Aghanistan geospatial analysis

I'm currently taking a graduate-level course on ArcGIS and geospatial analysis. As part of this course, I will be required to do a project on a topic of my choosing. Afghanistan will be my focus and rather than do something simply for a grade, I would like to help with a real-world problem in Afghanistan. So this is a call to the council for ideas/suggestions on a topic for my project. If you're in Afghanistan and there is some question you have that geospatial analysis can help answer, please PM me or post here.

Of course, since this is for a class, I have some significant LIMFACS:

1. Time: The course ends in late August, so I must have the project done by then. Additionally, if you need something by next week, I probably can't help you there either.

2. Classification: I have to use unclassified data, though FOUO might be possible.

3. Data: I need to be able to get the relevant data to answer the question. I don't really care what format it's in - I can process or format it as needed. There is a lot of data out there, but much of it is proprietary or held by NGO and other organizations and not directly published. Some of this I can acquire, some I cannot. If you have the data, that would be ideal.

4. Scope: It will just be me working on this project, so please bear that in mind.

Any and all suggestions and comments are welcome! Thanks!

In addition to posting this here, you might want to also contact the geospatial analysis team at the Open Source Center with this same information. They do great work and no doubt have a question/requirement that they could use you on. Everything they do is with open source.

Quote:

---

Originally Posted by Entropy

I'm currently taking a graduate-level course on ArcGIS and geospatial analysis. As part of this course, I will be required to do a project on a topic of my choosing. Afghanistan will be my focus and rather than do something simply for a grade, I would like to help with a real-world problem in Afghanistan. So this is a call to the council for ideas/suggestions on a topic for my project. If you're in Afghanistan and there is some question you have that geospatial analysis can help answer, please PM me or post here.

Of course, since this is for a class, I have some significant LIMFACS:

1. Time: The course ends in late August, so I must have the project done by then. Additionally, if you need something by next week, I probably can't help you there either.

2. Classification: I have to use unclassified data, though FOUO might be possible.

3. Data: I need to be able to get the relevant data to answer the question. I don't really care what format it's in - I can process or format it as needed. There is a lot of data out there, but much of it is proprietary or held by NGO and other organizations and not directly published. Some of this I can acquire, some I cannot. If you have the data, that would be ideal.

4. Scope: It will just be me working on this project, so please bear that in mind.

Any and all suggestions and comments are welcome! Thanks!

---

GeoInt, USGS, NOAA, etc for more depth in your project. Have fun

Entropy,

Coming from the Civil Affairs side of things Agriculture and Water are enduring issues with readily available data and may be of interest to you in this project. The Agricultural Engineering Department of your local university may be a good place for face-to-face help. One of my grad classes was on Evapotranspiration and we spent some time on developing Penman-Monteith and Priestly-Taylor evapotranspiration models. It was an interesting class and exercised database skills as well as modeling skills. Unfortunately, we did not have a GIS portion of the class but then I had my hands full with just the modeling and database portions. :eek:

Here are some links that might be of interest to you on this topic:

Simulation of winter wheat evapotranspiration in Texas and Henan using three models of differing complexity

Quote:

---

Crop evapotranspiration (ET) is an important component of simulation models with many practical applications related to the efficient management of crop water supply. The algorithms used by models to simulate ET are of various complexity and robustness, and often have to be modified for particular environments. We used three crop models, CROPWAT, MODWht and CERES-Wheat, to simulate ET of winter wheat (Triticum aestivum L.) grown at Bushland, Texas, and Zhengzhou, China, during multiple growing seasons in which ET was measured using lysimeter or soil water balance techniques

---

Comparison of methods used to estimate lake evaporation for a water budget of lake Seminole southwestern Georgia and Northwestern Florida

Seems to have a GIS portion...


Crop evapotranspiration - Guidelines for computing crop water requirements - FAO Irrigation and drainage paper 56


One of my class references....



Regards,

Steve

Thanks for all the replies and PM's I've received so far. I'm sorry I haven't replied yet - hit a deer on the way home this weekend at oh-dark-thirty, which turned into a bit of an adventure.

the guys at globalsecurity.com may be able to help

If you're just dropping by to distribute links to your Blog, please don't. If you do wish to comment, please go here (LINK), scroll up and see the introductions of others then add on a comment and tell as a little about yourself.

Ken,
I assumed that you had the signature option on the board just like any other board. I can't post comments?

Quote:

---

Originally Posted by Mark Eichenlaub

I assumed that you had the signature option on the board just like any other board. I can't post comments?

---

and your sig option is okay -- it's just that occasionally, people drop by and post a one-liner or a cryptic sentence and drop a link -- and do little else.

If you're going to comment, do remember it's not the best venue for communicating and it's better to be overly verbose than so concise that others aren't sure what was meant. Comment merrily away...

Quote:

---

Originally Posted by Ken White

and your sig option is okay -- it's just that occasionally, people drop by and post a one-liner or a cryptic sentence and drop a link -- and do little else.

If you're going to comment, do remember it's not the best venue for communicating and it's better to be overly verbose than so concise that others aren't sure what was meant. Comment merrily away...

---

Ok, thanks for the tips Ken.

I finished my project and am posting some of the results below.

To begin with, I would like to thank John McCreary of AFCEA for providing data for this project.

Before getting into the maps themselves, I need to discuss the data and, specifically, the limitations of that data:

1. The data is exclusively from open sources, primarily press reporting which is subject to collection bias, inaccuracy, probable undercounting and a variety of other errors. The data is for 2007 and 2008 only. I did not have access to data outside this 24 month period.

2. The location of incidents was not always specific or readily apparent, most often because the original reports had errors or did not give specific locations. It was not possible to determine the district in which an incident occurred for approximately 23% of the data – a statistically significant number. The district maps (included below) are therefore undercounting events by about a quarter. (Note: Total # of reported clashes at the provincial level for 2007-2008 is 4,633. Of those, 3,576 are included in the district-level data which was used to generate the maps below.)

3. Afghanistan has undergone one extensive administrative re-division of its province and district boundaries and is in the process of undergoing another. Some province and districts split, others merged or ceded territory as a result of these reforms. For incidents that occurred in or near those boundaries, it is often difficult to determine which model the original report was based on. For example, Uruzgan province was divided in two – the southern third remained Uruzgan and the remainder became Daykundi province. A report, even one today, of an attack in “Uruzgan” province could have taken place in Daykundi because not everyone is aware of the change, particularly the Western media, and others, for a variety of reasons (tradition, politics, etc.) may not acknowledge the boundary change in reports. To simplify this complex problem these maps use the older 32 province/329 district model. Standardization on this model caused some problems during transcription as some of the reporting utilized the newer model. In most cases, I was able to manually place the incident in the correct district or province, but in about two dozen cases at the district level this was not possible. Those incidents were removed from the district data.

4. The population and demographic data (ethnicity) are estimates from the Afghan Central Statistics Office (CSO) and vary in currency from last year back to 2002. Ethnolinguistic group data is primarily based on studies done prior to the beginning of the Soviet intervention in 1979 as well as estimates by the CSO based on interviews with local leaders, NGO studies, and other reporting. These estimates are subject to substantial error.

Because of these errors, I think one shouldn't look at the actual number of attacks per district, but the relationship between districts and where violence is concentrated.


Discussion:

Although not shown on the maps attached below, the data show a clear increase in violence in 2008 from 2007 in all areas, particularly IED attacks. There were a total of 1,704 reported attacks of all kinds in 2007. In 2008 there were 2,929 attacks of all kinds, an increase of almost 75%. IED attacks increased by 150%, from 239 to 596. Suicide bombings increase at a relatively smaller rate of 30%, from 124 to 162.

Map 1: This maps shows relative violence overlayed onto district ethnicity. It's important to note this shows which ethnic group holds a plurality, not a majority, since not every district has a majority ethnic group (this is actually one goal for the various district reorganizations - to provide each district with an ethnic majority). It shouldn't be a surprise that violence correlates strongly to the Pashtun ethnic group. All the districts with significant violence that do not have a Pashtun plurality have large Pashtun minorities. The exception is the Nuristani areas.

Map 2: This is a comparison of violence in Afghanistan for 2007 & 2008 by district. Again, no huge surprises, but it's interesting to note that levels of violence can vary widely over a single province.

Map 3: This map takes the violence from Map 2 and adjusts it on a per capita basis (based on the 2005 CSO population estimates), so the most attacks per capita are shown in red. The effect is that rural, less populated districts with significant violence are highlighted. Notice Kamdesh district in Nuristan province where last weekend's attack on FOB Keating took place. It has one of the highest violence per-capita ratios in Afghanistan.

I have some more maps from the project, but these three are the most interesting. I have all the data in ArcGIS geodatabase format that I'm willing to share for anyone who wants to PM me. I also have much of the data in excel format for anyone that's interested. I can also generate more maps if there's a specific request - at least for the next few weeks before my ArcGIS license expires.

and a great job, thanks.

Entropy,

Really nice job there :D! Just out of interest, is your dataset available?

Cheers,

Marc

Thanks Ken and Marc!

Marc, I have the data in excel or ArcGIS geodatabase. Send me a pm with your email and I'll send what I have to you.

Kivlonic:

Kudos.

FYI: One thing I learned from Iraq, as you saw in Afghanistan, is that sub-district and district boundaries, unlike in the US, are dynamic factors that must be tracked, updated in real time to have relevance. Actually, as with Kurdish areas in Iraq, the boundary changes themselves are typically leading or trailing factors of conflict---either creating them as one party jockeys for control over the area of another, or trailing in that a newly resettled majority has now gained control, and wants to affiliate with a different political structure (leaving the original remaining residents in a lurch).

One of the perennial problems (which you are not able to solve withoin your assignment) is that each change in districts actually represents a substantial change in demographics. In most cases, these census figures are based (and reported) on an underlying census block basis, so moving these blocks from one district to another moves the pop figures around.

The "disputed" district of Mandeli (Diyala) in Iraq was a circumstance where there were layers of depopulation/destruction as Kurds were pounded by successive attacks by Sadaam, then Iran, then Sadaam again. By 2009, you have a town which was once a large and important provincial district seat where, now, virtually everything has been destroyed, and only a few Kurdish stragglers exist or have returned. Questions: What data do you use to determine who belongs to this place? Who should be there? Who should control it (Erbil or Baghdad?)? How accurate and valid is the data?

Bear in mind, too, that once a change is made, all the place and town names can change, too. So dynamic administrative/political boundaries are reflected by the multiple alternative names given to places. Before the federal highway system (and Euro Counterparts) standardized names in the West, we did the same as them---used descriptive placenaming: Shiek Abdullah's place is only Shiek Abdullah's place if he lives there; once he moves somewhere else, the name follows him, and his old place is called Peshmerga HQ. Dynamic placenaming, and name tracking for all the alternative names is another big universe beyond your project.

One of the perennial, but unrecognized problem is conflicting administrative boundaries. Just because a province or district changes it's boundaries, what boundaries and "service areas" are used by national ministries, or other regional service providers? In Iraq, US recon folks couldn't understand a lot of what the national ministries were doing because, in part, they were looking at them from the provincial level. National roads (as opposed to provincial, district or local roads), and many other national-level services were managed based on a regional "service area" basis, or, in most cases, based on older provincial boundaries (they never changed them). Very confusing stuff if you don't realize these things as dynamic factors.

The US has a system for event reporting. In many cases, I found that there were substantial discrepancies between the event and the coded location. So your stuff is as good as it gets sometimes.

Great work.

Steve

PS:

I have never seen an accurate map of Iraq's provinces, districts and sub-districts in any open source, and seldom in any government source.

Any time you see a map of Ninewa's boundaries, and it does not show the district of Mahkmour as part of Ninewa as it really is (the subject of dispute), you are looking at a bad map.

Somebody on this site says: you have to keep learning.

Steve

Steve,

All great points, thank you!

The most frustrating part of this project was getting the data actually. Afghanistan is kind of a black hole and there is very little non-anecdotal public information. The Afghan CSO tracks some things (in fact, I got the provincial/district boundary shapefiles from their website), but they only provide information to NGO's and then only for a fee. You may be interested in looking over their site and contacting them as they are working on many of the issues you discussed in your comment.

There is also the problem of politically-skewed data. There are a couple of ethnic maps on wikipedia, for example, that appear to be biased to favor one group in certain areas. It's hard to refute stuff like that given so much about Afghanistan is prefaced by "estimate." Anything about Afghanistan needs to be double and triple checked. It reminds me, actually, of the research I did on the Balkans back in the 1990's prior to a deployment.

All in all it's a very tough nut. I had ambitious plans for this class project but was forced to scale back because of a lack of data. When very basic information on Afghanistan is difficult to come by, one wonders just how deep our ignorance of the country grows. Pretty damn deep I suspect.

I put together a powerpoint slide of Ninewa from 1950s to today, showing the many provincial, district, and subdistrict changes.

One slide with about 10 major layers. Te;lls the whole story in a minute.

It is consolidated from a huge amount of field research on the ground, and through provincial and ministerial governments. Collecting maps. Legislative enactments, governmental service boundary maps, etc... Dahuk (KRG) carved out in the north, Shirqat transferred to Salah ad Din (formerly Baghdad Province). The abolition of desert provinces, which brought Hatra and Ba'aj districts in. The disection of Irbil, which shifted Mahkmour to Ninewa, and other districts to Tam'im. Then, the internal re-disections of districts and sub-districts. Just in one province. Incredible.

I learned a while ago not to cast pearls before swine. NGA is great for geo-physical analysis, but his no capacity to provide geographic support for political/administrative boundaries, census data, etc...the non-combat stuff. I gave up long ago trying to explain how pop displacement tracking might help them know something about the place and people, or how, regardless of what system is used, land records and cadestral info tell as much as anthropology (people always record land transactions---even the bad guys).

(I would say non-mil stuff, but the military is so heavily involved in things civilian that the civ/mil distinction is one without a difference).

Send me an email if you would like a copy of the Ninewa ppt, or an accurate admin/pol map of Northern Iraq. (Too big to post).

Don't get me started on the crappy population/demographic data (my actual expertise). There are soldiers on the ground who knew more about how many people were where than the big shot decision-makers, but no systematic mechanism to collect and consolidate the data for real-time use. (Sure, you'll here from some spooky people here who might suggest it's somewhere else. Been there, done that. It ain't. Just a big intel/decision gap).

In a recent report, the Kagans cited that Afghanistan has a pop of between 26 and 30 million. The pundits usually say 33 million. They all make this stuff up and cross-cite each other. If the Afghan pop is plus or minus 7 million, how accurate do you think the local pop data is.

Entropy

any chance you could add the Oruzgan/Dai Kundi provincial boundary in (shouldn't change your data too much as the green bit is essentially Dai Kundi) ... I could then very definitely use these at work in an UNCLAS project we have going.

Jas

Glad to see it went well.

Quote:

---

Originally Posted by Spud

Entropy

any chance you could add the Oruzgan/Dai Kundi provincial boundary in (shouldn't change your data too much as the green bit is essentially Dai Kundi) ... I could then very definitely use these at work in an UNCLAS project we have going.

Jas

---

Unfortunately, I don't have shapefiles for the other, newer boundary models. I could probably wag the Daykundi provincial boundary, but the districts would mostly be wrong.

Entropy:

Please don't create a line on a map based on a wild ass guess.

That's what NGA did in Iraq in 1993 for provincial boundaries, and, once created, became gospel to anyone who got the maps. And the same happened after 2003 when the did the same for district and subdistrict boundaries. A complete fiasco.

Anyone with a GIS capability to create a map line should be reminded that, in many of these places, lines are drawn in blood, and creating maps based on wild ass guesses is where the flow starts. Most people don't understand that these oh-so-very-official-looking-maps in most of these conflict areas, are just made up.

I have a compendium of so-called "Green Lines"- demarcations of Kurdish territories, which people downloaded from the internet (from sites like this), and 10 years later, are the root of a crisis. Every one was different.

I come out of the tradition of school attendance boundaries, where any proposed change can bring out hundreds of "concerned" citizens, because a change of a few hundred feet change really will affect something that matters to them---where their kid goes to school, their property value, their definition of community and their place within it.

Just remember that when we move official political and administrative boundaries by potentially miles, it affects perhaps hundreds of people, and in a far more threatening way than just a school boundary change.

Sorry if that got to lectury, but there is too much of this going on in these very dangerous areas.

Check out this Iraq province map from a banner on this SWJ site.

http://www.understandingwar.org/file...ovinces.01.jpg

First, Ninewa and Irbil are incorrect (Makhmour).

Second, Baghdad, which is a province in its own right, is incorrectly shown as a microspec of red, and the actual Baghdad province land shown as a part of each adjacent province.

Third, Salah ad Din province (Tikrit) is north of Baghdad. Can you find the lines that delineate Salah ad Din's southern boundary? I can't. But the best extrapolation would put Taji, a part of Baghdad province, in Salah ad Din.

It is this map, (as repeated by everyone once incorrectly "published") that creates confusion and boundary disputes in the Ad Dujayl area (the portion of southern Salah ad Din that is primarily Shia) in which some would much rather affiliate with Baghdad vs. Sunni Salah ad Din. Note: Until Sadaam created Salah ad Din, all of it was a part of Baghdad province, so there is a lot of bewildering history.

Junk maps that create more confusion than they are worth.

I am increasingly coming to the belief that admin/pol maps in these areas are better provided as a layered power point slide vs. a static map, so they can see it as a dynamic element and understand the realities and issues in these places.