---
license: mit
---
# FLASHCast Dataset

This repository contains the training, validation, and testing datasets for **FLASHCast**, an auto-regressive unit streamflow and flash flood prediction deep learning architecture.

## Dataset Structure
*   **`data/unitq/`**: Sparse numpy arrays representing sequential geographical streamflow over time.
*   **`data/precip/`**: Matrix arrays recording precipitation and atmospheric forcings (MRMS/WoFS).
*   **`data/static/`**: Static geographical mappings such as flow accumulation and ground elevation properties.

## Origin Context
Processed via UNetDiffusion (`miles-credit` backbone) utilizing high-resolution continuous hydrological models covering the Continental US (CONUS).

_Total Size: ~151 GB_