Figure 4
The diagram is divided into four vertical panels in sequence from left to right. Each panel includes two horizontally aligned groups of ellipses labeled “Lower echelon” at the top and “Upper echelon”, at the bottom. The first panel is labeled “Train subproblem models separately” and contains ovals in the echelons connected by arrows. Both echelons are connected by a dotted line. The text below reads “Inputs: Demand, I P I, C I F in each subproblem. Output: Trained model”. The second panel is labeled “Flow calculation of lower echelon” and displays arrows within the lower group and a dotted red inter-echelon arrow, with “Inputs: Trained models, Demand, I P I, C I F. Output: Flows”. The ovals and arrows in the “upper echelon” are shown transparently. The third panel, “Flow balance in upper echelon country”, highlights only one oval and two arrows from the “Lower echelon” group, and the dotted red arrow from lower to upper echelon. The text below reads “Inputs: Flows from lower echelon to that country, Conversion factor. Output: demand in that node”. The fourth panel, “Flow calculation in upper echelon country”, contains directional arrows only within the lower group and a dotted red arrow from lower group, labeled with “Inputs: Trained models, I P I, C I F, Demand from previous step. Output: demand in that node”. The Lowe echelon” section is shown transparently. All the panels are connected by right arrows. A flow legend at the bottom defines the symbols and arrows, where red dotted line shows “Connections between echelon”, line shows “Boundary of subproblem”, ovals show “Country”, and arrows show “Flow”.Steps of SC reconstitution