Event inventories, VRSC, 11/13/2023, Alexander Dunkel

### Motivation

Note:
- landscape architect and landscape planner
- environmental impact assessments
- temporal characteristics of landscapes often missed

---

Hull & McCarthy (1987).  
_Change in the Landscape_.  
Landscape and Urban Planning 15 (3–4): 265–78.  
[DOI](https://doi.org/10.1016/0169-2046(88)90050-3)

Notes:
- pointed out in several publications
- in practice, no significant change

---

<br>
<span style="font-size:xx-large;"><strong>Types of changes (p. 266f):</strong></span>
<br><br><br>
<ul style="font-size:xx-large;text-indent:-40px;list-style-type: none">
<li class="fragment" data-fragment-index="1">slow (gentrification of neighborhoods, growth of vegetation)</li>
<li class="fragment" data-fragment-index="2">sudden (weather fluctuations)</li>
<li class="fragment" data-fragment-index="3">regular (seasonal in plants, animal migration, sunrises)</li>
<li class="fragment" data-fragment-index="4">frequent (presence of wildlife, wind, sounds)</li>
<li class="fragment" data-fragment-index="4">infrequent (fire, floods)</li>
<li class="fragment" data-fragment-index="4">long duration (buildings, roads, consequences of natural disasters)</li>
<li class="fragment" data-fragment-index="4">medium duration (harvesting of trees, seasons)</li>
<li class="fragment" data-fragment-index="4">ephemeral-periodic (wildlife, weather, hiking, evidence of other hikers)</li>
<li class="fragment" data-fragment-index="4">...</li>
</ul>
</span>

Notes:
- some examples of types of changes
- mixture of natural landscape characteristics, flora and fauna
- but also human-related changes

---

<strong>User-contributed geodata & social media</strong>

A complementary data source for assessing subjective landscape values

Dunkel (2015). [DOI](https://doi.org/10.1016/j.landurbplan.2015.02.022)

Notes:
- personal research
- focused on public geosocial media data
- supplementary source for subjective landscape valuation

---

Preprocessing steps:

<ul>
<li class="fragment" data-fragment-index="1">query data from Application Programming Interfaces (APIs)</li>
<li class="fragment" data-fragment-index="2">data cleansing, transformation into a common data scheme, <br><img src="images/logo.svg" width="50" style="bottom:0"> LBSN Structure <sup><a href="https://lbsn.vgiscience.org" class="smallsuper fa fa-external-link"></a></sup></li>
<li class="fragment" data-fragment-index="3">privacy-preserving & ethical data processing, Dunkel et al. (2020). <a href="https://doi.org/10.3390/ijgi9100607">DOI</a></li>
<li class="fragment" data-fragment-index="4">aggregation, visualization</li>
</ul>

Notes:
- several preprocessing steps necessary
- not all covered here

---

Dunkel et al. (2023).  
_From Sunrise to Sunset: Exploring Landscape Preference through Global Reactions to Ephemeral Events Captured in Georeferenced Social Media._  
PLOS ONE 18 (2): e0280423.  
[DOI](https://doi.org/10.1371/journal.pone.0280423).

Notes:
- focus here: temporal visualizations
- precursor study

---

Reliably identify landscape values, <br>independent of overall visitation frequency?

Note:
- main question of precursor study

---

Signed Chi

1. A generic query (expected; e.g. all Flickr images)  
2. A specific query (observed; e.g. "sunset"-reactions)

Visvalingam (1978). The Signed Chi-Square Measure for Mapping. Cartographic Journal 15:
93–98. [DOI](https://doi.org/10.1179/caj.1978.15.2.93)
</p> 
</div>
</div>

Notes:
This equation was originally developed by
Visvalingam (1978), UK Census Unit, to visualize over- and underrepresentation in spatial data.

---

<div style="height:99vh">
<div style="position:absolute;text-align:right;bottom:100px;right:50px;z-index:9999;font-size:x-large;">
<p style="vertical-align: bottom;background: rgba(255, 255, 255, 0.0); color:lightgrey; " >
Sunset in Oia, Santorini <br>
CC BY-SA 2.0 Pedro Szekely/Wikipedia
</p> 
</div>
</div>

Notes:
- over- and underrepresentation of reactions to the sunset and sunrise worldwide
- datasets: 700 Million georeferenced Flickr and Instagram photographs

---

<div style="position:relative;top:550px;z-index:9999">
<p style="vertical-align: bottom;text-align:center" >
▶ consensus among the top-ranked 20% of choices
</p> 
</div>
</div>

Notes:
- striking commonalities, despite the fundamental differences between these two networks
- consensus among the top-ranked places to view the sunset and sunrise worldwide
- some interesting differences: Burning Man example

---

<img class="plain" style="height:500px;" src="images/01_sel_2.png"/>
<br>
Relative ranking of places (sunset and sunrise) for Central Europe.  
Source: Instagram

Notes:
- different form of visualization
- confirm expectations: east/west coast patterns

---

<img class="plain" style="width:80%;" src="images/indonesia.png"/>
<br>
Relative ranking of places (sunset and sunrise) for Indonesia.  
Source: Instagram

Notes:
- identify surprising trends, e.g. Sunrise tours in Indonesia

---

<div style="height:99vh">
<div style="position:absolute;text-align:right;bottom:100px;right:50px;z-index:9999;font-size:x-large;">
<p style="vertical-align: bottom;background: rgba(255, 255, 255, 0.0); color:lightgrey; " >
"Sunrise Watchers, Mount Batur, Bali" <br>
Howard Somerville/Flickr
</p> 
</div>
</div>

Notes:
- obvious: data and behaviour is heavily influenced by online communication trends

---

<p style="z-index:9999;width:100%;position:relative;top:50px;"><span class="highlight">Global Spread of Information</span><br>
<span class="small fragment">Instagram "Mass invasions", Øian et al. (2018), p.1229 <sup><a href="https://doi.org/10.6027/TN2018-534" class="smallsuper fa fa-external-link"></a></sup></span></p>
<br>
<div style="text-align:center;height:500px;z-index:9998">
    <img class="plain" style="width:100%;height:auto;left:50%;" src="images/comparison_pois.png"/>
</div>

<div style="width:400px;position:absolute;top:400px;right:0px;text-align:right;z-index:9999;background-color:white">
<em><p style="vertical-align: bottom" class="small">
Frequentation of popular </br>vantage points in Europe </br>from Instagram data </br>Claudia Tautenhahn, </br>Master thesis, TUD 2019/20
</p></em>
</div>
</div>

Notes:
- global spread of information also has negative effects on landscapes
- examples: overfrequentation at popular tourist spots in Europe
- Dark Hedges, known from Game of Thrones
- seasonal periodicity + trends

## Event Inventories?

Catalog events and temporal valuation patterns  
for places and regions

Notes:
- question for this "scoping study"
- catalog events and temporal patterns for regions

---

Two key tasks:
1. Reduce <span class="highlight">bias</span> in the data to increase representativeness
2. Increase <span class="highlight">precision</span>, to derive more reliable and useful inferences

<p class="fragment" data-fragment-index="5">Rapacciuolo et al. (2021), p.1229 <sup><a href="https://doi.org/10.1111/oik.08215" class="smallsuper fa fa-external-link"></a></sup></p>

Notes:
- publication for biodiversity monitoring based on crowdsourced online data
- bias and precision often inversely related

---

Reduce bias:
- through aggregation
- for example, by sampling from a larger, more representative number of observers and by integrating
  data from different platforms

Notes:
- e.g. combining data from multiple sources
- complicates data analysis due to differences in data collection

---

Increase precision:
- through filtering
- for example, by querying based on time, space and semantics

Notes:
- filtering will reduce available data and thus, representativity

---

<div style="height:99vh;">
<br>
<strong>Study 1</strong> <sup><a href="https://kartographie.geo.tu-dresden.de/ad/wip/ephemeral_events/html/05_hotspots.html" class="smallsuper fa fa-external-link"></a></sup><br>
<img class="plain" style="width:auto;height:40%;left:50%" src="images/overview_hotspots.svg"/>

<div style="position:relative;top:-50px;z-index:9999">
<p style="vertical-align: bottom;text-align:center" >
30 Biodiversity Hotspots in Germany<br>
Data Source: Instagram, Flickr, Twitter, iNaturalist<br>
</p> 
</div>
</div>

Notes:
- primary goal: reduce bias, increase representativeness
- no filtering, except space

---

</div>

Notes:
- absolute measurements of observed user activities
- striking differences: iNaturalist (Hikers, Nature lovers); 
- areas popular among families in summer, etc.
- corroboration

---

<div style="height:99vh;">
<br>
<strong>Study 2</strong> <sup><a href="https://kartographie.geo.tu-dresden.de/ad/wip/ephemeral_events/html/02_reddit_nationalparks.html" class="smallsuper fa fa-external-link"></a></sup><br>
<img class="plain" style="width:80%;height:auto;left:50%" src="images/barplot_nationalparks.png"/>

<div style="position:relative;top:-50px;z-index:9999">
<p style="vertical-align: bottom;text-align:center" >
20 US Nationalparks <br>
Data Source: Reddit<br>
Comments (red) & submissions (blue)
</p> 
</div>
</div>

Notes:
- Data availability: Georeferences available at different granularities
- single data source, limited representativeness

---

<div style="position:relative;top:-100px;z-index:9999">
<p style="vertical-align: bottom;text-align:center" >
Relative seasonal communication patterns (Joyplot)
</p> 
</div>
</div>

Notes:
Mountain peaks are used as a
metaphor for the volume of monthly patterns that deviate from the norm (the average
monthly frequency for each park), and distance/order and absolute height of the
peaks are used to symbolize the total available volume of data, which increases with
distance

---

<div style="height:99vh;">
<br>
<strong>Study 3</strong> <sup><a href="https://kartographie.geo.tu-dresden.de/ad/wip/ephemeral_events/html/04_milvus_maps.html" class="smallsuper fa fa-external-link"></a></sup><br>
<img class="plain" style="width:50%;height:auto;object-fit:contain;left:50%" src="images/milvus.png"/>

<div style="position:relative;top:-50px;z-index:9999">
<p style="vertical-align: bottom;text-align:center" >
Observations of the red kite (Milvus milvus) in Europe <br>
Data Source: iNaturalist<br>
2007-2022
</p> 
</div>
</div>

Notes:
- identify long-term temporal trends for specific ephemeral values
- filtering significantly reduces the available variance, 
  richness, and representativeness of the data

---

<div style="position:relative;top:-50px;z-index:9999">
<p style="vertical-align: bottom;text-align:center" >
Temporal signed-chi evaluation for three regions
</p> 
</div>
</div>

Notes:
- expected trends (all photographs) vs. observed trends
- difficult to interpret
- shifts in platform use and community fluctuations
- recommendation: benchmark species; or benchmark group

<div style="height:99vh;">
<br>
<strong>Outlook: Towards <em>Event Inventories</em></strong><br>

<span class="small">Poor data availability</span>

<div style="position:relative;top:40%;z-index:9999">
<p style="vertical-align: bottom;text-align:center" >
Wikidata SPARQL Query: Events <sup><a href="https://kartographie.geo.tu-dresden.de/ad/wip/ephemeral_events/html/03_wikidata_event_query_nevada.html" class="smallsuper fa fa-external-link"></a></sup>
</p> 
</div>
</div>

Notes:
- combine different data sources to build Event Inventories, e.g. Wikidata
- very little available: town fairs, festivals, but also wildfires and mass shootings
- includes personal observations
- geosocial media data helps to underpin observations with data

---

<div style="height:99vh;">
<br>
<strong>Outlook: Towards <em>Event Inventories</em></strong><br>

What works well
<ul style="font-size:xx-large;text-align:left;width:80%;">
<li>visualization of periodic, recurrent, and seasonal patterns</li>
<li>small scale contexts (regional, continental or global)</li>
<li>well-defined events</li>
</ul>

Challenges
<ul style="font-size:xx-large;text-align:left;width:80%;">
<li>varying specificity as a key challenge in capturing landscape change through user-generated content</li>
<li>fluctuating user groups and platform incentives</li>
<li>data retrieval, API access</li>
</ul>
</span>

</div>

---

Thank You