Combining water point data and machine learning to predict the potential spatial distribution of specific type or group of drinking-water source/service

Weiyu Yu

by Weiyu Yu 09/25/2017 06:45 AM BST


Description

Although SDG monitoring on drinking-water progress is often based on national level indicators, sub-nationally and geospatially disaggregated indicators may become increasingly important as they could effectively reveal inequalities in services between different geographic locations and population groups. As more recent concerns are raised about issues such as water quality, the functionality of facility, and continuity of service, it may also become increasingly important to disaggregate data by specific type of water service. Currently, disadvantaged drinking water services are not often reported at high levels of geospatial disaggregation. Fortunately, as more geospatial data sources become available with the transition from the MDGs to the SDGs, predicting the potential spatial distribution of specific type(s) of drinking-water source using machine learning method becomes possible. DHS modelled surfaces as one of such novel data could potentially be the important sources of predictive covariates for modelling the potential distribution of specific type(s) of drinking-water service.

Co-authors to your solution

Jim Wright, Nicola Wardrop

Link to source code or original files

https://geoterry.github.io/GEOWAT-SDGinsights/case1_maxent_code.R

Please enter a link to your solution (working demo)

https://geoterry.github.io/GEOWAT-SDGinsights/

Submission status

Link to Solution / Working Demo


Comments

David Jeronimo Giraldo Atehortua

by David Jeronimo Giraldo Atehortua 10/10/2017 06:59 PM BST

hello!

that project have any API or WS? i want connect my world data viewer whit you data!

Spigit Reputation:User Reputation 69.44

Jorge  Martinez-Navarrete

by Jorge Martinez-Navarrete 10/11/2017 03:04 AM BST

We are starting the Collaboration & Teaming Phase, check out the home page for what to expect!

https://sdginsights.ideas.unite.un.org

Spigit Reputation:User Reputation 100.00

Weiyu Yu

by Weiyu Yu 10/31/2017 01:31 AM GMT

Many thanks

Marcelo LaFleur

by Marcelo LaFleur 10/30/2017 09:01 PM GMT

Hi Weiyu,

Your proposal is fascinating. Looking at the project files and code, I see that it is quite flexible to different specifications and to the use of other covariates. Have you tried other countries to get a sense of the additional effort needed? How easy would it be to correlate or visualize the predicted surface water distribution with  income/assets, population, etc? 

Spigit Reputation:User Reputation 57.34

Weiyu Yu

by Weiyu Yu 10/31/2017 01:00 AM GMT

Many thanks. You are right; it is flexible to different statistical or machine learning algorithms as well, and it is therefore possible to apply such predicted surface as a basis to produce ‘thematic population maps’, for example, mapping the population using boreholes as the main drinking source. Similarly, there are also scopes to correlate such surface with income/poverty as well. As a part of my PhD study I’ve also modelled tube wells/boreholes in other countries such as Kenya, Somalia and Tanzania.

David Jeronimo Giraldo Atehortua

by David Jeronimo Giraldo Atehortua 10/30/2017 11:46 PM GMT

Hello!

is a good idea!! you have any shapefile for i add to my Word Data Viewer?

thank!

Spigit Reputation:User Reputation 69.44

Weiyu Yu

by Weiyu Yu 10/31/2017 01:29 AM GMT

Sorry for being unable to reply your comments earlier. Many thanks for your questions and I also like your idea of Word Data Viewer. I do have some modelled surfaces (GeoTIFF format) in my hand, and we'd be delighted to collaborate with your nice work. My email is W.Yu@soton.ac.uk, we could keep in contact and see if our projects could express some values.