What has been written about the 2007-2009 drought in Iraq and why there is a need for another drought assessment

I’ve found two academic articles and a few different UN/IOM reports covering the drought in 2007-2009 in Iraq. I thought it would be a good exercise to summarize what is known about this drought in Iraqi Kurdistan, and to argue for the added value of another drought assessment – which is what I’m working on right now, due to a small refocus of the article on drought and migration (more drought, less migration).

Trigo et al (2010) assessed the drought in the whole Fertile Crescent region using the following data:

Data Source Spatial Resolution
Precipitation GPCC dataset 1 degree
Geopotential height, temperature, humidity and zonal and meridional wind components at different pressure levels NCEP/NCAR Reanalysis 2.5 degrees
Normalized Difference Vegetation Index SPOT VEGETATION 0.008928 degrees (about 1 km2 at the equator)
Water level for Lake Thartar Satellite radar altimetry from Topex/Poseidon, Jason-1 and Jason-2

The GPCC dataset were used to visualize the spatial extent of the drought. Trigo et al (2010) found a striking decline in precipitation, and that the decline was most severe in the first hydrological year, October 2007 to May 2008.

The use of GPCC in Iraq could, however, be risky, since it’s based on interpolations of station data, and there are very few,  usually no, stations in Iraq. As an example, there were no stations in Iraq in 2008 when the data measured a decline in precipitation of around 70-80% in both years.

GaugeStationsME2008 Scale

Atmospheric circulation variables from the NCEP/NCAR Reanalysis project were used to provide an explanation of the processes causing the drought. They found that the wet months, during the drought years, were dominated by high pressures that inhibited synoptic activity to enter from the Mediterranean. Furthermore, masses of dry air were advected from Iran and Russia causing dry weather. Convective instability was also weaker than normal, due to the dry conditions.

The analysis of vegetation conditions during the drought showed strong anomalies in April 2008, especially in the northern parts of the Fertile Crescent area. The first drought year also showed several areas with up to six consecutive months of vegetation stress. One of these areas were northern Iraq, but also eastern Syria, south-eastern Turkey and western Iran were affected. Trigo et al (2010) also presented a comparison of the latest drought to the 1998-2000 drought, which showed that the intensity and extent of the droughts were similar.

The authors also used the Global Land Cover 2000 (GLC2000) database to assess how much of the vegetation that were agricultural vegetation. This is of course very important, since agricultural drought has severe economic consequences, while a drought in unused lands might not be as severe on the economy. In 2008, 20% of cultivated land were affected by drought for at least 5 months. In 2009, the corresponding number was 9%. The majority of the pixels (56 and 69% respectively) belonged to the land cover sparse land.

The GLC2000 is a useful dataset, but has limitations. Temporally, it only covers 2000, and the land cover changes that have happened since then are neglected. Furthermore, GLC2000 is based on SPOT VEGETATION data (1 km2 resolution?) and “expert knowledge” of which the first is a coarse and limited data source, and the latter can be very subjective. A visual interpretation, based on my own (subjective) knowledge tells me that the GLC2000 is not very accurate for the Duhok Governorate in 2011-2013. I’m not saying that the GLC2000 is completely wrong, it definitely is not, but it’s very important to be careful about relying too much on, or drawing too strong conclusions based on, outdated or inaccurate data.

In addition to the land use data, national cereal production data for Syria, Iraq and Iran are used to show the notable dip in production in wheat and barley in 2008.

Lake Thartar was found to have a strong decline in height 1999-2001, and 2008-2009 and these results may indicate a hydrological drought, where water in reservoirs (both surface water and subterranean water) were affected by drought.

All in all, Trigo et al (2010) provide a thorough, large extent, drought assessment based on existing data, which is not perfect, but still give an idea of the characteristics of the 2007-2009 drought. The use of multiple datasets strengthens the analysis and the conclusions that there was a severe drought in 2007-2009, affecting both hydrology, vegetation and agricultural production. Furthermore they provide an explanation for the atmospheric processes behind the drought. Such a drought assessment is valuable and points to where it would be interesting to conduct a more thorough analysis of drought, incorporating more detailed data and information.

A “more detailed” drought assessment of the Erbil governorate, bordering the Duhok governorate, was presented by Fadhil in 2011. The study presents a reduction of about 70% in total annual precipitation between 2007 and 2008, based on data from the meteorological station in Erbil. No information about the number of stations (or is it only one?), their locations and their temporal properties are provided.

The satellite based drought analysis included two Landsat images (Path: 169/Row: 35) captured in June 2007 and 2008. This is where the big problems begin. First of all, Landsat provides snapshots of the Earth’s surface, which may be influenced by for example clouds or other short term phenomena. The data were used in five vegetation, soil, and water indices who were briefly described. NDVI was used to show the decline in vegetation during the drought year. The results (a comparison between June 23 2007 and June 9 2008) show that large parts of the study area had a negative change, and few had positive change. Looking at a table showing the areas with negative and positive NDVI (change in NDVI I assume, otherwise we would be talking about surface water), only 15% of the total area had a negative change in 2008, while 84% had no change (the definitions of no change being unclear). Further indices used where the Normalized Difference Water Index (NDWI), the Bare Soil Index (BSI), the Tasseled Cap Transformation Wetness (TCW) and the Land Surface Temperature (LST). They also show a clear decrease in soil/vegetation moisture during the study period (which is less than a year). As an example, NDWI were used to show the change in the extent of the Dokan Lake, which were 32.5% smaller in 2008, than the previous year. This is, however, not an indicator of a prolonged period of drought, but rather a short term change that could be due to a heavy rainfall in 2007, just before the image acquisition date. And now you might ask how common rainfall in summer is, but I tell you it has happened. July 2011, travelling in the mountains, we had a thunderstorm. A better measure of the lake levels and vegetation would be to look at a longer time perspective and compare the drought years (2007 was also a drought year and not good as a reference year) to the normal years. What I’m wondering is, why didn’t he turn to MODIS for a similar analysis?

This small “review of academic literature” on the drought shows a need for a drought assessment more detailed than Trigo et al (2010), using different meteorological, vegetation, and socio-economic data, but with a larger temporal extent than Fadhil (2010), using a longer period average as a reference, to compare the drought years to. This is what I’ve been working on for the past few weeks, in addition to a lot of teaching GIS and basic remote sensing (of drought).

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s