Saved in:
| Main Author: | |
|---|---|
| Format: | Recurso digital |
| Language: | |
| Published: |
Zenodo
2025
|
| Online Access: | https://doi.org/10.5281/zenodo.16652554 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Table of Contents:
- <p>This file is the random forest trained model for prediction of next-hour FRP given a set of Global Forecast System (GFS) inputs and gridded RAVE Fire Radiative Power (FRP) data points for the same corresponding hour. Inputs are listed below:</p> <table style="border-collapse: collapse; width: 516pt; height: 289.297px;"><colgroup><col style="width: 129pt;" span="4"> </colgroup> <tbody> <tr style="height: 41.3281px;"> <td style="height: 41.3281px; width: 129pt;">lat'</td> <td style="width: 129pt; height: 41.3281px;">'lon'</td> <td style="width: 129pt; height: 41.3281px;">'u_wind_10m'</td> <td style="width: 129pt; height: 41.3281px;">'snow_depth_water'</td> </tr> <tr style="height: 41.3281px;"> <td style="height: 41.3281px; width: 129pt;">'CAPE'</td> <td style="width: 129pt; height: 41.3281px;">'dsflx'</td> <td style="width: 129pt; height: 41.3281px;">'v_wind_10m'</td> <td style="width: 129pt; height: 41.3281px;">'soilt_10_40cm'</td> </tr> <tr style="height: 41.3281px;"> <td style="height: 41.3281px; width: 129pt;">'acc_precip'</td> <td style="width: 129pt; height: 41.3281px;">'soilt_surf'</td> <td style="width: 129pt; height: 41.3281px;">'usflx'</td> <td style="width: 129pt; height: 41.3281px;">'soilt_40_100cm'</td> </tr> <tr style="height: 41.3281px;"> <td style="height: 41.3281px; width: 129pt;">'24_hr_roll_frp'</td> <td style="width: 129pt; height: 41.3281px;">'QC'</td> <td style="width: 129pt; height: 41.3281px;">'ulflx'</td> <td style="width: 129pt; height: 41.3281px;">'soilt_100_200cm'</td> </tr> <tr style="height: 41.3281px;"> <td style="height: 41.3281px; width: 129pt;">'hour'</td> <td style="width: 129pt; height: 41.3281px;">'temp_2m'</td> <td style="width: 129pt; height: 41.3281px;">'p_surf'</td> <td style="width: 129pt; height: 41.3281px;">'soilm_frac_10_40cm'</td> </tr> <tr style="height: 41.3281px;"> <td style="height: 41.3281px; width: 129pt;">'dlflx'</td> <td style="width: 129pt; height: 41.3281px;">'dewpt_2m'</td> <td style="width: 129pt; height: 41.3281px;">'precip_water'</td> <td style="width: 129pt; height: 41.3281px;">'soilm_frac_40_100cm'</td> </tr> <tr style="height: 41.3281px;"> <td style="height: 41.3281px; width: 129pt;">'soil_moist_frac'</td> <td style="width: 129pt; height: 41.3281px;">'sphum'</td> <td style="width: 129pt; height: 41.3281px;">'snow_depth'</td> <td style="width: 129pt; height: 41.3281px;">'soilm_frac_100_200cm'</td> </tr> </tbody> </table> <p>Notes on inputs: </p> <ul> <li>QC is a quality control flag from RAVE</li> <li>24_hr_roll_frp is the average FRP value over a 24-hour rolling time time window for the gridded GFS gridcell. This value is always greater than 0 as a model requirement.</li> <li>lat and lon are centroid lat and lon values from GFS grid</li> </ul> <p>This model was trained on non-zero rolling 24-hour mean RAVE FRP data gridded to GFS resolution from 2019. The blind validation performance metrics are created from fires unseen in the training and testing process, for any models including this next-hour one, for August through December 2021. The random forest was set up with Model performance is listed below:</p> <table style="border-collapse: collapse; width: 937pt;"><colgroup><col style="width: 120pt;"> <col style="width: 201pt;"> <col style="width: 79pt;" span="3"> <col style="width: 89pt;" span="2"> <col style="width: 70pt;"> <col style="width: 131pt;"> </colgroup> <tbody> <tr style="height: 46.0pt;"> <td style="height: 46.0pt; width: 120pt;"><strong>Model initialization</strong></td> <td style="width: 201pt;"><strong>Input Summary</strong></td> <td style="width: 79pt;"><strong>MAE</strong></td> <td style="width: 79pt;"><strong>MSE</strong></td> <td style="width: 79pt;"><strong>Persistence MAE</strong></td> <td style="width: 89pt;"><strong>Training Count</strong></td> <td style="width: 89pt;"><strong>Testing Count</strong></td> <td style="width: 70pt;"><strong>Inputs</strong></td> <td style="width: 131pt;">Model Name</td> </tr> <tr style="height: 64.0pt;"> <td style="height: 64.0pt; width: 120pt;">Next-hour FRP</td> <td style="width: 201pt;">Rolling FRP + climo hr + GFS vars hr (RF)</td> <td style="width: 79pt;">201.81</td> <td style="width: 79pt;">62769.5</td> <td style="width: 79pt;">471.38</td> <td style="width: 89pt;">272253</td> <td style="width: 89pt;">116680</td> <td style="width: 70pt;">28</td> <td style="width: 131pt;">RF_GFS_RAVE.sav</td> </tr> </tbody> </table>