Knowledge of the size of the population during humanitarian emergency and/or development interventions is essential because it allows for better planning of assistance (food and non-food distribution, vaccination campaigns, etc.). Classical methods are based on the estimation of the target population from existing documents and data (census reports, list of households, etc.). But some contexts do not allow to have these documents to do this work. As a result, emergency response standards are used or individual counts are necessary and difficult to implement. For this reason, during a recent humanitarian emergency assistance to the internally displaced persons camp in Ngala (Borno State, Nigeria), we decided to combine the counting of dwelling done from satellite picture (Earth Observation) with a field survey, using different statistical methods to estimate the population size.

Since September 19th 2016, the Swiss section of Médecins Sans Frontières (MSF-OCG) supports the population of the camp of Internally Displaced Persons in Ngala (Borno state, Nigeria). A rapid survey organized in October 2016 show a very high mortality rates twice above emergency threshold, the mortality rate was below emergency threshold during a survey organized in February 2017 showing an improved situation. However, considering the population size of the camp estimated by MSF team at this date, the number of medical consultations and the number of person hospitalized was below what was expected. MSF teams questioned the actual size of the camp, then, we proposed and used this method to estimate the population size. Based on a satellite picture taken on December 31st, 2016, dwellings inside the camp were counted and classified in 3 categories (4724 White tents, 4592 Brown dwellings, 2582 Tukuls). We organized a survey on February 17th, 2017 to estimates the average number of people who slept under a shelter the night preceding the survey for each category. We sampled 600 shelters divided in 30 clusters. Each starting point of each cluster was chosen by spatial randomness, the shelters were then selected by proximity. We collected for each cluster the number of people who slept inside the night before the survey, the colour and the material of the roof. Because we did not find any obvious correspondence between classification by field observation and by earth observation, we tried different hypothesis. We applied 3 distinct statistical methods to calculate our estimates and its confidences intervals.

We collected information on 600 shelters. Fifty-six percent of the shelters had a white roof. Inside shelter with white roof the average number of occupants was 3.04 (95% CI: 2.81–3.28 design effect: 1.6) while it was 2.51 (95% CI: 2.26–2.76 design effect: 0.9) for non- white roof shelter. From the 4 different hypothesis and 3 statistical methods used our estimates of the total population size of the camp range from 32,413 to 33,433, and the minimum lower bound of the confidence interval is 29,043 while the maximum is 36,058.

We have estimated a population of 32,513 people who have slept in the camp the night preceding the survey (95% CI: 30,106–34,920). This estimate fall below previous estimates. Although it does not include the population living in large buildings (less than 2000 according to MSF teams). We were able to do this estimate by the combination of earth observation and field surveys; this was done in a short timeframe with limited resources. The methods can still be simplified. We suggest reducing to two types of shelter and a smaller sample could reach the expected accuracy.