## The seasonal variation of the DC global electric circuit
# The seasonal variation of the DC global electric circuit
In this repository you can find the measurement and modelling data, the research code, and a comprehensive description of the code developed for our two papers on atmospheric electricity:
First paper (**Part 1: A new analysis based on long-term measurements in Antarctica**) focuses on the seasonal variation of the global electric circuit (GEC) based on electric field measurements at the Vostok station in Antarctica during 2006–2020. The key findings include:
@ -15,21 +15,21 @@ Second paper (**Part 2: Further analysis based on simulations**) further analyse
@@ -15,21 +15,21 @@ Second paper (**Part 2: Further analysis based on simulations**) further analyse
> **_Note:_** As soon as the articles are published, we will add links to them here.
## Short Description of the Scripts
## A Short Description of the Scripts
> **_Note:_** We use some data sets containing ionospheric potential values simulated via climate and weather forecasting models. Since these raw data sets are very large (around 350 Gb), we only upload preprocessed lower-dimensional data sets (around 20 Mb) to the repository. This preprocessing can be reproduced using the script `0_prepare_data.ipynb`, but this would require downloading the raw data sets from https://eee.ipfran.ru/files/seasonal-variation-2024/.
> **_Note:_** We use some data sets containing ionospheric potential values simulated via climate and weather forecasting models. Since these raw data sets are very large (around 350 Gb), we only upload preprocessed lower-dimensional data sets (around 30 Mb) to the repository. This preprocessing can be reproduced using the script `0_prepare_data.ipynb`, but this would require downloading the raw data sets from https://eee.ipfran.ru/files/seasonal-variation-2024/.
* `1_Earlier_measurements_images.ipynb`: the seasonal variation of the GEC based on the results of earlier studies (with the data taken from other sources)
* `2_Vostok_measurements_images.ipynb`: the seasonal variation and a diurnal-seasonal diagram for the GEC on the basis of potential gradient data from Vostok (both the new data for 2006–2020 and the earlier data for 1998–2001)
* `3_WRF_T2_images.ipynb`: the seasonal variation of air temperature at the height of 2 m (`T2m`) averaged over different latitude ranges
* `4_IP_simulations_temporal_images.ipynb`: the seasonal variation of simulated IP values for different models and parameterisations
* `5_IP_simulations_spatial_images.ipynb`: the decomposition of the seasonal variation into contributions to the IP from different latitude ranges and a new parameterisation of the IP
* `3_CAPE_T2_images.ipynb`: the seasonal variation of air temperature at the height of 2 m (`T2`) averaged over different latitude ranges and the distribution of CAPE in the tropics in different models
* `4_IP_simulations_temporal_images.ipynb`: the seasonal and diurnal variations of simulated ionospheric potential values for different models and parameterisations
* `5_IP_simulations_spatial_images.ipynb`: the decomposition of the seasonal variation into contributions to the ionospheric potential from different latitude ranges and a new parameterisation of the ionospheric potential
> **_Note:_** The scripts should be executed sequentially one after another; at the very least, scripts 4 and 5 should be run after script 2. This is necessary because script 2 saves intermediate arrays of preprocessed data from the Vostok station, which are used in scripts 4 and 5.
## Detailed Description of the Scripts
## A Detailed Description of the Scripts
> **_Note:_** Here and throughout the script descriptions, it will be specified which calculations correspond to which figures. The figure number is presented in the format 1.x or 2.x, where the first digit refers to the article (Part 1 or Part 2) and the second digit indicates the figure number within the article. If the figure number starts with the letter "S" (e.g., 1.S1), it refers to a figure in the Supporting Information for the corresponding article.
> **_Note:_** Here and throughout the script descriptions, it will be specified which calculations correspond to which figures. The figure number is presented in the format 1.x or 2.x, where the first digit refers to the article (Part 1 or Part 2) and the second part indicates the figure number within the article. If the figure number starts with the letter "S" (e.g., 1.S1), it refers to a figure in the Supporting Information for the corresponding article.
### 1. The script `1_Earlier_measurements_images.ipynb`
@ -81,6 +81,8 @@ Since in paper we focus on studying monthly mean values, using hourly dataframes
@@ -81,6 +81,8 @@ Since in paper we focus on studying monthly mean values, using hourly dataframes
Such daily-resolution arrays are stored in the variables `daily_df` and `daily_edf`.
> **_Note:_** At this stage we also save the average hourly values in the temporary file `vostok_diurnal_2006_2020.npy` to facilitate plotting later figures.
#### 2.5. Constructing Figure 1.2, Figure 1.S1 and Figure 1.S2
Figure 1.2 illustrates the seasonal variation (based on the new data) across different multi-year periods. To create it, we used the prepared data (daily_df) and helper functions to compute the mean values, the number of fair-weather days and the standard errors for three data sets:
@ -89,7 +91,7 @@ Figure 1.2 illustrates the seasonal variation (based on the new data) across dif
@@ -89,7 +91,7 @@ Figure 1.2 illustrates the seasonal variation (based on the new data) across dif
2. The same data for 2006–2012.
3. The same data for 2013–2020.
> **_Note:_** The data from this figure are saved in the temporary file `vostok_2006_2020_results.npz` for use in the second article. This helps us to avoid duplicating the code or merging the code to build different entities in a single cumbersome file.
> **_Note:_** The data from this figure are saved in the temporary file `vostok_2006_2020_results.npz` for use in plotting later figures. This helps us to avoid duplicating the code or merging the code to build different entities in a single cumbersome file.
Figure 1.S1 illustrates the seasonal variation (based on the new data) for individual years. Using the same process as was employed in the case of Figure 1.2, we compute the mean values, the number of fair-weather days and standard errors, but this time for each year separately (we obtain 15 data sets, one for each year in 2006-2020).
@ -123,9 +125,11 @@ The procedure for identifying correlations, determining the regression coefficie
@@ -123,9 +125,11 @@ The procedure for identifying correlations, determining the regression coefficie
We apply the solar-wind-imposed correction to two data sets: `meteo_solar_df` (obtained from merging `meteo_df` with `swip_df`; here the potential gradient values are adjusted for both meteorological influences and the solar wind) and `solar_df` (which takes into account only the correction for solar wind).
##### 2.7.3. Construct Figure 1.5 and Figure 1.S3
> **_Note:_** The adjusted data are saved in the temporary file `vostok_2006_2020_results_adjusted.npz` for use in plotting later figures.
#### 2.8. Constructing Figure 1.5 and Figure 1.S3
Finally, we construct Figures 1.5 and 1.S3 using the prepared data in the same manner as it was done for Figures 1.2 and 1.3.
We construct Figures 1.5 and 1.S3 using the prepared data in the same manner as it was done for Figures 1.2 and 1.3.
Figure 1.5 presents the monthly mean fair-weather potential gradient at the Vostok station under different conditions:
@ -140,13 +144,27 @@ Each subplot visualises the seasonal variation for the corresponding data set, w
@@ -140,13 +144,27 @@ Each subplot visualises the seasonal variation for the corresponding data set, w
Figure 1.S2 shows the same data as panel 1 of Figure 1.5, but the analysis is based on our own calculations rather than on those by Burns et al. (2012).
### 3. The script `3_WRF_T2_images.ipynb`
#### 2.9. Hourly data adjustment for future use
Finally, we perform procedures similar to those described in section 2.7 above to adjust hourly values of the potential gradient at Vostok for meteorological influences and for the solar-wind-imposed potential difference (we will need these data later). To this end, we employ a 6-hour time series of meteorological parameters (namely, temperature and wind speed at 0:00, 6:00, 12:00 and 18:00 UTC) and interpolate them to estimate values at 0:30, 1:30, 2:30 UTC and so on (`full_temp_df`, `full_wind_df`); we merge these data with hourly electric field values (corresponding to averages over 0:00–1:00, 1:00–2:00, 2:00–3:00 UTC and so on) and with the hourly data from `swip_df` (also corresponding to 0:30, 1:30, 2:30 UTC and so on) to obtain the data set `full_meteo_solar_df`. To remove the unnecessary variability from the potential gradient data, we use the same regression coefficients earlier determined for the diurnal mean values.
> **_Note:_** The average adjusted hourly values are saved in the temporary file `vostok_diurnal_2006_2020_adjusted.npy` to facilitate plotting later figures.
### 3. The script `3_CAPE_T2_images.ipynb`
This script calculates the seasonal variation of air temperature at the height of 2 m (`T2`), derived from global simulations of atmospheric dynamics, and presents distributions of CAPE values in the tropics (more precisely, within 30°S–30°N) in different models.
This script calculates the seasonal variation of air temperature at the height of 2 m (`T2m`), derived from global simulations of atmospheric dynamics.
#### 3.1. Plotting Figures 1.4 and 2.4
In the script the average temperatures for different latitudes and months are loaded from `WRF_T2_MONxLAT.npy` (see the data description below).
The temperature is then averaged across the latitude ranges of 20°S–20°N, 30°S–30°N, 40°S–40°N and 50°S–50°N, taking into account the area variation with latitude (1° wide cells at higher latitudes are weighted with a diminishing coefficient). The results of the averaging (the seasonal variation of temperature within the specified latitude bands) are presented in Figures 1.4 and 2.3, consisting of four panels each.
The temperature is then averaged across the latitude ranges of 20°S–20°N, 30°S–30°N, 40°S–40°N and 50°S–50°N, taking into account the area variation with latitude (1° wide cells at higher latitudes are weighted with a diminishing coefficient). The results of the averaging (the seasonal variation of temperature within the specified latitude bands) are presented in Figures 1.4 and 2.4, consisting of four panels each.
#### 3.1. Plotting Figure 2.S1
The source data for CAPE distributions is downloaded from `WRF_CAPE_HIST.npz`, `INMCM_SCAPE_HIST.npz`, `WRF_CAPE_RAIN_HIST.npz` and `INMCM_SCAPE_RAIN_HIST.npz` (see the data description below).
On the basis of these data the script plots Figure 2.S1, which shows histograms of CAPE for all grid columns and hours within 30°S–30°N, as well as for those grid columns and hours within the same range for which hourly precipitation is non-zero.
### 4. The script `4_IP_simulations_temporal_images.ipynb`
@ -158,33 +176,39 @@ The import of libraries and helper functions in this script is similar to the pr
@@ -158,33 +176,39 @@ The import of libraries and helper functions in this script is similar to the pr
The first step in the script is loading the time series of the modelled ionospheric potential. Since there are quite a few of them (calculated for different models with various parameterisation settings), for consistency they are loaded into two dictionaries: `wrf_hourly_total_ip` and `inm_hourly_total_ip`. To facilitate temporal grouping, two auxiliary index arrays, `<model>_dt_indices`, of the same length as the arrays in the dictionaries, are also created; these arrays store the series of dates corresponding to the respective ionospheric potential arrays.
| Dictionary | Key | IP Time Series |
| Dictionary | Key | Ionospheric Potential Time Series |
|---|---|---|
| `wrf_hourly_total_ip` | 500 | WRF model, CAPE > 500 kJ/kg, new parameterisation with T2m > 25 °C |
| `inm_hourly_total_ip` | 600 | INMCM model with CAPE > 600 J/kg |
| `inm_hourly_total_ip` | 800 | INMCM model with CAPE > 800 J/kg |
| `inm_hourly_total_ip` | 1000 | INMCM model with CAPE > 1000 J/kg |
| `inm_hourly_total_ip` | 1200 | INMCM model with CAPE > 1200 J/kg |
##### 4.1. Constructing Figure 2.1
#### 4.1. Constructing Figure 2.2
Here we average the data from the loaded arrays in the dictionaries (data sets for both models with the usual parameterisation and CAPE thresholds of 800, 1000 and 1200 J/kg) over months and construct seven plots.
Here we average the data from the loaded arrays in the dictionaries (data sets for both models with the usual parameterisation and CAPE thresholds of 600, 800, 1000 and 1200 J/kg) over months and construct eight plots.
Each of these plots represents the seasonal variation for a specific model and set of parameters. Note that monthly grouping for seasonal averaging is implemented using the auxiliary index arrays described above.
Six plots use the modelling data, while the seventh is based on the data from the Vostok station. It should be noted that here we use the output from the script `2_Vostok_measurements_images.ipynb`.
Eight plots use the modelling data, while the final two are based on the data from the Vostok station (before and after adjustments). It should be noted that here we use the output from the script `2_Vostok_measurements_images.ipynb`.
#### 4.2. Constructing Figure 2.1
Here we plot average diurnal variaiton curves corrsponding to the same data sets which were earlier used to plot Figure 2.2. We also use some output from the script `2_Vostok_measurements_images.ipynb`.
##### 4.2. Constructing Figure 2.5
#### 4.3. Constructing Figures 2.S2 and 2.S3
Figure 2.5 is constructed similarly to Figure 2.1. However, here specific summer periods are selected for averaging prior to plotting.
Figures 2.S2 and 2.S3 are constructed similarly to Figure 2.1. However, here specific periods are selected for averaging prior to plotting.
### 5. The script `5_IP_simulations_spatial_images.ipynb`
This script performs a somewhat more complex calculation. It uses the same time-averaging approaches as the previous one but additionally decomposes the IP into contributions from different latitude ranges.
This script performs a somewhat more complex calculation. It uses the same time-averaging approaches as the previous one but additionally decomposes the ionospheric potential into contributions from different latitude ranges.
This script also introduces the results obtained using a new IP parameterisation, which includes the surface air temperature (`T2m`).
This script also introduces the results obtained using a new ionospheric potential parameterisation, which includes the surface air temperature (`T2`).
> **_Note:_** The description of some helper functions and helper arrays will not be provided here, as they are identical to those used in Scripts 1–4.
@ -213,11 +237,11 @@ The loaded arrays and dictionaries of arrays (see the description in Script 4) a
@@ -213,11 +237,11 @@ The loaded arrays and dictionaries of arrays (see the description in Script 4) a
After loading these data, various seasonal variation characteristics are calculated for different latitude ranges. These include the contribution of each latitude range to the total ionospheric potential (as a percentage), the amplitude of the seasonal variation, the months when the minimum and maximum of the seasonal variation are attained.
#### 5.3. Constructing Figures 2.2 and 2.4
#### 5.3. Constructing Figures 2.3 and 2.5
For Figure 2.4 we decompose the seasonal variation of the ionospheric potential into contributions and show them cumulatively. The total value in the figure is composed of contributions from specific latitude ranges: `["0-9N", "0-9S", "9S-18S", "18S-90S", "9N-18N", "18N-90N"]`.
For Figure 2.5 we decompose the seasonal variation of the ionospheric potential into contributions and show them cumulatively. The total value in the figure is composed of contributions from specific latitude ranges: `["0-9N", "0-9S", "9S-18S", "18S-90S", "9N-18N", "18N-90N"]`.
Figure 2.2 uses a similar approach to averaging the arrays; however, the contributions of individual latitude strips are plotted in separate subplots. Additionally, seasonal variation patterns for the average temperature across different latitude ranges are shown.
Figure 2.3 uses a similar approach to averaging the arrays; however, the contributions of individual latitude strips are plotted in separate subplots. Additionally, seasonal variation patterns for the average temperature across different latitude ranges are shown.
> **_Note:_** While the ionospheric potential contributions from individual grid cells within latitude ranges can be summed (they are additive), this is not the case for temperature. For temperature we must calculate a weighted average, where the weighting function takes into account the variability in the grid cell area.
@ -225,9 +249,9 @@ Figure 2.2 uses a similar approach to averaging the arrays; however, the contrib
@@ -225,9 +249,9 @@ Figure 2.2 uses a similar approach to averaging the arrays; however, the contrib
This figure, which illustrates the simulated diurnal and seasonal variations according to different parameterisations, introduces the data from the new parameterisation that takes into account surface air temperature.
The data processing procedure for the seasonal variation is similar to that used in Figure 2.4. For the diurnal variation the arrays are created based on the hourly time series of the total ionospheric potential (`wrf_hourly_total_ip`), which are used exclusively for this figure.
The data processing procedure for the seasonal variation is similar to that used in Figure 2.5. For the diurnal variation the arrays are created based on the hourly time series of the total ionospheric potential (`wrf_hourly_total_ip`), which are used exclusively for this figure.
Note that we again utilise the data from the Vostok station to construct the figure. However, here we load two preprocessed data sets from previous scripts: one for the diurnal variation (`vostok_diurnal_2006_2020.npy`) and one for the seasonal variation (`vostok_2006_2020_results.npz`).
Note that we again utilise the data from the Vostok station to construct the figure. Here we load two preprocessed data sets from previous scripts: one for the diurnal variation (`vostok_diurnal_2006_2020_adjusted.npy`) and one for the seasonal variation (`vostok_2006_2020_results_adjusted.npz`).
<!-- ### 6. Calculation of SWIP using the Weimer 2005 model -->
<!-- ... -->
@ -239,14 +263,17 @@ Note that we again utilise the data from the Vostok station to construct the fig
@@ -239,14 +263,17 @@ Note that we again utilise the data from the Vostok station to construct the fig
- `vostok_1998_2004_hourly_80percent_all.tsv`: Hourly potential gradient data for 1998–2004, based on earlier measurements with a different field mill. Contains two columns: `Datetime` (date and time) and `Field` (potential gradient values in V/m).
- `vostok_1998_2020_hourly_SWIP.txt`: Simulated (with the Weimer 2005 model) hourly solar wind imposed ionospheric potentials over the Vostok station for 1998–2020.
- `vostok_daily_pressure_mm_hg.csv`, `vostok_daily_temp.csv`, `vostok_daily_wind.csv`: CSV files containing daily averaged atmospheric pressure, temperature and wind speed at the Vostok station.
- `vostok_daily_pressure_mm_hg_6h.csv`, `vostok_daily_temp_6h.csv`, `vostok_daily_wind_6h.csv`: CSV files containing daily atmospheric pressure, temperature and wind speed at the Vostok station every 6 hours (the pressure data are not actually used here, given that it did not show a statistically significant correlation with potential gradient values).
#### `./data/WRF/`
- `WRF_HOURLY_TOTAL_IP_[500/800/1000/1200].npy`: `numpy` arrays containing hourly total ionospheric potential values simulated with the WRF model, assuming the CAPE threshold of 500, 800, 1000 or 1200 J/kg. Note that for the threshold of 500 J/kg a new parameterisation is applied, which also assumes a 25 °C threshold in temperature; the same applies to the next item on the list.
- `WRF_IP_[500/800/1000/1200]_LATxMON.npy`: `numpy` arrays with dimensions `(latitude, 12)`, containing monthly average contributions to the ionospheric potential from different latitude ranges simulated with the WRF model, assuming the CAPE threshold of 500, 800, 1000 or 1200 J/kg.
- `WRF_HOURLY_TOTAL_IP_[500_T2_25/600/800/1000/1200].npy`: `numpy` arrays containing hourly total ionospheric potential values simulated with the WRF model, assuming the CAPE threshold of 500, 600, 800, 1000 or 1200 J/kg. Note that for the threshold of 500 J/kg a new parameterisation is applied, which also assumes a 25 °C threshold in temperature; the same applies to the next item on the list.
- `WRF_IP_[500_T2_25/600/800/1000/1200]_LATxMON.npy`: `numpy` arrays with dimensions `(latitude, 12)`, containing monthly average contributions to the ionospheric potential from different latitude ranges simulated with the WRF model, assuming the CAPE threshold of 500, 600, 800, 1000 or 1200 J/kg.
- `WRF_NUMDAYS_MON.npy`: A `numpy` array indicating the number of fair-weather days included in the monthly averages.
- `WRF_T2_LATxMON.npy`: A `numpy` array with the shape `(180, 12)`, containing montly averaged temperatures at the height of 2 m (`T2m`) for each 1° wide latitude band. `T2m` are calculated with the WRF.
- `WRF_CAPE_HIST.npz`, `WRF_CAPE_RAIN_HIST.npz`: Values and 10-J/kg bin boundaries to plot histograms of CAPE within 30°S–30°N simulated with the WRF model (for all grid columns and hours and for those with non-zero precipitation)
#### `./data/INMCM/`
- `INMCM_HOURLY_TOTAL_IP_[800/1000/1200].npy`: `numpy` arrays containing hourly total ionospheric potential values simulated with the INMCM model, assuming the CAPE threshold of 800, 1000 or 1200 J/kg.
- `INMCM_IP_[800/1000/1200]_LATxMON.npy`: `numpy` arrays with dimensions `(latitude, 12)`, containing monthly averaged contributions to the ionospheric potential from different latitude ranges simulated with the INMCM model, assuming the CAPE threshold of 800, 1000 or 1200 J/kg.
- `INMCM_NUMDAYS_MON.npy`: A `numpy` array indicating the number of fair-weather days included in the monthly averages.
- `INMCM_HOURLY_TOTAL_IP_[600/800/1000/1200].npy`: `numpy` arrays containing hourly total ionospheric potential values simulated with the INMCM model, assuming the CAPE threshold of 600, 800, 1000 or 1200 J/kg.
- `INMCM_IP_[600/800/1000/1200]_LATxMON.npy`: `numpy` arrays with dimensions `(latitude, 12)`, containing monthly averaged contributions to the ionospheric potential from different latitude ranges simulated with the INMCM model, assuming the CAPE threshold of 600, 800, 1000 or 1200 J/kg.
- `INMCM_NUMDAYS_MON.npy`: A `numpy` array indicating the number of fair-weather days included in the monthly averages.
- `INMCM_SCAPE_HIST.npz`, `INMCM_SCAPE_RAIN_HIST.npz`: Values and 10-J/kg bin boundaries to plot histograms of surface CAPE within 30°S–30°N simulated with the INMCM model (for all grid columns and hours and for those with non-zero precipitation)
