The world-wide waste web

The study addresses how hazardous waste (HW) is traded globally and organized as a complex network—the world-wide waste web (W4). Annually, 7–10 billion tonnes of waste are produced, including 300–500 million tonnes of HW, of which about 10% is traded. Prior claims suggest dominant flows from developed to developing countries, though South–South and South–North exchanges also occur. The Basel Convention (BaC) monitors legal HW trade but faces reporting limitations and excludes illegal trafficking. The research question is to characterize the topology and dynamics of the W4, quantify risks of waste congestion in countries, and assess potential environmental impacts considering countries’ environmental performance. The study focuses on 108 HW categories (2001–2019; 2010 missing), aggregated into seven types, to analyze global structure, simulate congestion dynamics, and identify countries at high risk where improper handling may cause environmental and health impacts detectable via chemical fingerprints.

Background work highlights rapid growth in HW trade, complex economic and geopolitical drivers, and debates about directional asymmetries (developed-to-developing vs. more complex South–South and South–North flows). The Basel Convention provides the main legal framework and data source but suffers from accuracy issues and lacks coverage of illegal trade. Prior studies have examined waste economics, health impacts of HW exposure, and cases of illegal trafficking. The study builds on complex networks literature to analyze weighted, directed trade networks and leverages environmental performance indices to contextualize risk.

Data: Basel Convention Online Reporting Database (ERS) country-reported imports/exports for 108 waste categories, years 2001–2019 (2010 missing). Country codes standardized (ISO 3166-1 alpha-2), reports curated to correct naming/coding errors; transit countries and self-exports excluded. Waste categories grouped into seven types based on BaC Annexes I, II, and VIII: Types I (Y1–Y18), II (Y19–Y45), III (Y46–Y47), and IV–VII (A-list groupings). Network construction: For each type and year, build weighted, directed networks with nodes as countries/territories. If a reported transaction q includes multiple categories across types, split q into type-specific weights via proportional allocation using q_k^{ij}/Σ_k q_k^{ij}. For each pair (i,j), use the maximum of reported export E(i,j) and import I(j,i) to set A_ij = max[E(i,j), I(j,i)], allowing asymmetry. Normalize adjacency matrices by total flows as described. Structural metrics computed include in-/out-strength distributions (testing 17 candidate distributions), edge density, reciprocity (after binarization), average number of weighted directed triangles (trace of A^3), clustering coefficient for directed networks, average path length, and (sub)graph centrality measures including betweenness and subgraph centrality. Congestion dynamics: A logistic growth-based model with dynamic memory describes normalized waste accumulation w_i(t) with carrying capacity normalized to 1 per country-year (based on total traded waste). To capture temporal memory, replace first derivative with Caputo fractional derivative D^α with 0<α≤1. Two processes modeled: congestion at arrival (imports) and congestion at departure (exports), after a change of variables s(t)=log(1−w(t)). Due to parameter uncertainties, use a linearized worst-case upper-bound system involving the weighted adjacency (A or A^T), diagonal Ω=diag(1−w_0), and Mittag-Leffler matrix functions for the solution. Initial condition w_0=1−c/n with c=0.005. Simulation parameters: α=0.75, β=0.01, c=0.005. Congestion time t_i defined as time to 50% congestion (s_i(t)=0.5). Risk indices: Risk of waste congestion R_i=1−1/max{t_i^A,t_i^D}, bounded in [0,1]. Environmental underperformance index U_i = 1 − EPI(i)/100 (EPI from Yale/Columbia). Potential Environmental Impact of Waste Congestion (PEIWC): plot R versus U per country and waste type; define a tolerance zone via upper/lower 50% prediction bounds around the linear regression of U on R; countries above this zone are at high risk of improper handling and disposal of wastes (HRIHDW). Yearly, produce aggregated networks and temporal trends by correlating R and U with time (Pearson r). Chemical fingerprints (CF): compile literature evidence linking waste-related activities to heavy metals (HM), volatile organic compounds (VOC, e.g., BTEX with T/B diagnostic), and persistent organic pollutants (POP: PCB, PCDD/Fs) in HRIHDW countries.

• Scale and composition: From 2001–2019, 1,470,096,618 metric tonnes of waste were reported traded. Types I–III account for 95.41% of traded volume: Type I 40.4%, Type II 28.9%, Type III 26.1%.
• Trade asymmetry and concentration: For I–III wastes, most trade occurred among developed nations: 90.67% (Type I; 8.29×10^7 tons), 70.19% (Type II; 4.58×10^7 tons), 99.07% (Type III; 5.86×10^7 tons). Directional imbalances: for Type I, developed exported 4,340,000 tons more to developing and 25,500 tons more to least-developed than they imported; for Type III, 52,000 and 15,300 tons more, respectively. For Type II, developed nations imported 9,870,000 tons more from developing nations than they exported. Import/export distributions are fat-tailed, indicating few countries dominate flows and connectivity.
• Congestion risk and environmental performance: PEIWC identified 57 countries above the tolerance zone (HRIHDW) across Types I–III, with a subset of 28 top-risk countries selected for detailed CF analysis. Examples include China, Bangladesh, India, Nigeria, Senegal, D.R. Congo, Lesotho, Mozambique, Morocco, Sierra Leone, Madagascar, Benin, Niger, Liberia, Ethiopia, Pakistan, Djibouti, Mexico, Afghanistan, Mauritania, Papua New Guinea, Marshall Islands, Eritrea, North Korea, Mali, Burkina Faso, and Barbados. Many have low EPI and rapid congestion times. 15 HRIHDW countries reported zero imports (S_in=0) across years but significant exports, suggesting potential underreporting/illegal imports.
• Temporal trends (2001–2019): Few HRIHDW countries improved both waste congestion risk and environmental performance (bottom-left of trend plot). Many worsened on both (top-right). Example: Lesotho worsened in both; Bangladesh improved EPI and slightly reduced congestion risk.
• Network evolution: Edge density decreased (Pearson r with time = −0.69), clustering increased (r=0.59), average path length decreased (r=−0.36), reciprocity roughly stable (r=−0.05). Average number of weighted directed triangles fell strongly (r=−0.83), as did weighted subgraph centrality (r=−0.78), suggesting a breakdown of balanced cyclic trade patterns and shift toward more net importer/exporter roles.
• Net importer/exporter shifts: Germany, France, U.S., and Ukraine trended toward net exporters (ΔS=S_in−S_out vs. time r=−0.78, −0.72, −0.66, −0.44). Netherlands, Belgium, Spain, and Canada trended toward net importers (r=0.88, 0.57, 0.42, ≈0.34). From 2011–2019, strong shifts to net exporters: Slovenia, U.K., New Zealand, Germany; to net importers: Netherlands, Poland, Sweden, R. Korea. Among HRIHDW: China shifted to net exporter (r=−0.70); Mexico (r=0.66), India (0.62), Uzbekistan (0.47) toward net importers.
• Network roles: Developed countries have highest betweenness centrality (e.g., U.K., France, Germany, Austria, Netherlands, Belgium). 82.8% of HRIHDW countries have zero betweenness, indicating endpoint roles (net importers or exporters) rather than intermediaries.
• Chemical fingerprints (linking waste to impacts): • Heavy metals (HM): Widespread contamination associated with dumpsites and informal recycling (e-waste and lead-acid batteries) in HRIHDW countries. Examples include elevated Pb, Cd, Ni, Cr, Zn, Cu, As in groundwater/sediments/biota in Burkina Faso, D.R. Congo (ambient air Cd, Pb, Ni), Lesotho (As, Pb in fish above WHO limits), China (elevated maternal/child Pb; elevated metals in residents/workers), India (dermal exposure factors: Cr 192.6×, Cu 78.1×, Pb 30.9×, Zn 37.3×), Nigeria (high HM in soils at dumps/recycling sites), Madagascar and Senegal (battery recycling linked to Pb exposure, including child fatalities in Senegal). • VOC: Elevated BTEX and high toluene/benzene (T/B) ratios indicative of waste-related emissions in HRIHDW locales: Guangdong e-waste dismantling town (T/B≈3.15), Guangzhou (9.36), Dakar urban/semirural (4.51/5.32). High T/B also reported in Bangladesh (6.85), Benin (7.75), Burkina Faso (2.32), Ethiopia (2.3, 4.25), India (3.58–8.97), Mexico (2.19–6.59). • POP (PCB, PCDD/Fs): Large PCB inventories in several HRIHDW countries (e.g., Mozambique 240,571 tonnes of oil suspected PCB; Ethiopia, China, Bangladesh, Lesotho, Liberia, Morocco, Nigeria, D.R. Congo, Sierra Leone with substantial PCB-containing equipment/oils). PCB contamination observed in regional air/soil/milk and aquatic systems; West African offshore air shows major PCB emissions; immigrants’ serum PCB levels correlate with origin countries’ second-hand e-waste imports. PCDD/Fs releases markedly higher in HRIHDW: D.R. Congo 300,412 g TEQ/a; China 10,232; India 8,658; Nigeria 5,340; Lesotho 1,708; Sierra Leone 1,242. Mean TEQ for HRIHDW ~2,162 vs. ~587 for other countries.

The analysis demonstrates that while most HW trade occurs among developed countries, there is a persistent directional imbalance toward developing and least-developed countries for many waste types. The fractional-order congestion dynamics capture how waste stress propagates over the network and through time, showing that countries with low environmental performance can reach congestion rapidly, elevating risks of improper handling. The PEIWC framework integrates congestion risk with environmental capacity, revealing 57 HRIHDW countries (28 top-risk) that are disproportionately vulnerable. Independent chemical fingerprint evidence (HM, VOC, POP) supports the link between waste trade and environmental/human health impacts in multiple HRIHDW countries, consistent with open dumping, informal recycling, and inadequate infrastructure. Network evolution toward fewer weighted cycles and more polarized importer/exporter roles suggests decreased buffering and increased vulnerability to shocks (e.g., policy bans or demand shifts). Developed countries occupy key intermediary roles (high betweenness), whereas most HRIHDW countries are endpoints, limiting their influence over trade pathways and potentially exacerbating risk exposure. Overall, the findings support targeted interventions to improve waste management capacity and governance in vulnerable countries and to reconsider trade patterns that externalize environmental burdens.

This work introduces a network-based, fractional-order congestion model to quantify how hazardous waste trade stresses countries and to estimate congestion times and risks. Coupled with an environmental underperformance index, the PEIWC framework identifies countries at high risk of improper handling and disposal, highlighting 57 HRIHDW and a top subset of 28 for detailed analysis. Structural analyses reveal a less densely connected but more clustered W4 with a pronounced decline in weighted cycles, and shifting roles of countries toward net importer/exporter positions. Independent chemical fingerprint evidence across HM, VOC, and POP substantiates likely environmental and health impacts in several HRIHDW countries. The model can be used to assess scenarios such as import bans (e.g., China’s 2017 policy), pandemic-driven waste surges, and global increases in waste volumes. Future research should refine parameterization by waste stream and country-specific capacities, integrate illegal trade and transit data, improve temporal resolution, and validate model outputs against observed environmental and health outcomes.

• Data limitations: Reliance on self-reported BaC data; illegal trade and transit routes are not captured; 2010 missing; discrepancies between reported exports and imports necessitated using max(E, I) per dyad; some countries reported zero imports despite significant exports.
• Model assumptions: Waste types aggregated into seven groups; carrying capacity normalized to total annual traded waste per country; fractional derivative order fixed at α=0.75 and β=0.01 across contexts; linearized worst-case solution provides upper bounds rather than exact trajectories; risk defined at 50% congestion threshold.
• PEIWC framing: EPI as proxy for environmental capacity may not capture within-country heterogeneity or temporal policy changes; tolerance zone definition (50% prediction bounds) affects HRIHDW classification; selection of top 28 countries emphasizes largest deviations.
• Causality: CF evidence links waste activities to pollutants, but confounding sources (e.g., mining, industrial emissions, fuel use) can contribute; assigning pollutant burdens solely to traded waste is not always possible.
• Generalizability: Results depend on 2001–2019 trade patterns; evolving policies (e.g., bans), technological changes, and economic shifts can alter network structure and risks.