## What is STILT

A basic introduction to the Stochastic Time-Inverted Lagrangian Transport (STILT) model, an open-source lagrangian particle dispersion model

### Particles

STILT uses meteorological data to follow an ensemble of theoretical particles as they are transported backward in time through the atmosphere

### Footprints

These particles are used to calculate a **footprint** defining the upstream area that influences the air arriving at a given location

## How STILT works

### 1 Define a receptor

A receptor is defined as a location at a specific time that we are interested in. Often receptors are the location/time that a measurement is made, allowing us to investigate what happened to the air before it arrived at the measurement location.

### 2 Release and follow particles

A cloud of theoretical particles are released from the receptor and followed as they travel backward in time. The average trajectory of the particle cloud is calculated using meteorological wind fields and random velocities are introduced for each particle using a Markov chain process to approximate turbulent motions.

### 3 Calculate influence footprint

The surface influence footprint is calculated using the positions and height above ground of the particles at each time step. Footprints define the area upstream from the receptor as well as the strength of the influence from any given location within the model domain. These footprints can be easily convolved with surface flux inventories to model the contribution of near-field surface fluxes on the receptor.

### 4 Evaluate emissions

Using STILT to relate measurements to emissions estimates is a powerful method for mapping pollution, fine-tuning emissions inventories, and tracking changes in emissions over time. Get creative - this is where the science happens.

## STILT Improvements

This distribution contains a completely redesigned R codebase and proposes a centralized, collaborative platform for documentation and future model development.

### Gaussian Kernel Footprints

Influence of particles is spatially distributed using gaussian kernels. This method agrees more closely with idealized brute-force cases over prior dynamic grid coarsening techniques.

### Parallel Simulations

High level methods for single and multi-node parallelism take advantage of computing resources, enabling total simulation time to decrease linearly with the number of parallel threads.

### Effective Dilution Depth

Nested gaussian plume model rescales the effective dilution depth for fluxes in the hyper near field. This increases the influence of fluxes originating close to the receptor.

### Modernized Framework

Simple front end controls for transport and dispersion modeling, model parallelization, and footprint calculations provide a systematic, well-documented workflow.