Investigating the accuracy of georeferenced social media data for flood mapping: The PetaJakarta.org case study

Abstract

Georeferenced social media data are gaining increased application in creating near real-time flood maps needed to improve situational awareness in data-starved regions. However, there is growing concern that the georeferenced locations of flood-related social media contents do not always correspond to the actual locations of the flooding event. But to what extent is this true? Without this knowledge, it is difficult to ascertain the accuracy of flood maps created using georeferenced social media contents. This study aims to improve understanding of the extent to which georeferenced locations of social media flood reports deviate from the actual locations of floods. The study analyses flood-related tweets acquired as part of the PetaJakarta.org project implemented in the coastal mega-city of Jakarta and provides insight into the level of accuracy expected with using georeferenced social media data for flood mapping. Importantly, the results reveal that the accuracy of flood maps generated with georeferenced social media data reduces with increase in the size of the minimum mapping unit of the flood map. Finally, an approach is recommended for creating more accurate real time flood maps from crowdsourced social media data.Abstract-Georeferenced social media data are gaining increased application in creating near real-time flood maps needed to improve situational awareness in data-starved regions. However, there is growing concern that the georeferenced locations of flood-related social media contents do not always correspond to the actual locations of the flooding event. But to what extent is this true? Without this knowledge, it is difficult to ascertain the accuracy of flood maps created using georeferenced social media contents. This study aims to improve understanding of the extent to which georeferenced locations of social media flood reports deviate from the actual locations of floods. The study analyses flood-related tweets acquired as part of the PetaJakarta.org project implemented in the coastal mega-city of Jakarta and provides insight into the level of accuracy expected with using georeferenced social media data for flood mapping. Importantly, the results reveal that the accuracy of flood maps generated with georeferenced social media data reduces with increase in the size of the minimum mapping unit of the flood map. Finally, an approach is recommended for creating more accurate real time flood maps from crowdsourced social media data.

Analisis Penggunaan Media Sosial Dalam Mitigasi Bencana Banjir Di Dki Jakarta Tahun 2020

“…Social media is one type of technology that has the potential to facilitate rapid and effective disaster response (T. Holderness & E. Turpin, 2015). Of the many social media platforms that exist today, such as Youtube, Instagram, Facebook, and Twitter, which are included in the media that are widely used in Indonesia.…”

“…From some of the information provided by the community regarding floods, it helps people to avoid flood locations and be more careful. Jakarta is home to millions of residents who are very active on social media sites, in such a way that the city has been dubbed the "Twitter capital" of the world (T. Holderness & E. Turpin, 2015).…”

“…The impact of the frequent and high intensity flood events experienced in Jakarta is exacerbated by a number of factors, including physical geography, climate change, poor maintenance of drainage networks, subsidence, sea level rise, underfunding, aging and deteriorating flood control infrastructure, growth fast and unplanned cities, slums and illegal solid waste disposal in drains (T. Holderness & E. Turpin, 2015).…”

A review on the application of social media data in natural disaster emergency management

“…。严重的 自然灾害往往会造成高昂经济损失和重大人员伤 亡 [3][4][5] 。重大自然灾害频发对区域的可持续发展来 说， 不仅是沉重的压力， 也是严峻的挑战 [6][7] 。要完 全消除自然灾害的影响几乎是不可能的 [8] ， 但可以 通过灾害管理来降低灾害损失 [9] 。目前， 包括社交 媒体、 位置服务和机器学习等在内的新兴技术被证 实可用于灾害管理 [5] 。 随着互联网技术的发展和 5G 时代的来临， 大 数据受到了越来越多的关注。李德仁 [10] 总结出大 数据具有速度快、 体量大、 真伪难辨、 模态多样、 价 值大等特点， 并指出时空大数据的挖掘和研究将会 是未来的热点。作为一种时空大数据， 社交媒体数 据具有实时性和位置服务的特点 [11] ， 已经成为灾害 管理的研究对象之一 [12][13] ， 将社交媒体数据应用到 灾害应急计划和危机管理中， 能有效提高应急管理 的效率 [14] 。 社交媒体数据能够弥补传统遥感、 气象观测数 据在解析致灾程度、 灾情中的不足。遥感技术的优 势在于实时监测灾害进程， 反馈图像和波段信息， 不足在于采集灾害数据的时间周期较长， 且遥感产 品与受灾群众的交互性不高 [15] 。气象观测数据的 优势在于定点观测致灾强度信息， 进而采用数值天 气预报模型以及大气模型等模拟致灾强度的空间 分布， 不足在于模型模拟出的结果准确度不能保 证， 且该结果仅仅代表水文气象方面的考虑 [16] [17] ， 核心思想是 "人人都是传感器" [18] ， 即每一个体 都可以采集地理信息数据进行共享。VGI 通常是 指普通用户以移动互联网为媒介自发协助完成空 间地理信息数据的采集、 处理、 管理与维护 [19] ， 其特 点包括 Web 2.0 [20] 、 集体智慧 [21] 与新地理 [22] 。与 VGI 类似， 相关研究证明社交媒体数据能反映灾情信 息 [23][24] ， 感知灾害信息 [25][26][27][28][29][30][31][32] [33] ； ② 使用位置熵和 马尔可夫转移矩阵结合时空维度， 反映致灾因子动 态演变 [34] ； ③ 通过空间聚类识别受灾区 [35] 乃至重灾 区 [36] 。特别地， 对于地震灾害， 可以利用空间增长 模型快速估计震区烈度 [37] ， 但该方法对原始数据要 求较高。另外， 滤波分类 [38] 和时间序列与空间变化 分析 [39] [42] ， 但由于 缺乏官方的致灾强度图， 使优化结果无法验证等。 在利用社交媒体数据提取因灾损失信息方面 的主要方法有： ① 采用语义分析 [43] 、 语义分类 [44] 、 情 感分析 [45] 和主题标签 [46]<...>…”

An approach for assessing the quality of crowdsourced geographic information in the flood management domain

“…Degrossi et al (2014) and Moreira, Degrossi and Albuquerque (2015), for instance, estimated the quality of water level observations by comparing them with authoritative data (i.e., sensor data). Ogie and Forehead (2017) also used authoritative data to evaluate the positional accuracy of flood-related Twitter messages. However, the lack of authoritative data in many parts of the world hampers traditional quality analyses that compare CGI to reference data sources.…”