Environmental Geospatial Statistics of Zaatari Refugee Camp


  • Author: Lillian Pierson
  • Date: 03 Oct 2013
  • Copyright: Images appear courtesy of iStock Photo

Al Zaatari Refugee Camp was opened back in July of 2012 as a quick solution to the shelter needs of tens of thousands of Syrian refugees who were fleeing the violent and dangerous conditions of their mother country. But as time passed Syrians continued to flee in droves and, as of March 2013, the UNICEF estimated that there were over 363,031 people who are either registered as refugees or who are waiting to register to live in the 5.3 km2 camp. Zaatari is located in the dry and desolate reaches of the Jordanian desert, about 10 km east of Mafraq. Most people are living in tents, while a few fortunate ones get to live in trailers that are owned by the United Nations High Commissioner for Refugees (UNHCR). With thousands of refugees fleeing to Zaatari each month, the security situation within the camp has been steadily deteriorating.

thumbnail image: Environmental Geospatial Statistics of Zaatari Refugee Camp

While camp safety and security is a serious concern for aid workers at Zaatari, environmental health and sustainability is another huge issue. Back in March, UNICEF estimated that in one week they (and ACTED) supplied 3,322,000 L/day of water, removed 1,250 m3 of wastewater, and removed 1,400 m3 of solid waste from the camp. The environmental services needs are staggering and significantly underfunded. Water resources are very scarce, many of the bathroom facilities do not work, and the shower areas of the camp are subject to high rates of vandalism. To make matters even worse, the residents of Zaatari are exposed to dusty dry winds and frequent sandstorms that are common to the desert in which they are living. “But what does all of this have to do with statistics?” you say; potentially, a lot!

If you have access to data on geospatial variables, then the statistical analysis of point patterns can be used to uncover and monitor critical environmental health issues in refugee camp and early settlement populations. Al Zaatari Refugee Camp is a great candidate for these types of analyses because geospatial data has been collected consistently on the refugee population since its opening.

The use of statistics to explore spatial point data at Zaatari Refugee Camp has the possibility of uncovering unknown problem areas for security and environmental conditions within the camp. It also has the possibility to provide information that can help aid workers improve their distribution of goods and services of all types. Poisson distributions, inhomogeneous Poisson distributions, the Thomas process, stochastic kriging, trend surface analysis, kernel density estimates, and Monte Carlo testing can be used to identify relevant trends and clustering statistics within the refugee camp.

The aim of these analyses would be to uncover significant geospatial trends that can be turned over to aid workers for ground-truthing and further exploration. Our data includes point data that represents the tented population of people living in Zaatari Refugee Camp at several measurement intervals over the last year. This data was collected by UNOSAT and generously shared by Edouard Legoupil, Information Management Officer at UNHCR, and Lars Bromley, Principal Analyst at UNOSAT. The statistical and geographic methods that must be applied to analyze this data can be carried out using ArcGIS, GeoDa, and R Programming Language. Let’s examine how these statistical methods can provide insight into environmental and population dynamics within the Zaatari Refugee Camp.

Point Data

Traditional statistical methods are quite difficult to deploy in spatial data analysis. This difficulty rests upon the fact that traditional statistical methods assume randomly distributed source data when, in geographic data, randomness almost never happens. Geographic data is marked by its auto-correlative tendencies; two objects that are close together in space are more likely to share similar features than two objects that are far away from one another.

The Poisson Distribution

Poisson Equation

Despite geographic auto-coorelation, we can still use the independent random process (IRP), or complete spatial randomness (CSR), in order to test geographic point patterns for quantifiable clustering across a study area. By testing our point data against the IRP/CSR and a Poisson distribution, we should be able to quantify and map any significant spatial clustering of populations living in the Zaatari Refugee Camp. By having a clear map of significant population clustering within the camp, aid workers may be able to better plan the distributions of their environmental services.

The use of statistics to explore spatial point data at Zaatari Refugee Camp has the possibility of uncovering unknown problem areas for security and environmental conditions within the camp. It also has the possibility to provide information that can help aid workers improve their distribution of goods and services of all types. 

While IRP/CSR and Poisson distributions are great ways to model and test point pattern data for clustering against a random set of points, they are not able to provide information about the effects of environmental probability and density-based interaction-intensity between points in a point pattern. Geographic data exhibits both first- and second-order effects. First-order effects of geographic data are attributed to an uneven distribution of environmental probability across geographical areas. Second-order effects are derived from the intensity of interactions between points or events in space.

Inhomogeneous Poisson

Inhomogeneous Poisson Equation

Inhomogeneous Poisson Distribution is great for testing geographic point data for the first-order effects of environemtal probability. The inhomogeneous Poisson differs from a standard Poisson distribution in that, with inhomogeneous Poisson, event intensity (ʎ) varies with location. By testing our point data against an inhomogeneous Poisson distribution, it may be possible to uncover and track changes in environmental probabilities that are effecting the spatial distribution of tent populations in the Refugee Camp. When combined with an aid worker’s personal knowledge about environmental conditions at Zaatari, geospatial statistics about environmental probability and tent camp distributions have the potential to uncover unknown areas of environmental concern.

The Thomas Process

To test for second-order interaction-intensity between points in our spatial dataset, we can use the Thomas Process. The Thomas process makes use of homogenous and non-homogenous Poisson processes that generate “parent” events and a random number of randomly placed “children” events around them. The process then deletes the “parent” events, leaving only “children” points against which we can compare our control data. By testing our point data against the Thomas Process, it may be possible to uncover and track changes in density-interactions and interaction-intensity between tent camps at Zaatari. These findings can then be cross-referenced against the personal knowledge of environmental aid workers in order to determine how changes in density-interactions might relate to changes of environmental conditions in certain regions of the camp. In the future, this type of information could potentially be molded into a type of early-indicator metric for environmental need or concern within a refugee camp.

Water conditions at Zaatari camp

Kernel Density Analysis and Monte Carlo Testing

Kernel density estimation is a good way to avoid the problem of “edge effect” when quantifying results for analyses that utilize Poisson distribution, inhomogeneous Poisson distribution, and the Thomas process. This estimation method quantifies spatial variation of events in a study area by comparing the kernel density values of your control data with those of your model process output. Although KDE is a great way to test for spatial variation, it should be used with Monte Carlo testing in order to establish a good confidence interval for your conclusions.

Field Data

The field data behind a point pattern can be analyzed collectively in order to uncover trend surfaces and in order to interpolate what is happening in unsampled geographic regions. Trend surface analysis and stochastic kriging are two useful techniques for interpolating spatial field data.

Trend Surface Analysis

Trend surface analysis simply shows a fundamental trend of the distribution and intensity of one metric across your study area. For the point data that represents the Zaatari population, our z-value would probably be represented by the number of people that occupy each grid area within the camp. We could use this method to show the basic distribution of tents across the camp and also to show how that distribution has been changing over time. Although this method is far less precise than kriging, we can still generate residual coverage maps that show where and how predicted trend surfaces differ from the actual data values across space.

Stochastic Kriging

We can use stochastic kriging to form a coverage of our study area by interpolating over a small subset of control data. Isotropic and anistropic kriging should both be carried out in order to test the model output for fitness against the lateral drift in our study data, or the lack thereof. Of the several different types of kriging available, it appears that universal kriging would be the most appropriate method to use for analyzing the point data collected from Zaatari Refugee Camp. This is because universal kriging can be used to model isotropic geographic data that exhibits some extent of regional drift. By performing successive rounds of universal kriging across the study area over time, we could possibly use the resulting coverages to form a predictive model for the future spatial distribution of tent populations at Zaatari. Accurate predictions about future population distributions can be used to help anticipate need and to plan for the most effective distributions of environmental services in the future.

Related Topics

Related Publications

Related Content

Site Footer


This website is provided by John Wiley & Sons Limited, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ (Company No: 00641132, VAT No: 376766987)

Published features on StatisticsViews.com are checked for statistical accuracy by a panel from the European Network for Business and Industrial Statistics (ENBIS)   to whom Wiley and StatisticsViews.com express their gratitude. This panel are: Ron Kenett, David Steinberg, Shirley Coleman, Irena Ograjenšek, Fabrizio Ruggeri, Rainer Göb, Philippe Castagliola, Xavier Tort-Martorell, Bart De Ketelaere, Antonio Pievatolo, Martina Vandebroek, Lance Mitchell, Gilbert Saporta, Helmut Waldl and Stelios Psarakis.