This resource contains model code and data used in "Imagery classification of stream stage to support ephemeral stream monitoring". The "Model" folder contains the Logistic Regression model used in the paper to classify stream state as "high water" levels, "low water" levels, "no water", or "obstructed" for images ("Photos.zip") of Perry Creek, California from August 2017-November 2023. Within the "Model" folder, there is also a model that was optimized for running on personal computers; this will generate slightly different results compared to the model used in the paper, but uses less memory. The "Photos" .zip contains the "Labeled" photos that were used to train the model and the "Unlabeled" photos that were classified by the model. The "Data" folder contains surface meteorology, soil moisture, stream stage, and streamflow data used in the paper to understand the hydrologic context of the image classifications and develop a deeper understanding of the hydrology of the Perry Creek watershed. Instructions for how to use the method presented here to classify photos from the USGS Flow Photo Explorer are provided in the "USGS_Photo_Instructions" folder. Please refer to the manuscript and the metadata below for detailed descriptions of the data and methods.

Here are instructions for how to run the model (instructions for running the PC-optimized version are in guide.txt of "PC_optimized_model.zip"):
1. Install Python 3.10 here: https://www.python.org/downloads/release/python-3100/ (once installed, add to PATH; Note: later Python version may work, but has not been tested)
2. Open your preferred interpreter (Ex: Visual Studio Code)
3. Check that you have Python 3.10 with "python --version"
4. Download the "Model_in_Paper" folder and "Photos.zip" into a folder on your computer. Unzip "Photos.zip".
5. Check your paths to ensure that they look like: "working_dir/Labeled_Photos/PEC/*" and "working_dir/Unlabeled_Photos/PEC/*"
6. In your working directory, create a virtual environment (you can also use a package manager like Conda) by typing this in the Terminal: "python -m venv venv"
7. Activate the virtual environment (Windows: "venv\Scripts\activate" or Mac: "source venv/bin/activate")
8. Install the packages from the requirements.txt file: "pip install -r requirements.txt"
9. Update the paths in main.py to point to where the photos are saved ("photo_path") and where you would like figures saved ("working_dir").
10. Use Python 3.10 to run main.py to classify the photos (navigate to the "Model_in_Paper" folder in Terminal and run "python main.py"). Main.py uses many of the functions in the "Model_in_Paper/utils" folder. The code will generate figures from the manuscript and return image classifications for the unlabeled photos that should resemble those in the "Model_in_Paper/image_classifications_final_model.csv".
11. The user can adjust the model configuration in main.py and/or use the model to classify their own field camera stream photos (using a folder of identically-sized, labeled photos sorted by image label and a folder of identically-sized, unlabeled photos).