**A step‑by‑step indirect methodology for estimating traffic volume on rural local roads**

Below is a practical workflow that builds on the transport‑demand modelling concepts described in the literature and can be applied when direct counts are unavailable.

---

### 1. Delineate functional sub‑networks  
* **Goal:** Split the whole road system into self‑contained “functional networks” whose external borders are national or other main roads.  
* **How:** Use GIS to overlay the complete road catalogue, then clip out every cluster of local (non‑national) links that is bounded on all sides by a higher‑order route. Each resulting polygon becomes one functional network for which an independent demand model will be run.

### 2. Gather surrogate data for each sub‑network  
| Variable | Typical source |
|----------|----------------|
| Population & households per zone | Census or municipal registers |
| Employment / land‑use mix (residential, agricultural, commercial) | Land‑use maps, cadastral databases |
| Road geometry (lane width, shoulder, curvature) | Existing road inventory |
| Connectivity metrics (distance to nearest arterial, number of intersections) | GIS network analysis |

These variables will later serve as explanatory factors in the demand model.

### 3. Build a travel‑demand model for the study area  
* **Generation – Distribution – Mode split – Assignment** is the classic four‑stage structure used for aggregate modelling [7].  
* Because microscopic (per‑vehicle) data are rarely available on rural networks, start with an *origin–destination (O/D) matrix* at the zone level. If a detailed O/D set already exists (e.g., from VISUM or another planning model), import it directly – SUMO’s tools support such imports [6].  
* Where no O/D data exist, generate synthetic matrices using socio‑economic variables collected in step 2 (population → trip generation; employment & attractiveness → distribution).

### 4. Convert the O/D matrix into link‑level traffic volumes  
1. **Trip assignment** – Run a static or dynamic user‑equilibrium assignment on the functional network to distribute trips onto individual links.  
2. **Adjustment for rural characteristics** – Apply correction factors that reflect lower vehicle‑kilometres travelled per capita in sparsely populated areas (e.g., multiply by a “rural DVMT factor” derived from regional travel surveys).  

### 5. Calibrate & validate with any available counts  
* Even a handful of short manual or automated counts on representative segments can be used for calibration.  
* Use an *empirical Bayes* approach: combine the model‑based estimates (prior) with observed counts (likelihood) to obtain posterior AADT values that are statistically more reliable.  

### 6. Derive final traffic‑volume database  
For every local‑road segment within each functional network compute:

\[
\text{AADT}_i = \frac{\text{Assigned trips on } i}{365}
\]

Optionally attach a confidence interval derived from the calibration step.

---

## Why this works for rural areas  

* **Indirectness:** The method does not require continuous field counting; it relies on readily available demographic and land‑use data.  
* **Network segmentation:** By treating each functional network separately, the influence of major corridors is isolated, preventing spill‑over bias that often plagues aggregated models.  
* **Flexibility:** If later more counts become available (e.g., seasonal pneumatic tubes or temporary radar surveys), they can be incorporated through the same empirical Bayes update without re‑building the whole model.

---

## Supporting references from the literature  

* The four‑stage demand modelling framework is outlined in transport‑demand studies [7].  
* SUMO’s capability to import O/D matrices (useful for large‑scale rural applications) is described in its documentation [6].  
* Simulation of individual vehicle interactions on a single rural stretch using the RuTSim model demonstrates how calibrated demand can be fed into microscopic traffic simulators for validation purposes [1].

---

### Quick checklist for implementation  

| ✔︎ | Action |
|---|--------|
| 1 | Create GIS layers of national/main roads and clip local‑road clusters. |
| 2 | Compile zone‑level socio‑economic variables. |
| 3 | Generate or import O/D matrices; run a user‑equilibrium assignment per functional network. |
| 4 | Apply rural DVMT correction factors. |
| 5 | Calibrate with any spot counts (empirical Bayes). |
| 6 | Export AADT values + uncertainty for all local segments. |

Following this workflow will give you a defensible, reproducible estimate of traffic volumes on rural local roads without the need for extensive field counting campaigns.