and rainfall series”, Journal of Arid Environments,
Vol. 59, pp. 463-80.
Los, S.O. (1993), “Calibration adjustment of the NOAA AVHRR normalized difference
vegetational index without resource to component channel 1 and 2 data”, International
Journal of Remote Sensing, Vol. 14, pp. 1907-17.
New, M., Hulme, M. and Jones, P.D. (2000), “Global monthly climatology for the twentieth
century”, Data Set, available at: www.daac.ornl.gov
Nicholson, S.E. and Farrar, T.J. (1994), “The influence of soil type on the relationships between
NDVI, rainfall and soil moisture in Semiarid Botswana. I. NDVI response to rainfall”,
Remote Sensing of Environment, Vol. 50, pp. 107-20.
Olsson, L., Eklundh, L. and Ardo, J. (2005), “A recent greening of the Sahel – trends, patterns and
potential causes”, Journal of Arid Envirinments, Vol. 63, pp. 556-66.
Richard, Y. and Poccard, I. (1998), “A statistical study of NDVI sensitivity to seasonal and
inter-annual rainfall variations in southern Africa”, International Journal of Remote
Sensing, Vol. 19, pp. 2907-20.
Robinson, S., Milner-Gulland, E.L. and Alimaev, I. (2002), “Rangeland degradation in Kazakhstan
during the Soviet-era: re-examining the evidence”, Journal of Arid Environments, Vol. 53,
pp. 419-39.
Symeonakis, E. and Drake, N. (2004), “Monitoring desertification and land degradation over
sub-Saharan Africa”, International Journal of Remote Sensing, Vol. 25, pp. 573-92.
UNCOD (1977), Desertification: Its Causes and Consequences, Pergamon Press, Oxford.
Veron, S.R., Paruelo, J.M. and Osterheldt, M. (2006), “Assessing desertification”, Journal of Arid
Environments, Vol. 66, pp. 751-63.
Remote
monitoring
system
595
Wang, J., Rich, P.M. and Price, K.P. (2003), “Temporal responses of NDVI to precipitation and
temperature in the central Great Plains, USA”, International Journal of Remote Sensing,
Vol. 24, pp. 2345-64.
Weiss, E., Marsch, S.E. and Pfirman, E.S. (2001), “Application of NOAA-AVHRR NDVI
time-series data to assess changes in Saudi-Arabia’s rangelands”, International Journal of
Remote Sensing, Vol. 22, pp. 1005-27.
Wessels, K.J., Prince, S.D., Frost, P.E. and Van Zyl, D. (2004), “Assessing the effects of
human-induced land degradation in the former homelands of northern South Africa with a
1-km AVHRR NDVI time-series”, Remote Sensing of Environment, Vol. 91, pp. 47-67.
Xiao, J. and Moody, A. (2004), “Trends in vegetation activity and their climatic correlates: chin
1982 to 1998”, International Journal of Remote Sensing, Vol. 20, pp. 5669-89.
Yang, L., Wylie, B., Tieszen, L.L. and Reed, B.C. (1998), “An analysis of relationships among
climate forcing and time-integrated NDVI of grasslands over the US Northern and Central
Great Plains”, Remote Sensing of the Environment, Vol. 65, pp. 25-37.
Zhou, L., Tucker, C.J., Kaufmann, R., Slayback, D., Shabanov, N. and Myneni, R.B. (2006),
“Variations in northern vegetation activity inferred from satellite data of vegetation index
during 1981 to 1992”, Journal of Geophysical Research, Vol. 106, pp. 9-2001.
About the authors
P.A. Propastin is currently a Postdoctoral Research Associate in the Department of Cartography,
GIS and Remote Sensing at the Georg-August University, Go¨ttingen, Germany. His main
research activities focus on the integration of remote sensing and GIS in studies on climate
dynamics and their impacts on the vegetation cover in Asia. P.A. Propastin is the corresponding
author and can be contacted at:ppropas@uni-goettingen.de
M. Kappas is Head of the Department of Cartography, GIS and Remote Sensing,
Georg-August University of Go¨ttingen, Germany. His current research interest is in physical
geography and the quantitative exploitation of satellite remote sensing data for the
characterization of terrestrial surface properties.
N.R. Muratova is Head of the Department of Remote Sensing and Digital Image Analysis at
the Space Research Institute of the Kazakh Science Academy, Almaty, Kazakhstan. Her scientific
interests focus on the remote sensing of vegetation, the diagnostics and evaluation of
satellite-derived surface data, and climate-vegetation interactions.
MEQ
19,5
596
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints