Being able to project future spatial population distributions
is an important tool to tackle challenges following
continuous global population growth and possible
future climate challenges. This report presents
an answer to how a convolutional neural network can
be used to project future spatial population distribution
and what results can be achieved by using this
approach. This is investigated by designing a convolutional
neural network, PopNet. PopNet identies
complex spatial patterns in historical data, on a 250
meter resolution grid and projects the future spatial
population distribution. From this architecture, two
future scenarios are simulated based on IIASA SSP2
population projections for Denmark and France respectively
and the results evaluated. While the neural
network method does have aws, the results prove
how, and that, convolutional neural networks can be
used to project future population distribution. A number
of key challenges, strengths and weaknesses are
found and further alterations are proposed that could
improve PopNet precision and applicability for future
use.