Rapid inundation mapping using NWM, satellite observations, and a convolutional neural network - Demonstrated on California Atmospheric Rivers 2023

This dataset supports the analysis presented in our paper (https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2024GL109424) on flood prediction using the Novel Water Model-Convolutional Neural Network (NWM-CNN). Aimed at enhancing flood forecasting and resilience, this collection encompasses a comprehensive compilation of flood events across California during the Atmospheric River events during the water year 2023. Included are NWM-CNN predictions, Sentinel-1 flood observations, and historical flood damage reports for Sacramento County, California, during the significant 2023 atmospheric river event and over the past decades.

The dataset is structured to facilitate a detailed examination of the NWM-CNN model's performance in predicting surface water area with high temporal resolution and accuracy. By integrating satellite imagery with hydrodynamic modeling, the NWM-CNN model represents a significant advancement in flood modeling, offering an effective tool for damage assessment, flood forecasting, and supporting parametric insurance solutions.

Key components of the dataset include:
* NWM-CNN Predictions: Model predictions at a 250m grid cell resolution.
* Sentinel-1 Flood Observations: Satellite-derived flood extents used for comparison against model predictions.
* Flood Damage Reports: Historical records of flood damage within Sacramento County, providing a ground-truth comparison for model efficacy.

Code to analyze this data is available here:
Jonathan Frame. (2024). jmframe/NWM_CNN_california_AR_2023: GRL Paper Proof 1 (1.0.0). Zenodo. https://zenodo.org/records/13153247. DOI: 10.5281/zenodo.13153247

By making this data publicly available, we aim to contribute to the collective efforts in reducing the impacts of climate disasters through improved access to scientific information and resources.

本数据集支持我们发表于论文（https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2024GL109424）中关于采用新型水文模型-卷积神经网络（Novel Water Model-Convolutional Neural Network，NWM-CNN）开展洪水预测的相关分析。为提升洪水预报效能与灾害韧性，本数据集收录了2023水文年加州大气河流事件期间的加州全域系列洪水事件完整汇编数据，包含NWM-CNN预测结果、哨兵-1（Sentinel-1）洪水观测数据，以及2023年重大大气河流事件期间及过去数十年间加州萨克拉门托县的历史洪水损失报告。

本数据集的结构设计便于细致评估NWM-CNN模型在高时间分辨率、高精度地表水面积预测任务中的表现。该模型将卫星影像与水动力模拟相结合，是洪水模拟领域的重要进展，可为灾害损失评估、洪水预报及参数化保险方案提供有效支撑工具。

数据集核心组成如下：
* NWM-CNN预测结果：分辨率为250米网格单元的模型预测数据
* 哨兵-1（Sentinel-1）洪水观测数据：用于与模型预测结果比对的卫星反演洪水范围数据
* 洪水损失报告：萨克拉门托县境内的历史洪水损失记录，可为模型效能验证提供地面真值对比依据

本数据集的分析代码可参见：
Jonathan Frame. (2024). jmframe/NWM_CNN_california_AR_2023: GRL Paper Proof 1 (1.0.0). Zenodo. https://zenodo.org/records/13153247. DOI: 10.5281/zenodo.13153247

本数据集公开发布的初衷，是通过提升科学信息与资源的可及性，助力全球减少气候灾害影响的集体研究与实践工作。