O17-2020: Team H
This project (LandSwipe) aims to provide rainfall-induced landslide warning information by exploiting the use of Artificial Intelligence (AI) and involving citizen science reporters to develop digital crowd mapping. In case there is landslide warning, I wish people can have spare time to prepare themselves and keep away from landslide predicted area which lead to minimize the loss of life and materials. This project may also empowers stakeholders to systematically monitor and prepare planning for mitigation purposes.
Challenge #2: Predicting Landslide
Data showed an inclining trend of landslides which caused by rainfall in Indonesia from 2014-2018. The increase amount of rainfall has linear relationship with considerable number of landslide incidents. LandSwipe project aims to use Artificial Intelligence (AI) approach to predict rainfall-induced landslide pattern and involve citizen science reporter to develop digital crowd mapping. Rainfall data derived from Global Satellite Mapping of Precipitation (GSMaP), comprise of rainfall accumulation, intensity, and duration, used as input for AI models then overlay it over topographic map by using GIS. Ten configurations were performed using R, as programming language, to process the data. The performance of AI model will be tested using two statistics parameters: accuracy value and false alarm ration (FAR).
Hi I am Anistia Malinda Hidayat from Indonesia. I work as a single team and currently study about meteorology.
Indonesian National Board for Disaster Management stated that there is an inclining trend of landslides in Indonesia from 2014-2018. Compares to the other provinces in Indonesia, Central Java has the highest number of landslide accidents. The number of landslide accidents in Banjarnegara is second highest after Wonogiri, but the population density in Banjarnegara is higher (857,12 population/km2) than Wonogiri (551,48 population/km2). Therefore landslide that happen in Banjarnegara will has higher exposure/impact to the society. Predicting landslide could become more challenging because there are so many factors that have to be considered in this project. To make it more clear, below I listed some difficulties that need to be considered in this project:
The people who potentially most affected by landslide accidents:
Based on the analysis on the previous explanation and my background knowledge as candidate of meteorologist, here are 3 points that I assume becoming the major cause of problems:
I believe that 4 points that I have mentioned before as the main cause of the problems because:
1. Some studies showed that people insist to live around landslide-prone area because:
2. Aldrian and Susanto (2003) showed that based on identification of three dominant rainfall regions, rainfall pattern over Banjarnegara strongly influence by monsoonal activities. High rainfall pattern which observed during November until March (NDJFM) followed by the highest number of landslide accidents during 2014-2018. Based on Indonesian National Board for Disaster Management data, 78% (84 cases) out of 108 cases of landslide accidents happened because of rain. Therefore, I strongly believe that rainfall will be the main feature on this prediction model.
3. Automatic Weather Station (AWS) just put on Banjarnegara early in 2019 (March) and there are around 3 months data which is abnormal (the value is continuosly too high) so the accuracy of data is questionable -> using satellite observation data (GSMaP become an option)
4. News showed several scientists have made LEWS but I cannot found another news explaining us about the detail location of the instrument, how this instrument transfer/distribute the information to the society, how this instrument connect the government, related instances, and the society when disaster happen, etc. This is how the 'gap' formed, only certain or limited circles who know about this, while the users (society), who actually become target user, do not have any idea about this
MAIN IDEAS. Since Artificial Intelligence (AI) is well-known for its capability on handling big dimension of data and process non-linear data pattern, I plan to make near-real time landslide prediction by using AI approach over a specific coordinate point area. When there is landslide warning, this information will be automatically forwarded to the society who stay within 15 km radius.
REFERENCES. This idea mainly inspires by some of these research:
VALUE PROPOSITION OF LANDSWIPE PROJECT
There are two important things that need to be elaborated here, first is about how to develope accurate AI model prediction. A flow chart describing the way to develope AI model prediction (in this case I use Probabilistic Neural Network) is presented on the above picture. There 3 stages in general, where stage 1 consist of 2 substages.
Second issue that we need to address is how to give this information to the society, how to connect all elements (society, related agencies, government, and other stakeholders), and also how to involve society as part of citizen science. The answer is to develope mobile apps (LandSwipe). The application, design, and the way it works already explained in the previous section.
LandSwipe project considered to possess an impact if it fullfill several key metrics mentioned below:
Potential people that most likely to be supportive are:
Those who most likely to oppose are:
BRIEF EXPLANATION OF LANDSWIPE PROJECT
Aims: predicting landslide related to rainfall (rainfall-induced landslide)
Tools: Artificial Neural Network (PNN) involving citizen science reporters (community participation) and using of R programming language
Data: landslide historical data (response data) + rainfall data from GSMaP in 2014-2018 (feature) + citizen science reporter data (timing of landslide occurences -> validate response data)
Method: There will be two PNN model, one use to predict landslide in rainy season and the other for dry season (Indonesia has 2 season)
Target user: All people (especially people who live over landslide prone areas and familiar with the use of social media/internet/mobile phone/even AI and crowdsourcing) + connected to government/related instances/hospital
Location: One of the most landslide prone areas in Indonesia with highest population density in Central Java Province -> Banjarnegara -> if it success then will expand to another places
*Notes: detail explaination of how this project scientificly work was showed in theory of changes
What will the user experience?
Almost half of the total human population in the world are active users of social media. The Next Web published data per April 2018 that showed Indonesia always become the top five social media users in the world. So, I plan to design social media platform-look like that specifically created for discuss, report, comment status, or post photo about landslide information named LandSwipe. By doing this, people can enjoy social media and help other people at the same time (new way enjoying social media which is so meaningful and helpful). Additional features of LandSwipe besides its role as social media are:
Who has to do what. to make it happen?
Who are the key partners to execute? Key partners to help others evaluate your value preposition?
What are the precipitating events?
who else is in the field?
What's wrong? Missing? Not working?
Physical and intellectual resources needed (besides financial resources)
Next steps? Pilots?
Cost structure? Financial Sustainability? Revenue streams?
1. Research development fund
2. Marketing cost
*Notes: Research Budget Plan (RAB) is mostly based on the Regulation of Finance Minister Republic of Indonesia No. 78/PMK.02/2019 concerning Standard Budget for 2020
How might this go wrong? How might the problem evolve? What are the legal, cultural and other impediments?
How will i promote adoption?