Analytics Vidhya
Published in

Analytics Vidhya

Analysing the Influence of Meteorological Features and Human Activities in the Occurrence of Forest Fires

What are the main meteorological predictors that influence forest fires? How do human activities affect wildfires?

This article will provide a synthesis of human activities, metrological predictors and conclude with how data mining is being applied in detecting and eventual prevention of forest fires.

It is undeniable that human activities affect the environments is ways that science is able to prove causalities between climate change and our daily behaviours. While political subjects continue to downplay these interactions to guard the massive profits leading to higher GDPs for their countries.

The timeline graph below correlates human activity with the occurrence of global fires (NASA 2010). The start of the industrial revolution marks the start of a surge in fire activity.

Picture Courtesy of NASA

Fire activity based on charcoal records for the last millennium (shown in gray), this closely matches the wildfire model (illustrated in red) with the charcoal record. The combination of climate model suggests that global fire activity is likely to increase by between 5 and 35 percent depending on other predictors. Future greenhouse gas emissions, development patterns, and demographics are all significant players in this.

“There is causality between human activity with climate change thereby affecting the occurrence of forest fires.”

Prior to the world going in an apocalyptic mode (covid-19), a familiar place in Stephen King meet Alfred Hitchcock script, we witnessed the Amazon rainforest burn. Soon after that we had more than 50 fires in New South Wales and Victoria in Australia. This record fire is speculated to have killed over a billion animals leaving many species in severe jeopardy. Reports from the fires in 2020 in the United States have led biologists estimate the fires have killed 50% of the state’s endangered pygmy rabbits, which inhabit sagebrush flats that burned this year. They believe only about 50 of North America’s smallest rabbit remain. Officials estimate the flames have also killed 30% to 70% of the state’s sage grouse and sharp-tailed grouse, birds that also depend on sagebrush.

Picture Courtesy of BBC

A koala rescued from a forest fire in Australia, unfortunately a lot of his species were lost.

The map of forest fires shows the scare from low to extremely high with over 200,000 fires detection over a period of 12 months. South America, China, Russia, Central Africa, Australia and North America all face high damages due to the frequency of blazes.

Picture Courtesy of Fortune

The largest rainforest, the Amazon has faced serious damaged due to forest fires with over 9,000 km2 burnt area. The cost of recovery is almost a trillion dollars. The cause is the slash-and-burn approach to deforest land in preparation for agricultural endeavours in 2019. Brazil’s National Institute for Space Research said that the number of fires in the country — largely set by humans — had jumped 84% this year over the same period in 2018. The amount of Amazon forest cover lost in Brazil in that span spiked 39%. Greenpeace estimates that massive blazes in Siberia this year have released more than 166 metric tons of CO2, nearly equal to the annual emissions of 36 million cars.

Metrological Factors Affecting Forest Fires

Major environmental issues caused by fire fires create ecological and economical damage as well endanger human lives. struggle with understanding and detecting meteorological conditions such as temperature, wind, soil moisture, tree density, shrubs and other potential fuel sources. According for Center for Climate and Energy Solutions climate change has been a partly responsible for increasing the risk and extent of wildfires in the Western part of United States. The by-products of climate changes the creation of organic fuels which burn and help spread wildfires.

Case Study I: Weather and Human Impacts on Forest Fires

A century-long study carried out by (Zumbrunnen et al., 2011) based on the most fire-prone regions of Switzerland. Consideration of land use and meteorological data alongside forest fires for a period ranging between 1904 to 2008. The mountain canton states of Valais and Ticino where observed as they possessed distinct climatic conditions. The availability of fire ignitions was analysed, road density was considered as an ignition source, livestock density and the change in forest area that is fire-prone. This study found that while road and livestock densities influence in the two cantons, the increase in forest area was well correlated with fire occurrence in Ticino but not in Valais. Land abandonment and forest cover change being less may have led to less extensive fire occurrence.

The figure above shows the changes in the predictors for the entire study area (Valais and Ticino) from 1904 to 2008: Nesterov ignition index (10-year moving average), road density, change in forest area (during the preceding decade), and livestock density. The reference surface for calculating the road and livestock densities corresponds to the area in the canton under 2500m a.s.l.

A conceptual framework for analysing the potential impact of human and weather on fire recurrence is shown below. The assumption is that variables with most prediction powers are ignition, fuel load and fire weather.

Conceptual model of the potential impact of human factors and weather on fire occurrence related to ignition and fuel conditions.

The main emphasis of the result project a non-linear nature of the relationships between fire occurrence and anthropogenic drivers. No thresholds above which road density was no longer correlated with fire occurrence. Therefore, the predicted increase in human population in both cantons will not result in higher risk in forest fire outbreaks. To conclude, the study presents a case that the impact increased human activities and populations in the two cantons would have led to a different result if the climate warmer. Hence the need to not refute the correlation between human activities to forest fires is not the outcome from the study.

Case Study II: Potential Influence of Meteorological Variables on Forest Fire Risk in Serbia

This recent publication focuses on establishing the relationship between meteorological variables and forest fires indices in Serbia (Tosic et al., 2019). Factors considered were daily temperature, precipitation, relative humidity and wind speed data from 15 meteorological stations across the country. The results provided are evident that interannual variability in climate has impact on fire activity. The increase of forest fires in Serbia between 2000–2012 follows the patterns observed in Mediterranean countries; the Italian Alps and the Spanish Pyreness where forests cover less than 50% were greatly affected.

In conclusion the study clearly presented potential influence of meteorological variables on forest fires in Serbia. The interaction between climate dynamics with high air temperatures, low humidity and a lack of precipitation being the most influential. Table 1 are the results of the regression model.

Modeled (dashed line) and observed (solid line) annual number of forest fires in Belgrade for the evaluation period 2011- 2017: a) using meteorological variables (retained precipitation — P (mm) as predictor, and b) using fire indices (retained L as predictor).

Monthly analysis indicated that the most frequent fires occurred in August, March-April, and September. This trend spiked between 2007 and 2012 with the maximum number of wildfires being 1627. This resulted in over 22,000 hectors. An attribution of high air temperatures in 2007 and prolonged heat waves and the worst ever drought since record times begun.

Table 1: Results from stepwise regression models for number of forest fires in Belgrade with retained predictor (meteorological variables: relative humidity-RH, precipitation-P and wind speed-V, and fire index)

To identify meteorological variables responsible for forest fires a combination of require of input variables in a stepwise regression model for Belgrade. This led to precipitation or relative humidity were found to be the variable with the highest predictor power.

Case Study III: A Data Mining Approach to Predict Forest Fires using Meteorological Data

Using modern tools and techniques to predict forest fires stretches long before we had advanced data analytics with computing and storage resource that only mainframe had two decades ago. With this in mind, it is now more than ever vital to maximise the potential of analytics in encouraging the better ecological outcome. As illustrated in the previous sections, the damage cause by forest fires is monumental, the cost of recovery is beyond any monetary figure.

This study reveals how data mining techniques such as Support Vector Machines (SVM), Random Forest (RF) and four distinct feature selection setups are successfully applied to predicting forest fires (Cortez and Morais, 2007). Using spatial, temporal, Fire Weather Index components and weather attributes, were tested on recent real-world data collected from the northeast region of Portugal. The best configuration uses SVM and four meteorological inputs (i.e. temperature, relative humidity, rain and wind) and it is capable of predicting the burned area of small fires, which are more frequent. Such knowledge is particularly useful for improving firefighting resource management (e.g. prioritising targets for air tankers and ground crews).

Table 2: The predictive results in terms of the MAD errors (RMSE values in parentheses; underline — best model; bold — best within the feature selection)
The REC curves for the M-SVM, M-RF and Naive models (left); and the real values (black dots) and M-SVM predictions (gray dots) along the y−axis output range (right)
Picture courtesy of AAAS

Future Research Areas

This article only touches the surface with regard to meteorological predictors and human-ecological interactions that increase wildfires. A more detailed and globally spread synthesis of these causalities must be done. As acknowledged throughout this article, the loss from these blazes must be a concern of many governments and entities. Climate action policies including moderating land use for logging, agriculture and developmental undertakings must be more stringent. The gap between forest preservation and civil actors must be resolved.

Conclusion

Environmental damaged caused by forest fires is colossal, having means to prevent and detect potential fires is paramount. Substantial effort in automating the detection that aid Fire Management System (FMS). Three major trends using satellite data, infrared scanned and local sensors with metrological data have created a great base ground to apply data mining techniques in implementing real-time fire warning systems. As seen in the final case study, a regression model using SVM and feature selection achieved exemplary results.

According the World Health Organisation, wildfires are usually caused by human activities or an act of God such as lightning and can happen at random. However, in 50% of wildfire cases recorded, the cause remains obscure. Wildfire risks increases exponentially in dry conditions with drought-light conditions or during high winds. The impact of wildfires varies from disrupting transportation, communications, power and water supply. The deterioration of air quality affect both animals and humans. A combination of wildfires and volcanic activities affected over 6 million people between 1998–2017 with 2400 attributed to global fatalities from suffocation, injuries and burns. The frequency of wildfires has grown with the increase in the meteorological effects of climate change.

References

1. Cortez, P. and Morais, A. (2007) ‘A Data Mining Approach to Predict Forest Fires using Meteorological Data’, Proceedings of 13th Portugese Conference on Artificial Intelligence, pp. 512–523. Available at: http://www.dsi.uminho.pt/~pcortez/fires.pdf.

2. Tosic, I. et al. (2019) ‘Potential influence of meteorological variables on forest fire risk in Serbia during the period 2000–2017’, Open Geosciences, 11(1), pp. 414–425. doi: 10.1515/geo-2019–0033.

3. Zumbrunnen, T. et al. (2011) ‘Weather and human impacts on forest fires: 100 years of fire history in two climatic regions of Switzerland’, Forest Ecology and Management. Elsevier B.V., 261(12), pp. 2188–2199. doi: 10.1016/j.foreco.2010.10.009.

--

--

Analytics Vidhya is a community of Analytics and Data Science professionals. We are building the next-gen data science ecosystem https://www.analyticsvidhya.com

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Chibili Mugala

A nerdy data scientist with a passion for artificial intelligence, computer vision & autonomous vehicles. https://linktr.ee/qibili