Skip to content

Latest commit

 

History

History
52 lines (33 loc) · 2.39 KB

README.md

File metadata and controls

52 lines (33 loc) · 2.39 KB
Raw

A multiscale deep learning model for soil moisture integrating satellite and in-situ data

Installation

This repository was written and tested for Python 3.7, CUDA 10.1, and Pytorch 1.7.1. It can be run in linux or the user needs to create the environment manually. As the data is large, you need to download it from here.

unzip hydroDL_zenodo.zip
cd hydroDL_zenodo
conda env create -f hydroDL-rel.yml python==3.7 
conda activate hydroDL-rel 
pip install torch==1.7.1+cu101 torchvision==0.8.2+cu101 torchaudio==0.7.2 -f https://download.pytorch.org/whl/torch_stable.html
(linux): wget https://data.pyg.org/whl/torch-1.7.0%2Bcu101/torch_scatter-2.0.7-cp37-cp37m-linux_x86_64.whl
(linux): pip install ./torch_scatter-2.0.7-cp37-cp37m-linux_x86_64.whl
(windows): wget https://data.pyg.org/whl/torch-1.7.0%2Bcu101/torch_scatter-2.0.7-cp37-cp37m-win_amd64.whl
(windows): pip install ./torch_scatter-2.0.7-cp37-cp37m-win_amd64.whl

Dataset Setup

The tutorial includes demo data located in ./example/multiscale/, which contains three folders fine_resolution, coarse_resolution, and insitu.

How to match coarse-resolution with fine-resolution data :

coarse-resolution and fine-resolution data are matched by index_bsf_fsg.json, which is as follows:

[[0,1,2...,15], [16, 17,18,…,31],...]

Concept

The position of each sub-list corresponds to the index of the coarse cell, and the content of the sub-list corresponds to the index of the fine resolution data contained in the coarse cell. A part of the data can be selected for training if the data is too large. The method is to pick the index of the coarse grid (e.g.[0, 2, 3, ...]) and save it as index_bsg_nv.json. Then, save the new sub-list of fine resolution grid index as index_bsg_fsg.json. It also works for in-situ data.

How to train the model:

multiscale.py can train demo data directly.

How to train custom data:

Prepare your own data according to the format in . /example/multiscale/, and modify the following configuration:

  • The data path: rootDB, rootFine, and rootSite.

  • The period of data: dBRange.

  • The training period: tRange.

  • The variable name: varT, varC, and target.

  • The validation period: testRange.

  • The epoch number of validated model: testEpoch.