The majority of Sudan is characterized as semi-arid region and thus susceptible
to degradation or even desertification; semi-arid regions are subject to regular
seasonal dryness and large inter-annual variability in precipitation. This results
in variable vegetation cover on annual and inter-annual timescales, as both
natural ecosystems and non-irrigated crops rely on soil moisture derived from
seasonal rains or springtime snow melt (Evans and Geerken, 2004; Weiss et
al., 2004). Figure 1a shows average annual rainfall in
Climate-induced variability in semi-arid vegetation is a matter of both ecological
interest and economic concern, as strong sensitivity to climate can result in
rapid land use change (Vanacker et al., 2005) and vulnerability to human-induced
degradation (Evans and Geerken, 2004). Climate is one of the most important
factors affecting vegetation condition. Therefore, evaluation of the quantitative
relationship between vegetation patterns and climate is an important object
of applications of remote sensing at regional and global scales. The Normalized
Difference Vegetation Index (NDVI) is established to be highly correlated to
green-leaf density and can be viewed as a proxy for above-ground biomass (Tucker
and Sellers, 1986).
The causes of variance of relationship between NDVI and its explanatory variables are known to be spatial variations in properties such as vegetation type, soil type, soil moisture (Ji and Peters, 2004; Foody, 2005). Vegetation cover processes play a crucial role in the water balance over a wide range of spatio-temporal scales (Betts et al., 1996). Unfortunately, vegetation dynamics and their interaction with climate are still largely unexplored. Under this framework, satellite data have proved to be very useful for collecting realistic data about land use change and vegetation trends from local to global scale (IGBP-IHDP, 1999). In particular, NOAA-AVHRR (Advanced Very High Resolution Radiometer, onboard National Oceanic and Atmospheric Administration satellites) data can provide useful information on such changes over climatic spatio-temporal scales. The long time series of observations can be very useful for studying vegetation dynamics over inter-annual scales.
Vegetation is one of the most important parameters for human environment assessment and monitoring due to their specific role in geo-sphere, biosphere and atmosphere interactions and plays an important role in global climate change. The vegetation amount controls the partitioning of incoming solar energy in the sensible and latent heat fluxes and consequently changes in vegetation amount will result in long term changes in the global and local climate, which will in turn affect the vegetation growth as a feedback. Vegetation has special characteristics due to its distinct annual and seasonal changes it is a sensitive indicator on the study of global and local environment change caused by climate or human activities. Thus it comes as no surprise that the detection and quantitative assessment of green vegetation is one of the major applications of remote sensing.
In this decade, human beings consequently realized the significance of global change monitoring, several international organizations such as IGBP, HDP and WCP, have lunched very important programs, among which land cover and vegetation change monitoring is a key project. The method for studying land use and vegetation change is developed very quickly as the progress of remote sensing technique in the world.
This study aim to employ GIS to examine the relationship between rainfall and the normalized difference vegetation Index (NDVI) in Sudan, during the time period from 1982 to 1993 and the value of NDVI is taken as a tool for drought monitoring. The objective of this work is to examine whether there is a relationship between rainfall and NDVI in Sudan. Once a positive relationship is established, the work analyses the use of NDVI as a proxy indicator for the occurrence of meteorological drought, that is, when precipitation is significantly below what is normally required by vegetation. This is done by integrating multi-source geo-referenced datasets in a GIS platform in order to facilitate analysis and the generation of cartographic, statistical and modeling products. The final output comprises the spatial analysis products and aims to be useful in the decision making process for drought monitoring and to avert its consequences on lives and livelihoods.
DATA AND ANALYSIS METHODS
The AVHRR satellite, with its 12 year data record (1982-1993) and reasonably spatial resolution (8 km), provides an excellent tool for the analysis of regional vegetation. AVHRR 15 day composites of surface reflectance and maximum NDVI were downloaded from World Meteorological Organization (WMO) website in this study.
We removed noisy pixel areas characterized by exceptionally low NDVI values
relatively to their pixel neighborhood. This pixels represented large cloud
areas and were replaced by a mean value calculated from the temporal neighboring
NDVI layers. The 15 day NDVI composites were integrated to mean monthly and
then to mean growing season values for each of the analysis years.
||(a) Average annual rainfall in Sudan (1982-1993) and (b) mean
NDVI in Sudan (1982-993)
||(a) Average annual rainfall in growing season (JASO) in Sudan
(1982-1993) (b) mean NDVI in growing season (JASO) in Sudan (1982-1993)
Analysis of seasonal and inter-annual vegetation dynamics and trends of Sudan
region is based the normalized difference vegetation index (NDVI). This index
is calculated from AVHRR measurements in the visible and infrared bands as follows:
Where, ρr and ρnir are the surface reflectance`s
in the 550-700 nm (visible) and 730-1000 nm (infrared) regions of the electromagnetic
The foundation for using NDVI data in monitoring arid and semi-arid lands was
based on a large body of research in 1980s in a wide range of arid land regions,
which demonstrated a close relationship between NDVI and rainfall variations
on seasonal to inter-annual time scales (Tucker and Nicholson, 1999). This relationship
between NDVI and rainfall provided the basis for using time series NDVI data
for drought monitoring and development of famine early warning systems in regions
with sparse terrestrial rainfall networks (Hutchinson, 1991). The compiled 12
year time series is now exploited to examine the linkages between climate variations
and ecosystem dynamics (Lotsch et al., 2003) and more recently to study
long-term trends in vegetation (Eklundh and Olsson, 2003; Slayback et al.,
2003). For this study we subset the Sudan country from the continental data
set for the period 1982-1993. Figure 1b shows an example of
the average of all data for complete years from 1982 to 1993 showing the long-term
mean in Sudan. Since the evolution of NDVI in Sudan is closely related to rainfall
seasonality, the analysis in this paper only focuses on NDVI patterns during
the growing season. The growing season was defined by examining the long-term
mean patterns of NDVI. The months of July through October, referred to here
as JASO, were selected to represent the average start and end of the growing
season. In order to reduce the amount of data to be examined, we created a long-term
NDVI climatology by averaging data for all cloud free pixels for July-October
months from 1982-1993. The year to year variability in the NDVI patterns was
examined by calculating yearly JASO anomalies as follows:
||The respective JASO percent anomalies.
||Individual seasonal JASO means.
||The long-term JASO mean (Fig. 2b) and Fig.
2a shows Average Annual Rainfall in Growing Season (JASO) in Sudan in
the time period of 1982-1993.
RESULTS AND DISCUSSION
The spatial NDVI anomaly patterns in Sudan are shown in Fig. 3.
These series of images show the JASO percent NDVI anomaly patterns for selected
growing seasons during the 1982-1993 periods. These series of NDVI anomalies
shows the spatial coherence and temporal persistence of drought conditions during
the 1980s, a noted feature of the persistence in rainfall departures throughout
the recorded climate history of the region (Rasmusson, 1988). In 1982 and 1983,
a patchy pattern of below normal NDVI showed the prevalence of drought conditions
across the country, especially in the western and eastern areas and most of
the pronounced greenness was concentrated in the middle of Sudan. This pattern
was enhanced in 1984, 1990 and 1991 and the whole region showed a low NDVI level
of below normal conditions and the most extreme negative departures reached
80% lower than the normal conditions. These magnitudes of negative departures
agreed with rainfall departure patterns for the region shown by Nicholson (1985).
In 1985, it still showed negative departures in NDVI ranging between 10 and
40%. Region-wide drought conditions returned in the growing season in 1987 and
the negative departures in NDVI were on the order of 10-60%, which were concentrated
from the east to the west in the central Sudan, although it showed normal to
above normal vegetation conditions in the most part of south-east Sudan. During
the growing season in 1988 and 1992, the whole region showed a high NDVI level
of above normal vegetation conditions and had positive anomalies ranging between
20 and 100%. The above normal greening in 1988 was associated with positive
rainfall anomalies during the months of August, September and October (Nicholson
et al., 1996). Presented in Fig. 4 Monthly NDVI and
mean monthly rainfall in Sudan in 1988.
||NDVI anomaly patterns in the growing season (JASO) during
the time period of 1982-1993
||Mean monthly NDVI and monthly rainfall in Sudan during (1988)
Monthly Time Series Patterns in the Time Period from 1982 to 1993
The monthly time series of NDVI and rainfall from 1982 to 1993 in Sudan are
shown in Fig. 5 and 6. On average most of
the rainfall occurs between July and October, with a maximum in August. Approximately
83% of the annual rainfall falls between the July and October (Lamb, 1980),
so averaging NDVI data for these months fairly represents the growing season
for the region. From Fig. 7, monthly relationship of rainfall
and NDVI for that time series in Sudan showed that the NDVI values had a correlation
(0.598) as a linear relation to the rainfall,, Both rainfall and NDVI show a
maximum in August-September.
Mean monthly NDVI ranged from 0.2 to 0.35 across the country throughout the
time series in Fig. 8 and the NDVI values greater than 0.2
corresponded well with the rainy season from July to October. These high NDVI
values persisted towards the end of year in November and December indicating
the lagged response of vegetation to rainfall in this region (Nicholson et
al., 1990). The extent of these values across the region is an indicator
of rainfall conditions.
The low NDVI values lower than 0.3 were distributed from the east to the west in the central Sudan in 1984, 1990 and 1991, indicating the prevalence of drought conditions. The extent and duration of these NDVI values can be used as an indicator of the strength and duration of the rainfall to produce mechanisms associated with the ITCZ (Intertropical Convergence Zone) since vegetation growth in the region is primarily controlled by rainfall, although other factors including potential evaporation influence the fluctuating boundary (Milich and Weiss, 2000).
During the growing season in 1988 there was an easing of the severe drought conditions with above normal NDVI, following a good rainy season in Sudan. From 1989 to 1993 it showed low NDVI values of below normal NDVI except 1992. This is a western extension of the drought that affected eastern Africa in 1991-1993.
The persistence and spatial coherence of drought conditions during the 1980s
is well represented by the NDVI anomaly patterns and corresponds with the documented
rainfall anomalies in Sudan during the time period from 1982 to 1991. The time
series was dominated by low NDVI values of below normal conditions, with 60%
of the years showing below normal NDVI conditions and the severest departures
in NDVI occurred in 1984, which situation persisted for 6 years in the time
series, with exceptions in years of 1983, 1985, 1986 and 1988. These patterns
are shown in Table 1 and Fig. 9. Between
1987 and 1993, 71% of the years showed low NDVI values of below normal vegetation
conditions, but it showed high NDVI values of above the long-term mean in 1988
and 1992 (Fig. 9).
||Mean monthly NDVI in Sudan (1982-1993)
||Mean monthly rainfall in Sudan (1982-1993)
||Scatter plot of the rainfall and NDVI in Sudan (1982-1993)
||Average NDVI in the time period of July-October in Sudan (1982-1993)
||NDVI anomaly scores (+/-) showing persistence patterns of
above normal or below normal vegetation condition
||NDVI Anomaly index (NAI) in Sudan (1982-1993)
The persistent nature of these patterns in NDVI in the time series of 1982-1991
is in agreement with the historical patterns of rainfall anomalies in the region.
The NDVI time series data indicated that there was a gradual and slow but persistent
recovery from the peak drought conditions that affected the region in the early
to mid-1980s. It was corroborated by the decrease in the magnitudes of negative
rainfall departures from 1984 to 2000 (Nicholson, 2001). The correlation between
the NDVI and rainfall anomaly time series for the 1981-2000 period is positive
and significant (R = 0.78) indicating the close coupling between rainfall and
land surface response patterns over the region. The large scale and coherent
changes in anomaly patterns between 1984 and 1993, a difference of 10 years
might suggest some large-scale climatic influence on Sudan vegetation dynamics.
In the process it has been demonstrated that in regions such as Sahelian Africa,
where there is a dearth of digital data from which useful monitoring and management
information can be drawn, GIS using remotely sensed data obtained from satellites
is technically feasible. Furthermore, it is a relatively low cost system, as
it uses free data for input and can be run on an ordinary desktop computer.
The training required for running the system is also limited since once set,
it uses pre-processed data inputs. The analysis has shown that NDVI is a complex
indicator, difficult to interpret, as well as being a delayed outcome indicator.
NDVI is a crude indicator of drought risk and needs to be related to other socio-economic
and bio-physical data in order to be useful. The precision of NDVI as a vegetation
index also needs to be strengthened through establishing its relationship to
the growing season, for each specific climatic zone, on the basis of local vegetation
and crop types. The GIS can help in understanding and analyzing complex environmental
situations, such as the Sudan, even where data is relatively scarce and there
is a limited knowledge base. The GIS tools used, however, have highlighted
the shortcomings of the data sets and the method. For any system for monitoring
environmental change, the objectives need to be specific. In particular, it
should be clear whether the aim is to monitor environmental change across years,
in which case a long time-series data is required, or whether it is to monitor
vegetation and crop changes within seasons for purposes of drought warning.
An effective drought warning system using NDVI should take advantage of remote
sensing sources in using real time data, in order to facilitate timely decision
making. If this were to be done, NDVI can be a valuable first cut indicator
and provide a key input for cost-effective, reliable and timely drought monitoring
systems. The patchy nature of the increase in NDVI will require the use of higher
spatial resolution data from LANDSAT, SPOT and MODIS in order to determine the
driving factors of change at the landscape scale. Further studies examining
combined climate data including rainfall and sea surface temperature patterns
and continued gathering of long-term satellite data sets will help in understanding
the long-term changes in the climate and land surface conditions of this sensitive