We have seen what rapid urbanization in low-and Middle-income countries has led to, although there have been a lot of advantages to this effect but we cannot rule out the issues as regards the disadvantages. It has led to an ever growing number of slums dwellers with inadequate access to basic amenities. Looking into the continent of Africa, as regards the paper, Slum dwellers are categorized by socioeconomic vulnerability, including poverty and unemployment and are situated in areas that are prone to risks such as flood zones.

Earth observation (EO) was termed the sustainable source that would give detailed information about slums in order to track their progress, compare indicator outputs across countries and informing policies and plans.

This research gives an insight into two main objects/elements that contribute to successful EO that is based on slum mapping. They include:

• Contextualizing slums which allows understanding of local context and user requirements
• Conceptualizing and building models which allows local slum characteristics to be translated into image features.

The EO community methods have been usually data driven with an inadequate understanding of different end-users. Map makers are often uncertain on the required map scale, accuracy measures and level of details required by end users. Also, end-users have misconceptions and low awareness of the potential and limitations of EO.

The main contribution to the paper were in two folds:

First, it identifies knowledge based features including local context-knowledge, end user requirements and geo-ethics to contextualize slums. Most EO studies have shown that classification of slums are done without fully understanding the end users requirements and often do not ask important questions on how maps should be/ produced to ensure ethical data sharing.

Secondly, it proposes a user-driven EO based approach that integrates the above issues into the machine learning based mapping model. Here, the approach aggregates slums to an appropriate unit to avoid unnecessary details and ensure ethical data sharing.

The case study in this paper was the capital of Ghana which is Accra where it is estimated to have a population of about 4.9million and about 38.4% of this population lives in the slum.

The input data for this region was in two folds; Primary data and the secondary data. For the primary used SPOT 6 Imagery that was gotten from ESA through a third party grant. Also primary data also includes field photos and expert interviews to understand the local context, end-user requirements and geo-ethics.

The secondary data included the official slum data set used for training and validation of machine learning-based mapping and Open street Map (OSM) data. OSM includes features such as railways, streams drains, etc, used for creating street blocks.

The methodology for this research was also divided into two main steps:

• Data driven slum mapping, here it begins with a machine learning based mapping and then passes through a series of steps till in gets to analysis with a random forest classifier, which help to give a high prediction for accuracy and can handle high data dimensionality. This method due to EO based application can have misclassification and thus an uncertainty analysis was also carried out. In this research, uncertainty analysis focused on slum street blocks only. It can be grouped into two types:

1. Existential uncertainty that refers to the possibility that a street block is classified as a slum but does not correspond to a slum on the ground or the possibility that a slum street block is not detected.
2. Extensional uncertainty refers to the level of confidence a street block is classified as a slum.

• User driven approach, this simply were gotten from field observation and expert interviews. Here in-situ observation were conducted to investigate the causes of misclassification in the data driven model. The paper gave 14 location points that were purposefully sampled using specific criteria based on visual assessment of data-driven classification results. This method also gave a level of analysis based on three areas: Technology level that deals with ethical issues related to the machine learning classification, input features and accuracy metrics. Product level that referred to how information was represented and how it can be disseminated. Then finally the Application level that tells us how the information would be used to treat slum dwellers.

It was concluded that developing policy-driven geo information is essential to support SDG.

Sources: Maxwel Owusu, Monika Kuffer, Mariana Belgiu, Tais Grippa, Moritz Lennert, Stefano Georganos, Sabine Vanhuysse. (2021). Towards user-driven earth observation slum mapping. https://doi.org/10.1016/j.compenvurbsys.2021.101681