Case Tallinn | Spatial and Logistics Optimisation for Household Biowaste Management System

Author: Srinivasa Raghavendra Bhuvan Gummidi, University of Southern Denmark

By implementing a spatial optimisation model instead of Tallinn’s traditional district-based biowaste allocation, we successfully addressed the underutilisation of the high-value biogas plant. This data-driven approach, which maximized monthly profit based on facility revenues and reduced transport costs, led to a full utilisation of the biogas plant’s capacity. 

In TREASoURcE Task 6.4 (see report), we worked on a very applied question from Tallinn, Estonia. We asked how we can allocate household biowaste between two designated facilities, a biogas plant and a composting site, in a way that improves economic performance and still remains realistic from a logistics point of view. We used geocoded household locations, road network distances, facility capacity limits, and an optimisation model where monthly profit is defined as facility revenue minus transport costs.

The Situation in Tallinn that Motivated our Work

Tallinn follows district-based rules for allocating household biowaste. These rules were reported as effective in late 2024. In practice, the household’s district decides the destination facility. The rules are not designed based on road distance to the facilities, and they do not directly aim to maximise profitability. Because of this, we built a baseline that follows these district rules and we compared it with a spatial optimisation approach.

The report lists the district allocation clearly (see Figure 1). Haabersti (District 1), Mustamäe (Districts 3 and 4), and Pirita (District 13) are allocated to the biogas plant. Nõmme (2), Kristiine (5), Põhja-Tallinn (6 and 7), Kesklinn (8), Kesklinn Old Town (9), and Lasnamäe (10, 11, 12) are allocated to the composting site.

Data Use

For this case, we combined district-level biowaste information with route data from the biogas company and the composting company. We also used Tallinn district boundaries, a 100 m population grid, and OpenStreetMap road network data for calculating network distances. The analysis also includes facility information such as locations, capacity constraints, and revenue assumptions reported in the deliverable.

For the baseline supply estimation, the report explains that monthly household biowaste generation was approximated using April 2024 route data from the biogas company and October 2024 route data from the composting company. This process produced an estimated dataset covering around 24,624 household locations.

We also stayed consistent with the operational limits stated by Tallinn Municipality. The report notes a maximum of 250 addresses per collection route and 26 tons per trip.

District-Based Baseline

In the baseline, each household point was linked to a district polygon and then assigned to the biogas facility or the composting facility using the district rules described above (see Figure 2).

For the approximate 2024 baseline conditions, the report provides the following results for district allocation. The biogas facility received 302.38 tons per month. This corresponds to 30.2% utilisation of the assumed 1,000 ton per month household capacity. The composting site received 1,636.78 tons per month. This corresponds to 98.2% utilisation of the assumed 1,667 ton per month capacity. Monthly profit was €8,670.42 and monthly transport costs were €2,754.69. The average waste-weighted one-way distance was about 20.4 km.

These numbers show that composting is close to full utilisation, while the biogas facility is far below its capacity in the baseline. In the same report, the revenue per ton used in the model is higher for biogas than for composting. So, in the model logic, the low utilisation of the biogas facility matters for the overall system value.

Different Approach for Optimisation

We designed a workflow (see Figure 3) that is meant to remain practical for city planning while also handling a large number of households. Households were grouped using K-Means clustering. Then clusters were split or merged so that the operational limits, 250 addresses and 26 tons, are respected. For each cluster, a representative point was selected. The report states that this representative point is an actual household point that lies closest to the cluster centroid.

After clustering, a Mixed-Integer Linear Programming model was used to decide the allocation. The report states that the model is implemented with PuLP and solved with the COIN-OR CBC solver. The model assigns each cluster either to the biogas facility or to the composting facility. The objective is to maximise monthly profit using facility revenues and transport costs based on network distance.

For distances, Task 6.4 uses a road network approach. The report states the use of OSMnx for creating the street network and NetworkX for shortest path calculations.

Changes after Optimisation

When the optimisation model was applied to the same approximate 2024 waste supply, the allocation changed strongly. The optimised result assigned 94 clusters to the biogas facility. This corresponds to 1,000.00 tons per month, which fully uses the assumed 1,000 ton per month capacity. The remaining 73 clusters were assigned to the composting site. This corresponds to 939.15 tons per month, which is about 56.4% utilisation of the assumed 1,667 ton per month compost capacity.

In terms of economics and logistics, the report states that the maximum monthly profit increased to €16,050.39 compared to €8,670.42 in the baseline. Monthly transport costs decreased to €2,406.76 compared to €2,754.69 in the baseline. The report also notes that this is about a 12.6% reduction in transport costs. For the optimised solution, the report reports an average one-way distance of 14.33 km for clusters assigned to the biogas facility (see Figure 4).

Future Scenarios for 2035 and 2050

Task 6.4 also explores future conditions for Tallinn for the years 2035 and 2050. The report considers three scenario families for future household biowaste amounts. These are BAU, market-driven, and city-oriented scenarios. The scenario values are linked to population forecasts, and the report states that per-capita waste generation is kept constant in these projections.

In these future tests, the report assumes that facility capacity increases over time. It states a +25% increase by 2035 and a +50% increase by 2050, relative to the baseline capacity values.

Across all the future scenarios reported, the results remain economically positive. The report states monthly profits in the range of €19,200 to €22,900 (see Figure 5). It also highlights that the biogas plant capacity becomes a consistent limiting factor, and composting mainly receives the remaining amount. The average one-way distances are reported as stable at around 15 km for biogas and around 23 km for compost.

Benefits for Cities and Operators

From an environmental data science perspective, the Tallinn case is useful because it shows how spatial data can support planning decisions. The baseline follows district rules, and it is easy to apply, but it can lead to uneven facility utilisation. In the optimisation results reported in Task 6.4, the model fills the biogas capacity and improves the reported monthly profit while reducing the reported transport costs. This is based on the assumptions and parameters that are clearly stated in the report.

For replication, the main requirement is not a specific software tool. The requirement is to have the right input information. The report shows the need for geocoded supply points or a reliable way to represent them, road network distances, facility capacities, and clear economic parameters used in the objective function.