TELECOUPLING WEBINAR #1: The FABIO model with Dr. Martin Bruckner

Thanks a lot, Julie. Thank you for the opportunity to give this webinar. A very warm welcome also from my side, and I'm really glad that I can present today the Food and Agriculture Biomass Input-Output database to all of you. So what can you expect from this webinar? First, I will start with what is FABIO, then how can it be used by each one of you, then I will show you some current applications that we are working on and give an outlook on further research that will come later. Okay, so let's start with an introduction. What is an input-output table first? FABIO is a set of physical multi-regional input-output tables. What you can see here is an exemplary input-output table. FABIO comprises 191 countries plus rest of world and 130 commodities. This input-output table shows which... let's look at the column, let's say one, at the cattle column of Germany would show the inputs, for example the soybean inputs and the wheat inputs and the feed inputs that are required and from which countries these inputs come from. So that's what the input-output table shows us. FABIO comprises time series of input-output tables starting from 1986 to 2013. As I said, it covers 130 commodities: 127 of them are agricultural commodities, 64 crops, 32 processed products mainly vegetable oils and cakes and sugars and beverages, 14 animal groups, and 17 livestock products such as different kinds of meat, milk, hides, and fats. It currently covers three forestry commodities. It also covers environmental extensions. At the moment, the database includes data for harvested area in hectares, harvested biomass in tons, and blue and green water in million cubic meters. The data sources that we build this database from are mainly FAO data, and from FAO we source many different datasets, but the most important ones are the commodity balances and the bilateral trade data. For biofuels production, we used data from the International Energy Agency and from the Energy Information Administration. For biofuels trade, we used UN Comtrade. For feed use, we built feed balances for all countries worldwide based on data from the IMAGE model from the publication of Wirsenius 2010. Water data is from a continent. Actually, the starting point for this database of tables was to build supply and use tables in physical units. Here you can see a little example for a few commodities: soybeans, soybean oil, soybean cake, cattle, and beef, and how such a supply and use table can look like. You see that soybeans are supplied by the process called soybean production. Soybean oil extraction supplies soybean oil and soybean cake. Cattle husbandry supplies cattle, while cattle slaughtering supplies beef. So that's the supply table. Then we have the use table showing where these amounts of commodities that are supplied are then later used. Soybeans are used in this example as a seed for soybean production, as an input into soybean oil extraction. Then the oil is going to demand. At the very right side of the use table, you see food use, change, and other uses. So these 20 units, let's say tons, 20 tons of soybean oil are used in this example for food and for other uses, which is for example in the chemical industry or as feedstock for biofuel production. Then soybean cake is going into cattle husbandry, cattle is going into slaughtering, and beef is used by final demand for food. So that's how these supply and use tables look like. The full multi-regional supply and use tables for all 192 regions and 130 commodities will also be available online together with... Now I'd like to show how supply and use tables are converted into symmetric input-output tables. Here you can see that now we have a table with products by products. Before we had products in the rows and processes in the columns. Now we have products both in rows and columns. We see soybeans: 5 tons of soybeans going into soybean production, 80 tons into soybean oil, 80 tons into soybean cake. These 80 tons of soybeans that go into soybean cake you can find again in the soybean cake row and the cattle column: 80 tons of soybean cake going into cattle. Then 30 units of cattle (livestock is in thousand heads) go into beef production. This input-output table is using mass allocation, meaning that the soybean input into soybean oil extraction is split among soybean oil and soybean cake according to their mass. FABIO also provides input-output tables based on value allocation, which is shown now below. Here you see that the split of soybeans between soybean cake and soybean oil is now 33 to 67, not 20 to 80. The price of soybean oil is higher than the price of soybean cake, and that's why with value allocation, more soybeans are allocated to soybean oil than before. It was 20%, now it's one third. There would be additional options such as energy allocation, currently not implemented but can be easily implemented, which would then give a different allocation. For example, soybeans would be distributed 40% to soybean oil and 60% to soybean cake. The FABIO database currently comprises multi-regional input-output tables using mass allocation and using value allocation. We will see examples later where you can see how this affects the results. I already mentioned that there are the final demand categories of food use and other uses. Actually, other uses is not really a final demand, but in FAO data it is considered a final demand. It is used... In this example, 10 tons of soybean oil going into other uses would go into the chemical industry and would be further processed and would go into potentially international product supply chains. So it does not end at this place of industrial use. That's why we link FABIO with the monetary MRIO database EXIOBASE. This describes a hybrid multi-regional input-output table with FABIO in the upper left quadrant, EXIOBASE in the right quadrant, and in the upper right quadrant with inputs of agricultural commodities into non-food industries of EXIOBASE. In total, this hybrid IO table has roughly 35,000 rows and columns, so that's quite large. On the next slide, you can see a comparison of different available MRIO databases. Here you see that FABIO is among those with the highest country detail, with 191 countries plus rest of world. It's the one with the highest detail also for agriculture and food products, and also for forestry, although three products is not much yet. We will increase the detail for forestry products later. Of course, FABIO does not capture any other products or services, which are of course captured by all the other MRIO tables. But when combining FABIO with EXIOBASE, we actually can cover 172 non-food industries and services. FABIO is in tons and heads for livestock, as I mentioned, and has a long time series which will also be extended in the future. How you can use FABIO: first, you can go on GitHub. I've provided you some links here where you can find all the R codes that we used to build FABIO, to build a hybrid FABIO-EXIOBASE model. There is also a repository called FOGPA which gives the codes for the structural path analysis that we did to analyze international supply chains of agricultural commodities. We have another repository with IO visualizations, which I will also show you later. The whole database will be published on Zenodo. Here's the link, which is currently not online. We have to do some checks and will then publish the whole database under the GPL license. The FABIO paper is already available online, published in Environmental Science and Technology, and you can also look at FABIO to find the print of global to get some further information. Now I would like to show you our online visualizations. For this purpose, I switch to my browser. I hope you can all see now. This is a Sankey diagram for Indonesian oil palm. This tool is still under construction, but it will be available on fabio.earth. You can simply select a country. You see here the list of 191 countries and rest of world. You can also select a continent. Let's pick Asia-Pacific. Here under 30 products: primary crops, crop products, non-food... Okay, see the full list here. There are the wood products: industrial roundwood, coniferous, non-coniferous. These are the only three forestry products. Here the animal products: cattle, pigs, poultry, edible offal, milk. Let's select hides from Mexico. Then here you also choose a year. It's still under construction as I mentioned, and we will load the whole time series from 1986 to 2013. Then select mass or value allocation. Then the environmental indicator that shows which is the land use embodied in the supply, or the primary biomass, or the product unit. The difference between product unit and biomass is for example in our case we have selected hides and skins. The product unit would be the tons of hides and skins. Biomass is actually the biomass that is embodied in those hides and skins, which is the feed. Now we can see the Sankey diagram here. You see that hides from Mexico are mainly processed in Mexico itself, and only some smaller quantities are traded. They are processed mainly into leather products, which are also consumed domestically, but then also exported to the United States. You can also change the detail that is shown. We have a cutoff value which here is 4%, where you can reduce that to show more detail and show more regions. Otherwise, regions that do not receive a lot would be cut off not to get it too messy. Maybe this tool is to be used for teaching purposes or just to play around. Then I go back to the slides. Now I'd like to show you some of our current applications. We are working on an analysis of international cotton supply chains and the embodied water stress. You see that, which was surprising for us, only one third of the cotton harvest is cotton lint, and two thirds is cottonseed. Cottonseed is used as an animal feed mainly. Most of it is converted into cottonseed oil and cake. This cake is then used as a feed for cattle, poultry, pigs, and so on. But also the seed itself is used as a feed. One third is the cotton lint, or it's 35% of the total harvest, which then goes mainly into the textile sector to produce yarn, apparel, furniture, and so on. This Sankey is made using the FABIO version with mass allocation. Now if we switch to value allocation, you can see that it turns upside down. Now it's two thirds going into non-food, into textiles mainly, and only one third going into food. This is because the lint, although being only one third of the harvested quantity, is two thirds of the value, and the seed has a much lower price than the lint, making only one third of the value. So in this case, two thirds of the water stress would be allocated to textiles and other non-food uses of cotton. Here you see how it changed from mass allocation to value allocation. With mass allocation, a little less than half is going into non-food. With value allocation, it's more than two thirds going into non-food. Another application that we are working on is food waste and diets. A master student, Anna-Lena Fuchs, was doing her thesis on the footprint of French food wastage. In the left upper corner, you can see a figure showing the cropland wastage along the food supply chain at different stages: harvest, storage, transport, processing, distribution, and final consumption. We compared waste percentages given by FAO in an FAO report on food waste with waste found in the literature specific for France, so that's why you see these differences for the yellow and orange bars. You can see that the largest part of the cropland is lost at the consumption stage and at the harvest stage, but that depends a lot on where the products come from. In Europe, the losses at the harvest stage are much smaller than in Africa. On the right hand side, you see this pie chart showing the source of the cropland in French food waste. Most of it is from Europe, but there are 31 square meters wasted by each person in France coming from Latin America, and also 31 coming from Southeast Asia, and so on. In this study, we include only cropland, but also green and blue water, which you see below in this colorful bar chart. Here you see for example that fruits, although their share is 6% in the total wastage, they have a share of 15% in the total blue water losses, while cereals have a relatively high share in the losses of cropland. One more slide showing an application that we are working on, also on food waste combined with diets for Germany. Hannah Helander is doing her thesis work, and she is showing here on this figure what would happen if Germany would switch to the recommended diet. You can see that the biomass footprint would decrease, but actually the quantity of food waste would increase because the waste shares are higher for fruits and vegetables than for meat and milk. So the amount of waste would increase, but its footprint would be reduced still. One final slide on further work that we are currently doing. We are now revising and polishing the codes, and we will make all data available as I mentioned, still this year in the coming months. We will also do a nowcasting of FABIO for the years 2014 until 2016, maybe if possible until 2017. This will all be made available later on. We are also working on an environmental extension for greenhouse gas emissions, also for fertilizer use, nutrient input, and energy use. We will continue going subnational. We want to capture subnational commodity flows. We will start with Brazil and particularly with Mato Grosso next year, and add more countries and regions later on. We will add more wood and paper products. So that's it. Here's again the link where you can find more information about the FABIO database. If you have any questions or queries, please just send me an email. Thanks a lot.

I think that for non-academic uses, especially the online tool will be useful, where people can play around and find out the results for all different products and countries whatever they are interested in. Of course, they are also invited to download the database and do calculations on their own, but I think that probably most non-academic users will rather rely on interactive visualizations that we provide online.

That's a good question. Actually, the FABIO paper presenting the purely physical MRIO was published now. We just need to do some final checks, and then the database will go online hopefully in the next two to six weeks. That's the plan. But that's only for the physical MRIO. The hybrid MRIO, we will not publish the MRIO tables, but we published the codes that can be used together with the FABIO database and the EXIOBASE to build the hybrid tables. Actually, this code is already available on GitHub. So once FABIO is online on Zenodo, everyone can build the hybrid model.

Definitely, yes. We will also provide the Leontief inverse online on Zenodo, so you can simply take the Leontief inverse and multiply with the environmental extension that you are interested in and use it for multiplier analysis.

I personally, for the attribution of responsibility, prefer value allocation because I think that for example in the cotton case, if we would not buy the textiles, nobody would produce the cotton. So the lower value byproduct, which is seeds used for feed purposes, in my opinion should not receive the full two thirds of responsibility as it would receive when using mass allocation. But if you want to study global physical flows of commodities or of embodied energy or nutrients or such, then you should use mass allocation because that's the actual mass flows which you get then.

Absolutely. I would love to collaborate. I'm in contact with some people from SEI, and I hope that we can work together in the future.

Yes, in principle that's possible. You can't really see it in the online visualization, but if you download the database, you can derive these numbers.

That's also a very good question. We haven't done uncertainty analysis yet, and we have not come up with any quantitative measures. There are some parts of FABIO which are more uncertain than others. In our paper, we describe a little bit which parts we would expect higher uncertainties, mainly animal products I would say, but not so much for the others. We are directly taken from FAO data, of course there is also uncertainty in the FAO data. The commodity balances might also be sometimes... We will have to do carefully some uncertainty analysis, and we wanted to give the users some more information about uncertainties later.

Actually, the construction of FABIO was done in several projects. We joined the forces of several fundings to be able to construct FABIO. At the moment, we have the funding to continue, and I'm confident that we can keep maintaining FABIO and updating it, especially because with the R codes it's rather simple. If there are new data published on FAOstat, we just run the codes again and we should have the additional year. For example, if the data for 2014 is published, a few days later we could have the FABIO dataset for 2014 available.

Yes, I can.

Thanks for the opportunity. Bye.