Developed for World Bank Group

Results

Bole Lemi

Bole Lemi Industrial Zone is located in the outskirts of Addis Ababa, Ethiopia, directly to the south east of the city. It represents Ethiopia first hub targeting export-oriented manufacturers aimed at supporting the government's bid for economic growth.

The 156 hectares site started operations in 2014, with all pre-erected factories rented-out to more than 12 different corporations.

The ethiopian government sold the initiative as a way to create jobs, increase export revenue and help share best practice with local investors. This IDZ particularly focussed on attracting textile and apparel investment. The site is located along key economic corridors, surrounded by Infrastructures such as airports, railway lines, dry ports, universities etc.

The site remains fully occupied. Shoe and garment manufacturing is prevalent at the site, with investment from the likes of Phillips-Van Heusen (PVH) Corporation (A US clothing company that owns Tommy Hilfiger, Calvin Klein and Heritage brands) On the back of this SEZ, Bole Lemi Phase 2 (186 hectares) is currently being developed in collaboration with the World Bank Group. In the future, a residential and recreational area adjacent to the site is intended for development.

  • Size analyzed: 50 hectares
  • Established: 2014
  • Phase: Creation
Google image of Bole Lemi I IDZ, Ethiopia
Google image of Bole Lemi I IDZ, Ethiopia

Analysis

The Bole Lemi I IDZ site reflects a major investment on the outskirts of Addis Ababa. The venture’s targets are aimed squarely at accelerating the capital’s (and hence the nation’s) economic development. Specifically, cited goals are to create job capacity and to increase export revenue.

As with all sites, analysis of employment levels is eminently estimable (with high confidence) from CDR transaction logs if and when such datasets are available. However, building footprint, electricity consumption estimates and land-use model, as derived from EO data, all provide useful support to estimate impact on employment.

Unlike other SEZ site, infrasturctue is already strong in the region. Thus, usage of this infrastructure is the key micro-indicator for assessing manufacturing/distribution success. For phase II, identification of freight movement on key roads promises to underpin relevant analysis with clear access points to the site.

Given the proximity to central Addis Ababa, ‘success’ of this SEZ ought also be expressed in the context of its broader impact to the city. Reduction of poverty levels, unemployment and, importantly, elimination of slums across the city are key goals to the local government. Assessing Bole Lemis impact on these issues therefore remains key. To this end, an ongoing and iterative analysis of land use development from EO imagery (with specific tracking of informal/formal residential distributions across the extent) would be extremely valuable.

Furthermore, analysis of CDR dataset across Addis Ababa would provide unparalleled insight of impact - understanding the mobility of local population groups (from the entropy of their cell tower usage), to their range of travel, to analysis of where IDZ employment is commuting from all provides compelling evidence as to the sites impact.

Land Classification

A third scene was added to the land classification to illustrate the changes to the zone in the first year of it's creation, where most of the construction occurs. The brown color indicates an Other category that is some partially built up region that is made up of impervious surfaces and bare earth. In the second image the foundations of the buildings show up in the Other category.

The classification performs very well in this image since the zone is small and there are a limited number of materials present. The largest conversion of earth into built areas occured within the first year of it's creation.

Transport Infrastructure

The total length of roadway (in km) was calculated using imagery from Digital Globe's Geoeye-01 and Worldview-02 satellites. Here, Skynet gives a reasonable road mask though there is room for improvement as many of the roads are incorrectly severed.

Total km of Roads: 2014 to 2017

Building Footprints

The total building area (in m^2) was calculated using imagery from Digital Globe's Geoeye-01 and Worldview-02 satellites. Here, Skynet misses many buildings. This could be because it was trained using a large training data set from a different area of the world where there are few white roofs.

Total m^2 of building: 2014 to 2017

Building Counts

The trained Tensorflow object detection neural net was tested on GeoEye imagery in Feb 2014, we detect 11 buildings in the scene, and while tested on Worldview imagery from Mar 2017, 24 buildings are found. The model results turned out pretty close to independent buildings number in the site. The object detection neural net performs relatively well in this site since the images and buildings are pretty high quality in both dates.

We detected 11 buildings and 24 buildings in 2014 and 2017 respectively

Electricity Consumption

Electricity consumption is highly correlated to a region's economic activity. To estimate electricity consumption, we used the Visible Imaging Radiometer Suite (VIIRS) Day/Night Band (DNB) imagery to correlate nighttime light radiance. Specifically, monthly VIIRS DNB product from The Earth Observations Group (EOG) at NOAA/NGDC was used.

Monthly mean radiance values (in nanoWatts/cm2/sr) from Jan to July in 2017

Pollution

Air quality and air pollution can also be monitored through VIIRS satellite imagery. Aerosol optical thickness (AOT) of VIIRS was obtained and processed as a proxy for air pollution indicator. Because it is related to the particulate matter, e.g. PM2.5, concentrations in the atmosphere, AOT is widely used for air quality and pollution applications. AOT imagery is recorded and updated daily at 6 km resolution.

The mean values of aerosol optical thickness in 6944 (Jan 1st, 2016) and 65535 (Jan 1st, 2017)