Climate variability factors
This project was carried out as part of the TechLabs “Digital Shaper Program” in Aachen (Winter Term 2021/2022)
1. Introduction
“It is quite impossible to avoid a drought from happening, but we can avoid a drought turning into famine or displacement of people.” The former Director-General of the Food and Agriculture Organisation of the United Nations, Graziano da Silva, opened the 2nd International Seminar on Drought and Agriculture (FAO 2019) with this quote. In Rome in 2019, during the World Day to Combat Desertification and Drought 2019, potential hazards and damages already caused by land degradation were discussed. Desert regions like the Sahara are often associated with famine and drought, but those may also pose substantial problems for Europe. In 2020, eleven EU countries registered to decrease in agricultural production numbers. Many of them were the biggest producers like France, Germany and the Netherlands (Eurostat 2021). Those tendencies are not new nor unforeseeable for the European countries, since the same trend of decreasing agricultural harvest goods has occurred over the past ten years and is expected to continue for the upcoming decade (European Commission 2021).
In the ongoing discussions about less meat-based consumption and overproduction in the food industry, this trend may seem pleasing. In contrast to this, the numbers probably did not decline intentionally, but as a direct consequence of climate change. Since rising temperatures cause lower water availability for plants and inhibit the soil quality, fieldwork is heavily affected by global climate change. As stated by Graziano da Silva, a hard hit on the agricultural economy of a country may cause serious problems in regards to feeding entire nations.
This blog post is going to raise the status-quo of Europe’s agricultural well-being and will analyse the dependencies of climate, agriculture and humankind. This follows the motto “Making climate change visible”. The methods to approach those problems come from the field of Data Science, so the findings will be exclusively numeric. The main aspect of this work was the production of meaningful visualisations to gain new perspectives on climate change and especially its connection to agriculture.
2. Gaining Data
In the following, the methodological approach of this work will be explained. Since climate change as a topic itself is too rich in information to process nearly everything, the reader should gain a basic understanding of the selection process for the data used in this project.
2.1 Finding the right research area
Before actually collecting the data the geographical scale and region of this study design had to be discussed. It was decided to reference the European continent and work on the national scale. This brings some major advantages:
1. The statistical coverage of Europe is excellent due to the work of the European Commission. This advantage makes Europe far more favourable to use for this study than e.g. Africa or South America, where it would be really hard to gain consistent, generally applicable data for most of the countries.
2. Next to the omnipresent data of Europe it is well suitable due to its geography. By studying Europe on the one hand you have a pluralistic system with far over 30 countries of different sizes, developmental phases and market power. On the other hand — since the main topic of this work is climate change — you have climatological diversity in Europe. You can compare economic giants like France or Germany with their smaller EU partners and by doing so you can implement the climatological dimension, which gives completely different results for Sweden in the North and Greece in the South of Europe.
3. Finding an appropriate scale was rather hard because there are obvious disadvantages to each scale you could choose. For example, going very small with the NUTS2-Level, which is supported by most but not all of the EU-wide surveys would generate high-resolution information, which would account for a more detailed perspective. Unfortunately, most of the smaller countries do not contribute their data for this level, so choosing NUTS-2 would shrink the country list a lot. Therefore we decided to trade the smaller resolution for a bigger contingent of countries to study and research on the national scale. This has the downside, that regional disparities in big countries get cancelled out and the whole country is represented by only one number.
Nevertheless, since climate change is a global phenomenon the goal was to maximise the area covered by this project. Below you can see a map of all 22 European countries that are included:
2.2 Generating the Air-Pollution-Dataset
The scientific field of Climate Variability Factors consists of a broad variety of possible data sources describing different factors. Since air pollution is closely related to both climate change and agriculture and has a very direct impact on human and animal health, an attempt was made to compile a data set of the most important air pollutants. For the following pollutants weekly data was collected: — NO2: Nitrogen oxide, which is produced for example by combustion in motor vehicles — SO2: Sulfur oxide, which is produced, for example, by the combustion of oil and coal — CO: carbon monoxide, which is produced by combustion in motor vehicles — PM10 and PM2.5: Particles with a maximum diameter of 10 and 2.5 micrometers, which are produced by humans or by erosion in the form of dust, etc. — O3: Ozone, as a result of human emissions, for example, sulfur oxides or organic solvents SOURCE: umweltbundesamt.de
The large-scale collection of air pollution data has only been practiced in Europe since the beginning of 2014, which is why air pollution data refer to the period from the beginning of 2014 to the end of 2021. The data were extracted from https://discomap.eea.europa.eu/map/fme/AirQualityExport.htm. This required downloading several thousands of .csv-files and assembling them correctly without losing any data. The downloading of the .csv-files was handled with the request module of python, which allows downloading a file by giving the corresponding URL. With this module and several loops, it was possible to gather a huge amount of individual .csv-files. A dataset was created for each of the 40 European countries for each of the 6 pollutants. The several gigabytes of data now had to be reduced to the necessary information. The data sets were cleaned by deleting unnecessary columns and removing erroneous values and outliers. After the data sets had been cleaned, they could be merged into one data set per country and then into a common data set for Europe.
With this data set for air pollution, graphs could be created showing the different characteristics and development of the air pollutants SO2, NO2, CO, PM10, PM2.5, and O3. It distinguishes between a) the countries and b) the different types of regional settings (rural, urban, suburban).
2.3 Assembling further data
In addition to the air pollution data, several other statistics from the public, and official sources were collected. The main contributor was the website faostat.org, which publishes all kinds of food- and agriculture-related data including relevant information in the context of climate change. Since the research was all topic-based, the final dataset to work with had to be created first. The downloadable data were all fractions of the final dataset, so every table had to be converted to CSV format and then get modified.
One difficulty in bringing the tables together was the unequal timeframe for all tables. The solution for this was adding the specific year of the specific data as the suffix (e.g. ‘erosion_2005’). Whilst merging all datasets based on the country’s name you create a big, hardly plottable dataset. Converting it from wide- to long-format makes the final dataset far easier to work with since every category (e.g. ‘erosion’) is now a unique column (‘stubname’-argument) and the year-suffix gets included as a subindex via the j-argument of Pandas’ wide_to_long-function.
With merging all the tables in the first place, it was secured to minimize missing data. If one country was not included in a dataset and you try to merge based on ‘country_name’, this country got excluded from the overall analysis. This explains why countries like Germany and France are not in the final dataset. This secures comparability of our graphs because no country has a black spot of data for a whole column and therefore at any given plot all 22 countries get visualized.
3. Results:
In the following chapter, the main results of the visualization-based analysis will be presented. First, the graphs will be described and afterwards, some context regarding climate change and agriculture will be given.
3.1 Anomaly and Sea level:
To start this EDA off, by what degree global changes due to rising temperatures may affect agriculture. Two suitable variability factors to describe the climatic status quo are the actual temperature anomaly and the sea level rise: Figure 1 shows the development of the temperature anomalies over the last 30 years (one ‘climate cycle’). The anomaly describes the offset of a 20-year-median, so the bigger the offset, the warmer the year in comparison to the surrounding years. You can see a rising trend over the years with their maxima in the last five years.
This trend does not depend on the European region, since the temperatures in all continental parts of Europe during the observed time (Fig.2). All four regression lines rise strongly positively, which indicates the overall trend. Eastern Europe stands out as the region with the steepest incline.
One parameter directly linked to rising temperatures is the global mean sea level. This again addresses climate change as a global process, since the most molten ice mass comes from the poles but directly influences every country’s shoreline. Figure 3 shows the development of the global mean sea level with applied global isostatic adjustment, which means the global differences in this planet’s gravity field are accounted for when measuring the sea level. The referential timespan for the calculation of the mean is again 20 years. Aligning to the temperature anomalies you can see a straight trend indicating sea-level rise.
A rising sea level is problematic for a) the fresh-water reserves of groundwater and b) preserving land-area of the coastal regions. This is an extreme danger for topographically low-levelled countries e.g. the Netherlands. Following the current trends of sea-level rise, one-third of the Netherlands’ landmass is expected to fall below sea level, which threatens the future of several million inhabitants. ## 3.2 Exploratory Data Analysis: After gathering different, climate-related datasets, the first step of analysis demands getting an overview of your dataset. A very compact and insightful graph for this task is a heat-map. Figure 4 shows the overall heat-maps containing all used climate variability factors.
By distinguishing between the continental regions you can gain a more detailed overview of the independent developments. By displaying the Pearson correlation coefficient you can assume the strength of relationships of factors based on the heat-maps.
For example, the relation between irrigated and irrigable land area shows a strong correlation with coefficients above 0,9 for all regions. This seems to be an interregional similarity. Besides those similarities, you can easily see the significant differences. While irrigated & irrigable land and landcover erosion appear positively correlated in Southern Europe, it is the direct opposite for North-European and West-European countries. A similar difference occurs in the comparison of Northern Europe and Eastern Europe regarding the relationship between agricultural production and irrigated & irrigable land so you can see reversed trends here as well.
Summing up the findings of the heat-maps the European regions can be clustered into two groups. On the one hand, you have Western Europe and Northern Europe, which contain economically advanced and wealthy countries, that do not rely on their agricultural production to be the backbone of their economy. On the other hand, you have Eastern Europe and Southern Europe, for whom the agricultural economy is far more important (Figure 5).
Take for example Spain and which is strawberry production, Greece’s olive plantages or Europe’s biggest farmer for Apples, Poland. Those production sites are far bigger than in North- and West Europe. Overall a specialisation in agrarian-economy can be either an advantage or disadvantage in the combat against climate change. Those tendencies will be explored via detailed factor analysis.
3.3 Agriculture and water:
One of the relationships, that can already be seen on the heat-map, is the one between irrigated and irrigable land. Figure 6 shows the developments for each country over the time of eleven years. The trend goes with a decline in irrigated areas and a corresponding increase in irrigable areas. Keeping in mind the heatwaves of the years 2016 and 2017, which had a devastating impact on agriculture, it is very likely this shows a negative trend.
While you could say it is good that less land is irrigated, it is very likely that this only occurs due to an absolute growth of farming land and problems with the water supply. Because the temperature increases as shown via the anomaly more area could or should be irrigated (an increase of irrigable land), but the water shortages that are reported more frequently over the past years for Southern countries reduce the actual irrigated area (decrease of irrigated land).
This stresses the importance of water as a valuable resource in Europe and maybe a foreshadowing for future crises. Water in agriculture is needed to keep soils fertile and to keep the plants growing. Problems with the water supply during the farming season would have horrendous impacts on food prices. A desirable trend would be an increase in water efficiency. Figure 7 gives an overview of the ongoing tendencies:
You can see for every single country in this dataset, that water exploitation has increased over the last three decades. Additionally, this graph supports the clustering between the regions, since you can see that both the production numbers and the water exploitation index are the highest in Southern and Eastern Europe and the lowest in Northern and Western Europe.
Overall, this indicates a rather problematic future progression for those agrar-heavy countries. Although they could handle temporarily harsh weather conditions probably better because of their bigger capacities, longer periods of heat and drought might be a severe danger for their national economies. You can see this dependency in the heat-map of Eastern Europe, where the real Gross Domestic Product per capita and the agricultural production are directly correlated.
3.4 Analysis of several air pollutants:
Another big field of climate variability factors is air pollutants. Those can both be a consequence or a catalyser of climate change. Too high concentrations result in a harmful environment for humans and animals. In the first graph, the weekly mean concentration of the different pollutants from all over Europe is divided according to the region, in which the data was collected. A distinction is made between urban, suburban, and rural to see if the concentration of the air pollutants shows a regional dependency. The observed air pollutants are SO2, NO2, CO, PM10, PM2.5 and CO.
In the following graphs, the concentration in the urban environment is higher than in the rural environment. One reason for this could be that in cities more people congregate in a small area and therefore more emissions of any kind occur. SO2, NO2, and CO emerge particularly strongly in the combustion reaction of automobiles, which is why there is increased emission of these pollutants in the general car-rich cities of Europe. Compared to this, the density of this emittance is lower in rural areas. PM10 and PM2.5 are a consequence of the remaining pollutants, which are not always present as a gas, but also as particles. Thus, increased SO2, NO2, and CO concentrations are accompanied by increased PM10 and PM2.5 concentrations. In addition, the large number of people creates increased turbulence in the atmosphere near the ground, which in turn causes particles to deposit more slowly on the ground and thus remain in the air longer, where the concentration rises.
The only pollutant not yet discussed in this context is ozone. Here it is noticeable that the ozone concentration is the only one that is not highest in urban regions but rural regions. Ozone is mainly formed as a secondary product of other emissions under intense solar radiation. One of the relevant pollutants is nitrogen oxides (e.g., NO2). One source of NO2 is fertiliser, which is used on agricultural land. Since here, in addition to nitrogen oxides, there is intense solar radiation due to the large open areas in agriculture, ozone is more likely to be generated in rural areas than in cities. In cities, solar radiation is strongly impaired by buildings, etc. which is why less ozone is produced here. SOURCE: umweltbundesamt.de
Besides the different expressions due to the region, the course of the concentrations of the pollutants follows a certain cycle. At the end of each year, a maximum can be seen. Since the emissions of the main sources, such as traffic or energy production, are approximately constant during the year, there must be other reasons for this course.
One reason for the concentration maxima in winter is that there is less protective vegetation on the ground in winter, so more erosion can occur. As a result, more dust is carried into the atmosphere from the open areas, which consists of previously emitted Air Pollutants. This dust and the substances emitted from the ground cause the concentration of Air Pollutants (SO2, NO2, CO, PM10, PM2.5) to reach a maximum in winter. It should be mentioned that ozone is the only air pollutant that deviates from this observation. One reason is that for the formation of ozone intensive solar radiation is necessary. This is not the case in Europe in winter. In addition, ozone is a secondary product of the other air pollutants, which is why a time-delayed effect is to be expected. This means that the concentration maximum of ozone can be expected some months after the maxima of the air pollutants. Again, special attention should be paid to NO2 emissions, as this has a significant influence on ozone formation, as mentioned above. SOURCE: klimanavigator.eu and umweltbundesamt.de
To conclude the consideration of these graphs, the limits of the EU for the individual air pollutants are discussed below. The limits here refer to the EU specifications as provided by umweltbundesamt.de. First, it has to be said that the limits are only to be seen as a guideline, because in the graphs the weekly mean concentrations are shown, and the limits often refer to smaller periods. In addition, for many, the air pollutants are a maximum number of days per year on which the limit value may be exceeded. This problem is not apparent in the graphs shown since they are averaged over a week and thus individual days are not shown separately. An example of this is the graph of the weekly mean concentration of SO2 and O3 in Europe because SO2 and O3 do not exceed the limit values for many days in a row and therefore no exceedances can be seen in the weekly mean value. The same applies to CO, which is why the limit value was not plotted here. For the remaining air pollutants NO2, PM10 and PM2.5, however, exceedances of the limit value can also be detected in the weekly mean. For no air pollutants in rural areas, the limit value was exceeded in the weekly mean. In contrast, the limit value is regularly exceeded in urban and suburban areas. This shows that the Air Quality is much worse in cities than in rural areas. SOURCE: umweltbundesamt.de
Since the individual air pollutants have been discussed so far, the correlation between the air pollutants will now be explained in more detail. To show this visually, the relevant part of the heat-map from above with the correlation coefficients was created.
As mentioned before, ozone is a special air pollutant. This can also be seen in the heat-map, as ozone has a negative correlation coefficient to all other Air Pollutants. One interpretation is that ozone is influenced by different mechanisms than the other Air Pollutants and thus shows an opposite trend. The question arises whether a correlation can be seen by adding a time variable since it can be assumed that ozone is formed from for example NO2, but it takes time for these reactions to take place.
For the Air Pollutants PM10 and PM2.5, a strong positive correlation can be seen. This is not surprising since there is an overlap. PM10 includes all particles up to 10 micro-metres in diameter and PM2.5 includes all particles up to 2.5 micro-metres in diameter.
For the air pollutants NO2, SO2, and CO, there is consistently a positive correlation, meaning that all three pollutants have similar trajectories and are determined by similar phenomena.
4. Conclusion and reflection:
The results previously presented and explained show reliable evidence to evaluate the situation of European agriculture as problematic. Whilst some countries still increase their production it seems that they simultaneously strengthen their dependency on their harvest output. This may be crucial in a future world, where hazardous events are expected to appear more frequently and have more devastating impacts. The economically smaller countries of Europe are more vulnerable to these tendencies since they rely on their agricultural economy the most.
The analysis of this work does not indicate a relevant increase in regional resilience, which would be one future-oriented solution to prevent one’s nation from falling apart due to climate change and its consequences. Nonetheless, the selection of climate variability factors that were applied in this study is comparatively small in contrast to the broad variety of parameters climate change can be measured. Since the goal of this study was to show the effects of climate change, the choice of factors might have been biased by the researchers, which accounts for the solely negative evaluation of the status quo.
Overall, this project has pointed out the developments of the past climate period and stresses the need to work properly and fast in the fight against climate change. Due to the categorisation of continental regions and the distinction between rural. urban and suburban surroundings some detailed insights have been accomplished, that show the regional disparities and differences of a global occasion — climate change.
5. Future Work / Outlook:
Many of the reasons for the characteristics and behaviour of the Climate Variability Factors mentioned could not yet be examined in detail during this project. However, the existing data provide a good basis to still search for numerous correlations and quantify them with data. One idea already mentioned would be to get to the bottom of the relationship between ozone and NO2, since it is expected that these two air pollutants are related, and this has not yet been shown. The idea here would be to include a time variable to see if after a certain period after an increase in NO2 emissions there is an increase in ozone emissions. In addition, it should be mentioned that due to the broad subject area, the resolution of the data sets had to be greatly reduced. It is conceivable that from the air pollution datasets alone a separate project could be developed, in which not only the weekly mean concentration but also small-time windows could be considered. It would be interesting to know, for example, how often a certain limit value is exceeded per year. Because this can also cause damage to humans, animals, and the environment.
In addition to the pure visualisation, the next step could be to try to predict the development of the various climate variability factors for the future with the help of linear regression. In this way, a statement could be made about the effects that can be expected if no changes are made to people’s lifestyles today.
6. References:
[1] European Commission (2021): Agricultural output of EU down by 1% in 2020. Available under: https://ec.europa.eu/eurostat/web/products-eurostat-news/-/ddn-20211115 -2#:~:text=In%202020%2C%2011%20out%20of,Netherlands%20(both%20%2D 3.1%25) (last visit: 02.04.2022).
[2] European Commission (2022): EU agricultural outlook 2021–31: lower demand for feed to impact arable crops. Available under: https://ec.europa.eu/info/news/eu-agricultural-outlook-2021-31-lower-demand-feed-impact-arable-crops-2021-dec-09_en (last visit 31.03.2022).
[3] Food and Agriculture Organisation of the United Nations (2019): Unlocking the potential of agricultural innovation to improve farmers’ resilience to drought. Available under: http://www.fao.org/news/story/en/item/1198259/icode/ (last visit 02.04.2022).
[4] umweltbundesamt.de
[5] klimanavigator.eu
TechLabs Aachen e.V. reserves the right not to be responsible for the topicality, correctness, completeness or quality of the information provided. All references are made to the best of the authors’ knowledge and belief. If, contrary to expectation, a violation of copyright law should occur, please contact journey.ac@techlabs.org so that the corresponding item can be removed.
