Datasets:
ArXiv:
License:
File size: 10,486 Bytes
85b6170 3c0d1b0 85b6170 4baa9de 85b6170 4baa9de 85b6170 3c0d1b0 85b6170 3c0d1b0 85b6170 5e5f5ad 85b6170 977d955 3afdcd9 977d955 85b6170 9f5d263 85b6170 3afdcd9 85b6170 3afdcd9 85b6170 3c0d1b0 3afdcd9 9f5d263 3afdcd9 9f5d263 85b6170 9f5d263 3afdcd9 85b6170 9f5d263 85b6170 3c0d1b0 85b6170 3afdcd9 85b6170 9f5d263 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 | ---
license: mit
---
# NASA Power Weather Data over North, Central, and South America from 1984 to 2022
<!-- Provide a quick summary of the dataset. -->
This dataset contains daily solar and meteorological data downloaded from the [NASA Power API](https://power.larc.nasa.gov/)
## Overview
The dataset includes solar and meteorological variables collected from January 1st, 1984, to December 31st, 2022.
We downloaded 28 variables directly and estimated an additional 3 from the collected data. The data spans a 5 x 8 grid covering
the United States, Central America, and South America. Each grid rectangle contains 160 data points spaced 0.5 degrees apart in
latitude and longitude.
## Citation
<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
**BibTeX:**
```
@misc{hasan2025vita,
title={VITA: Variational Pretraining of Transformers for Climate-Robust Crop Yield Forecasting},
author={Adib Hasan and Mardavij Roozbehani and Munther Dahleh},
year={2025},
eprint={2508.03589},
archivePrefix={arXiv},
primaryClass={cs.LG},
url={https://arxiv.org/abs/2508.03589},
}
```
### Dataset Description
Here are the descriptions of the 31 weather variables with their units:
| Parameter Name | Symbol | Unit |
|--------------------------------------------------------------|------------------|--------------------|
| Temperature at 2 Meters | T2M | C |
| Temperature at 2 Meters Maximum | T2M_MAX | C |
| Temperature at 2 Meters Minimum | T2M_MIN | C |
| Wind Direction at 2 Meters | WD2M | Degrees |
| Wind Speed at 2 Meters | WS2M | m/s |
| Surface Pressure | PS | kPa |
| Specific Humidity at 2 Meters | QV2M | g/Kg |
| Precipitation Corrected | PRECTOTCORR | mm/day |
| All Sky Surface Shortwave Downward Irradiance | ALLSKY_SFC_SW_DWN| MJ/m^2/day |
| Evapotranspiration Energy Flux | EVPTRNS | MJ/m^2/day |
| Profile Soil Moisture (0 to 1) | GWETPROF | 0 to 1 |
| Snow Depth | SNODP | cm |
| Dew/Frost Point at 2 Meters | T2MDEW | C |
| Cloud Amount | CLOUD_AMT | 0 to 1 |
| Evaporation Land | EVLAND | kg/m^2/s * 10^6 |
| Wet Bulb Temperature at 2 Meters | T2MWET | C |
| Land Snowcover Fraction | FRSNO | 0 to 1 |
| All Sky Surface Longwave Downward Irradiance | ALLSKY_SFC_LW_DWN| MJ/m^2/day |
| All Sky Surface PAR Total | ALLSKY_SFC_PAR_TOT| MJ/m^2/day |
| All Sky Surface Albedo | ALLSKY_SRF_ALB | 0 to 1 |
| Precipitable Water | PW | cm |
| Surface Roughness | Z0M | m |
| Surface Air Density | RHOA | kg/m^3 |
| Relative Humidity at 2 Meters | RH2M | 0 to 1 |
| Cooling Degree Days Above 18.3 C | CDD18_3 | days |
| Heating Degree Days Below 18.3 C | HDD18_3 | days |
| Total Column Ozone | TO3 | Dobson units |
| Aerosol Optical Depth 55 | AOD_55 | 0 to 1 |
| Reference evapotranspiration | ET0 | mm/day |
| Reference evapotranspiration | ET0 | mm/day |
| Vapor Pressure | VAP | kPa |
| Vapor Pressure Deficit | VAD | kPa |
### Grid coordinates for the regions
the location indices in the dataset refer to the order of these coordinates. For instance `usa_0` refers to
the first rectangle of the USA in the list below. For the pytorch data, location indices 0-34 refer to the data from the USA
grid, 35-110 refer to the data from the South America grid and the rest refer to the data from the Central America
grid.
#### USA
```
((29, -109), (24, -101)),
((29, -101), (24, -93)),
((29, -93), (24, -85)),
((29, -85), (24, -77)),
((34, -125), (29, -117)),
((34, -117), (29, -109)),
((34, -109), (29, -101)),
((34, -101), (29, -93)),
((34, -93), (29, -85)),
((34, -85), (29, -77)),
((34, -77), (29, -69)),
((39, -125), (34, -117)),
((39, -117), (34, -109)),
((39, -109), (34, -101)),
((39, -101), (34, -93)),
((39, -93), (34, -85)),
((39, -85), (34, -77)),
((39, -77), (34, -69)),
((44, -133), (39, -125)),
((44, -125), (39, -117)),
((44, -117), (39, -109)),
((44, -109), (39, -101)),
((44, -101), (39, -93)),
((44, -93), (39, -85)),
((44, -85), (39, -77)),
((44, -77), (39, -69)),
((49, -133), (44, -125)),
((49, -125), (44, -117)),
((49, -117), (44, -109)),
((49, -109), (44, -101)),
((49, -101), (44, -93)),
((49, -93), (44, -85)),
((49, -85), (44, -77)),
((49, -77), (44, -69)),
```
#### Central America
```
((29, -117), (24, -109)),
((24, -117), (19, -109)),
((24, -109), (19, -101)),
((24, -101), (19, -93)),
((24, -93), (19, -85)),
((24, -85), (19, -77)),
((19, -109), (14, -101)),
((19, -101), (14, -93)),
((19, -93), (14, -85)),
((19, -85), (14, -77)),
```
#### South America
```
((-51, -77), (-56, -69)),
((-51, -69), (-56, -61)),
((-46, -85), (-51, -77)),
((-46, -77), (-51, -69)),
((-46, -69), (-51, -61)),
((-41, -85), (-46, -77)),
((-41, -77), (-46, -69)),
((-41, -69), (-46, -61)),
((-41, -61), (-46, -53)),
((-36, -85), (-41, -77)),
((-36, -77), (-41, -69)),
((-36, -69), (-41, -61)),
((-36, -61), (-41, -53)),
((-36, -53), (-41, -45)),
((-31, -85), (-36, -77)),
((-31, -77), (-36, -69)),
((-31, -69), (-36, -61)),
((-31, -61), (-36, -53)),
((-31, -53), (-36, -45)),
((-26, -85), (-31, -77)),
((-26, -77), (-31, -69)),
((-26, -69), (-31, -61)),
((-26, -61), (-31, -53)),
((-26, -53), (-31, -45)),
((-26, -45), (-31, -37)),
((-21, -85), (-26, -77)),
((-21, -77), (-26, -69)),
((-21, -69), (-26, -61)),
((-21, -61), (-26, -53)),
((-21, -53), (-26, -45)),
((-21, -45), (-26, -37)),
((-21, -37), (-26, -29)),
((-16, -85), (-21, -77)),
((-16, -77), (-21, -69)),
((-16, -69), (-21, -61)),
((-16, -61), (-21, -53)),
((-16, -53), (-21, -45)),
((-16, -45), (-21, -37)),
((-16, -37), (-21, -29)),
((-11, -85), (-16, -77)),
((-11, -77), (-16, -69)),
((-11, -69), (-16, -61)),
((-11, -61), (-16, -53)),
((-11, -53), (-16, -45)),
((-11, -45), (-16, -37)),
((-11, -37), (-16, -29)),
((-6, -85), (-11, -77)),
((-6, -77), (-11, -69)),
((-6, -69), (-11, -61)),
((-6, -61), (-11, -53)),
((-6, -53), (-11, -45)),
((-6, -45), (-11, -37)),
((-6, -37), (-11, -29)),
((-1, -85), (-6, -77)),
((-1, -77), (-6, -69)),
((-1, -69), (-6, -61)),
((-1, -61), (-6, -53)),
((-1, -53), (-6, -45)),
((-1, -45), (-6, -37)),
((-1, -37), (-6, -29)),
((4, -85), (-1, -77)),
((4, -77), (-1, -69)),
((4, -69), (-1, -61)),
((4, -61), (-1, -53)),
((4, -53), (-1, -45)),
((4, -45), (-1, -37)),
((9, -85), (4, -77)),
((9, -77), (4, -69)),
((9, -69), (4, -61)),
((9, -61), (4, -53)),
((9, -53), (4, -45)),
((14, -85), (9, -77)),
((14, -77), (9, -69)),
((14, -69), (9, -61)),
((14, -61), (9, -53)),
```
## Dataset Structure
<!-- This section provides a description of the dataset fields, and additional information about the dataset structure, such as criteria used to create the splits, relationships between data points, etc. -->
**raw:** unprocessed data dump from NASA Power API in the JSON format.
**csvs:** Processed data in the CSV format.
**pytorch:** Pytorch TensorDataset objects ready to be used in training. All of the daily, weekly, and monthly data have been reshaped
so that the **sequence length is 365**. Each sample is a tuple of the following data:
* weather measurements (shape `sequence_length x 31`)
* coordinates (shape `1 x 2`)
* index (`1 x 2`). the first number is the temporal index of the current row since January 1, 1984. The second number is the temporal granularity,
or the spacing between indices, which is 1 for daily data, 7 for weekly data, and 30 for monthly data. Note: this means the daily data
contains 1 year of data in each row, weekly data contains 7.02 years of data in each row (`7.02 * 52 = ~365`), and monthly data contains 12.2 years of
data (`12.2 * 30 = ~365`).
## Dataset Creation
### Source Data
<!-- This section describes the source data (e.g., news text and headlines, social media posts, translated sentences, ...). -->
NASA Power API daily weather measurements. The data comes from multiple sources, but mostly satellite data.
#### Data Processing
<!-- This section describes the data collection and processing process, such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->
The `raw` data is in the JSON format and unprocessed. The `csvs` and the `pytorch` data are processed in the following manner:
- Missing values were backfilled.
- Leap year extra day was omitted. So, each year of the daily dataset has 365 days. Similarly, each year of the weekly dataset has 52 weeks, and the monthly dataset has 12 columns.
- Data was pivoted. So each measurement has x columns where x is either 365, 52, or 12.
- `pytorch` data was standardized using the mean and standard deviation of the weather over the continental United States.
|