The power of prediction with social media
Summary (2 min read)
Sea level change
- Global mean sea level is rising at a rate of approximately 3.2 millimeters per year (Church et al. ).
- Direct consequences of sea level rise on coastal areas include an increase in flooding area, an increase in erosion, and an increase in salinity and changes in ecosystems (Nicholls et al. ).
- Estuarine circulation is mainly driven by freshwater flow, tides, and density differences (Garel et al. ).A studybyChua&Xu founda stronger longitudinal salinity gradient in estuaries due to sea level rise, which in turn drives a stronger gravitational circulation.
- The increase in salinity will cause the water to become denser, and thus increase the stratification of the water column.
- Changes in estuarine stratification and circulation will further cause oxygen depletion (Hong & Shen ).
Physical characteristics of the Guadiana Estuary
- The Guadiana Estuary is formed at the interface of the Guadiana River and the Gulf of Cadiz.
- This large reservoir was completed in 2002 and since then the freshwater flow into the estuary has been reduced from a yearly average of 143 to 16 m3/s (Garel et al. ).
- When there is a higher tidal amplitude and lower discharge from the Guadiana River, tidal processes control the water circulation of Downloaded from http by guest on 19 April 2021 the estuary and the estuary becomes well-mixed (Garel & D’Alimonte ).
- The water column is stratified only under extreme conditions (Basos ).
Model setup
- The present study implements the same general setup of the model by Mills et al. and uses the same Cartesian computational grid of 1,400 × 350 cells with a resolution of 30 m.
- Two months of simulation time is required due to the high residence time when the freshwater flow rate is low (Oliveira et al. ).
- The present model consists of two separate bathymetries: (1) a bathymetry in which coastal management strategies are implemented to keep the coastline as it is (Figure 1) and (2) a bathymetry that allows for geomorphological changes caused by sea level rise and thus allows flooding around the estuary (Figure 2).
- The 5th assessment report includes sea level rise forecasts up to the year 2100 for different Representative Concentration Pathways (RCP), more commonly known as greenhouse gas emissions (Church et al. ).
Areas of inundation
- Areas of inundation were computed for the various scenarios of sea level rise and highest freshwater discharge scenario (100 m3/s) over the bathymetry allowing flooding.
- The methodologies follow those of Mills et al. who computed flooding area as a function of the number of hours of land submersion during one tidal cycle, but only for a freshwater discharge of 500 m3/s.
- The present study includes an analysis of flooding area for a river discharge of 100 m3/s as well as each spring and neap tide scenario.
- Histograms and flood distribution maps were computed based on the percentage of time a land cell was covered by water in the grid during one tidal cycle.
Temporal evolution of salinity
- Time series graphs were produced at each location shown in Figure 3, allowing for an assessment of the evolution of salinity over two tidal cycles (approximately 24.48 h) for each sea level rise scenario.
- The average change in salinity every 30 years is summarized in Tables 1 and 2.
- The simulations for salinity distribution assume a salinity value of 0 for freshwater and a value of 36 for seawater.
Horizontal distribution of salinity
- The following section examines the horizontal distribution of salinity along with water velocity direction at a time instant 1 h before high tide.
- As can be seen in the time series results, changes in salinity throughout the scenarios of sea level rise vary for the present bathymetry, whereas results from the alternate bathymetry reveal a correlation between sea level rise and salinity.
- Thus, all horizontal distribution maps of salinity are shown for the present bathymetry, whereas only the present year compared with 2100 are shown for the bathymetry allowing flooding.
- High freshwater discharge at spring tide See Figures 6 and 7.
DISCUSSION
- The results obtained from the MOHID model have shown the dynamics of the Guadiana Estuary to be complex, especially with respect to the tides.
- Areas further upstream of the estuary portray an increase in salinity when the river discharge is low.
- Especially for the present bathymetry, decreases in salinity coincide with decreases in water velocity.
- The results of this model indicate that bathymetry, freshwater flow, and spring-neap tide variability impact the horizontal distribution of salinity intrusion caused by sea level rise.
- In terms of land inundation, all varying hydrodynamic factors result in an increase in inundation due to mean sea level rise.
CONCLUSION
- All results portray an increase in salinity in response to sea level rise.
- When the freshwater flow is low in the spring and summer months, areas located upstream of the estuary increase in salinity.
- A limitation of this work is the use of a two-dimensional model instead of a three-dimensional model.
- Future studies should use a three-dimensional model with real tidal signals Downloaded fr by guest on 19 April 202 to allow for a more complete evaluation of the Guadiana Estuary and how it responds to climate change.
Did you find this useful? Give us your feedback
Citations
532 citations
379 citations
286 citations
265 citations
Cites background from "The power of prediction with social..."
...Stevenson et al. (2000) disclosed that an unfavourable attitude towards online advertising was related to low purchase intention, whereas Wolin et al. (2002) proposed that a favourable attitude towards online advertising usually resulted in more recurrent online purchasing and greater online spending....
[...]
...Hence, a longitudinal approach would yield more complete results, as inferred by Kalampokis et al. (2013) and Schoen et al. (2013)....
[...]
...Stevenson et al. (2000) disclosed that an unfavourable attitude towards online advertising was related to low purchase intention, whereas Wolin et al....
[...]
205 citations
References
22,840 citations
18,802 citations
"The power of prediction with social..." refers background or methods in this paper
...Researchers can unobtrusively approach social media to observe the public’s behavior and then derive their intention or opinion from the observed behavior (e.g., using machine-learning techniques, cf. Bishop, 2006)....
[...]
...Unfortunately, their methods are quite different from those applied by Asur and Huberman and, thus, in the absence of further research, it is difficult to ascertain if and under what conditions consumer patterns can be predicted from social media....
[...]
...Machine learning methods are commonly used to face such tasks and, hence, any researcher or practitioner interested in the area of social media-based forecasting should be familiar with them (cf., Bishop, 2006)....
[...]
7,513 citations
7,452 citations
4,515 citations
Related Papers (5)
Frequently Asked Questions (8)
Q2. What have the authors stated for future works in "The power of prediction with social media" ?
If there is anything that the human experience has taught us is that predicting the future is both highly desirable and extremely difficult. While the authors do not have reasons to doubt it, they are consciously cautious about the validity of their own arguments regarding the future of forecasting using social media data. In the future, one might identify the existence of a new, fourth prediction model that is made possible by the idiosyncrasies of social media.
Q3. What is the importance of reporting on market design issues?
When publishing their results, it is of utmost importance to report decisions concerning market design issues, including resistance to tampering, as they might influence prediction outcomes.
Q4. What are the main challenges of the proposed methods?
The main challenges of the proposed methods lie in finding an automatic way to determine the best keywords and their associated weights to fit user-generated data to the ground truth available for training.
Q5. What is the main reason why this is of interest?
This is of interest because sentiment analysis has become inextricably associated with social media-based prediction –although up to now it has been applied under the form of very simple methods.
Q6. What is the way to assess the reliability of a prediction model?
despite of the model of choice, social media poses problems regarding the quality and credibility of the collected data, and those problems must be addressed with techniques which are independent of the predictive models.
Q7. What is the main focus of the research needed in electoral forecasting?
electoral forecasting from social media is a field where further research is needed, and where the main focus should be put on providing general models that could be applied not to a single election but to multiple elections.
Q8. What is the significance of the research on predicting the stock market from Web data?
Probably reflecting its financial importance, research on predicting the stock market from Web data largely predates the existence of user-generated content.