thesujitroy
commited on
Commit
•
080fa0c
1
Parent(s):
20f8a12
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,50 @@
|
|
1 |
-
---
|
2 |
-
license: apache-2.0
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
tags:
|
6 |
+
- Pytorch
|
7 |
+
- gravity wave
|
8 |
+
- Weather & Climate
|
9 |
+
- Foundation model
|
10 |
+
datasets:
|
11 |
+
- Prithvi-WxC/Gravity_wave_Parameterization
|
12 |
+
base_model:
|
13 |
+
- Prithvi-WxC/prithvi.wxc.2300m.v1
|
14 |
+
---
|
15 |
+
This repository contains pretrained model for Gravity Wave Flux Parametrization downstream task.
|
16 |
+
|
17 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/6488f1d3e22a0081a561ec8f/lOFP_1dAVKCw90uLpj2vu.png" alt="Gravity Wave" width="1024"/>
|
18 |
+
|
19 |
+
|
20 |
+
|
21 |
+
|
22 |
+
### Model
|
23 |
+
The pretrained [Prithvi WxC](https://huggingface.co/Prithvi-WxC/prithvi.wxc.2300m.v1) parameter model is finetuned to predict momentum fluxes from
|
24 |
+
the [Gravity Wave Parameterization dataset](https://huggingface.co/datasets/Prithvi-WxC/Gravity_wave_Parameterization).
|
25 |
+
|
26 |
+
<b>Input:</b> 491 (3 + 4x122) channels.
|
27 |
+
1. latitude (1)
|
28 |
+
2. longitude (1)
|
29 |
+
3. surface elevation (1)
|
30 |
+
4. zonal winds \\(u\\) (122)
|
31 |
+
5. meridional winds \\(v\\) (122) 6.
|
32 |
+
6. temperature \\(T\\) (122)
|
33 |
+
7. pressure \\(P\\) (122)
|
34 |
+
|
35 |
+
<b>Output:</b> 366 (3x122) channels.
|
36 |
+
1. potential temperature \\(\theta\\) (122)
|
37 |
+
2. zonal flux of vertical momentum \\(u'\omega'\\) (122)
|
38 |
+
3. meridional flux of vertical momentum \\(v'\omega'\\) (122)
|
39 |
+
|
40 |
+
### Code
|
41 |
+
Code for fine-tuning is available through [Github](https://github.com/NASA-IMPACT/gravity-wave-finetuning).
|
42 |
+
|
43 |
+
### Results
|
44 |
+
For the Andes (mountain waves) and the Southern Ocean (non-mountain waves),
|
45 |
+
the fine-tuned model achieves correlation coefficients of 0.99 and 0.97, respectively, when compared to the observed fluxes.
|
46 |
+
|
47 |
+
### Inference and demo
|
48 |
+
The github repo includes an inference script that allows to run
|
49 |
+
the [gravity_wave_model](https://huggingface.co/Prithvi-WxC/Gravity_wave_Parameterization/blob/main/magnet-flux-uvtp122-epoch-99-loss-0.1022.pt) model
|
50 |
+
for inference on [sample dataset](https://huggingface.co/datasets/Prithvi-WxC/Gravity_wave_Parameterization/blob/main/wxc_input_u_v_t_p_output_theta_uw_vw_era5_training_data_hourly_2015_constant_mu_sigma_scaling05.nc).
|