Clearskytool is a tool in Matlab to select clear sky conditions from UV irradiance surface at 340 nm measured intervals of one-minute. We describe two methods to identify clear sky days.
1. Wavelet transform method
The algorithm detects the spectral signatures of the clouds and the magnitude of the noise in the daily. Wavelet method is based on the decomposition of the UV irradiance measured at 340 nm with one-minute interval. This method is based a similar approach to Djafer et al. (2017). The decision if we have or not a clear sky day is taken from the decomposition analysis considering some criterios. Firstly, a Gaussian adjustment curve is generated from the UV irradiance data for each measurement day. In our case, e.g., as a condition for a day to be selected as a clear sky condition, it is required that the determination coefficient (rsquare_in) be greater than or equal to 0.982, the Root Mean Square Error (rmse_in) less than 0.025 W m-2 nm-1 and the measurements per day (len_in) greater than 600 minute measurements per day. Once these conditions are satisfied, the wavelet transform method is applied to identify measurements influenced by clouds.
2. Normalized method
Normalized method is based on the calculation of normalized UV irradiance measured at 340 nm with one-minute interval, using the power law equation of the cosine of the solar zenith angle (SZA). This method was used to identify clear sky Global irradiance by Long and Ackerman (2000) and UV irradiance by Suárez Salas et al. (2017). The power law equation depending a and b regression coefficients. The a coefficient determine maximum (sup_lim) and minimum (inf_lim01, inf_lim02) threshold for our data set. The automation process to detect the a and b (bcoef) coefficients, which are representative of the entire irradiance dataset, lies in iteration. For first iteration, it can be proposed initial coefficients take from the scientific literature or found clear sky day using wavelet transform method. In our case, e.g., for first iteration, these initial coefficients were obtained from the scientific literature (e.g., Suárez Salas et al., 2017), bcoef equal to 1.30, sup_lim equal to 0.82 W m-2 nm-1, inf_lim01 equal to 0.62 W m-2 nm-1 and inf_lim02 to 0.58 W m-2 nm-1. After the first iteration, the algorithm uses the results from fitting the previosly detected clear sky measurements, and succeeding iterations refine the process to the actual characteristics of the themselves. The final values for the coefficients are obtained after the automation process.
Both methods identify clear sky days from UV irradiance measured at 340 nm with one-minute interval. In addition, normalized method can identify short periods by day and its process for find the a and b coefficients is automatic. In contrast, wavelet method only identifies completely clear days and the process for find rsquare_in and rmse_in coefficients is manual.
Christian Torres, [email protected]
Jose Flores, [email protected]
Luis Suarez [email protected]
wavelet_transform_method - The following code, it read data set contain in file of name e.g. guv_data_all_f.txt and identify clear sky days using wavelet transform method. In addition, it generate plots of the select days. Finally, it save a file of name e.g cs_data_uv_min_days_met1.txt, which contain only data clear sky days select.
normalized_method - The following code, it read data set contain in file of name e.g. guv_data_all_f.txt and identify clear sky conditions using normalized method. Finally, it save a file of name e.g cs_data_uv_min_met2.txt, which contain all data fitting to clear sky coefficients.
plot_clear_sky_days - The following script plot clear sky days identifies by wavelet transform and normalized methods. In addition, it identifies the days that overlap between the two methods.
wavelet_transform_method
dir work section
dir_out = dir output e.g. ['data/output/']
dir_graph = dir graphics e.g. ['graphics/'];
dir_in = dir input e.g. ['data/input/']
filename_in = file name data set e.g. ['guv_data_all_f.txt'];
filename_out = file output name days wavelet transform method e.g. ['cs_data_uv_min_days_met1.txt']
filename_out_thres = file output name threshold normalized method e.g. ['threshold_in_met2.txt']
input years and location section
years_total = years work e.g. [2018,2019,2020]
lon_s = logitude e.g. [-75.30]
lat_s = latiude e.g. [-12.04]
elv_s = elevation e.g. [3314.0]
zone_s = zone UTC e.g. [-5]
conditions select day section
rsquare_in = determination coefficient e.g. [0.982]
rmse_in = Root Mean Square Error e.g. [0.025]
len_in = the measurements per day e.g. [600]
normalized_method
dir work section
dir_out = dir output e.g. ['data/output/']
dir_graph = dir graphics e.g. ['graphics/'];
dir_in = dir input e.g. ['data/input/']
filename_in = file name data set e.g. ['guv_data_all_f.txt'];
filename_out = file output name min normalized method e.g. ['cs_data_uv_min_met2.txt']
input years and location section
years_total = years work e.g. [2018,2019,2020]
lon_s = logitude e.g. [-75.30]
lat_s = latiude e.g. [-12.04]
elv_s = elevation e.g. [3314.0]
zone_s = zone UTC e.g. [-5]
conditions select short period section
bcoef = b coefficient initial e.g. [1.30]
sup_lim = threshold max e.g. [0.82]
inf_lim01 = threshold min < 78.5° SZA e.g. [0.62]
inf_lim02 = threshold min > 78.5° SZA e.g. [0.58]
plot_clear_sky_days
dir work section
dir_out = dir output e.g. ['data/output/']
dir_graph = dir graphics e.g. ['graphics/']
dir_in = dir input e.g. ['data/input/']
filename_in = file name data set e.g. ['guv_data_all_f.txt']
filename_in1 = file output name days wavelet transform method e.g. ['cs_data_uv_min_days_met1.txt']
filename_in2 = file output name min normalized method e.g. ['cs_data_uv_min_met2.txt']
filename_out = file output name days normalized method e.g. ['cs_data_uv_min_days_met2.txt']
input years and location section
years_total = years work e.g. [2018,2019,2020]
lon_s = logitude e.g. [-75.30]
lat_s = latiude e.g. [-12.04]
elv_s = elevation e.g. [3314.0]
zone_s = zone UTC e.g. [-5]
conditions select short period section
len_in = the measurements per day e.g. [600]
To test the two methods, we used a data set UV irradiance collected one-minute intervals at the Observatory of Huancayo, localized in the central Andes of Peru. UV irradiance was measured a GUV-511 multi-channel filter radiometer manufactured by Biospherical Instruments Inc., San Diego, California. The radiometer has four channels in the UV region, centered at 305, 320, 340, and 380 nm. Data are available from April 2018 to January 2020 with some periods of inactivity. In the following table is showed header of the file guv_data_all_f.txt.
header data file guv_data_all_f.txt
| C | Variable name | Short Name | Unit |
|---|---|---|---|
| C1 | Year | year | - |
| C2 | Month | month | - |
| C3 | Day | day | - |
| C4 | Hour | hour | - |
| C5 | Minute | min | - |
| C6 | Second | sec | - |
| C7 | Irradiance at 305 | irra_305 | W m-2 nm-1 |
| C8 | Irradiance at 320 | irra_320 | W m-2 nm-1 |
| C9 | Irradiance at 340 | irra_340 | W m-2 nm-1 |
| C10 | Irradiance at 380 | irra_380 | W m-2 nm-1 |
| C11 | Erythemal dose maximum | dose_max | - |
| C12 | Erythemal dose mean | dose_mean | - |
| C13 | Erythemal dose minimum | dose_min | - |
Changes dir work, year and locate, and condition select selctions
The main changes should be made in the dir work, year and locate, and condition select sections. For example, for our data set and location:
wavelet_transform_method
dir work section
dir_out = ['data/output/']
dir_graph = ['graphics/'];
dir_in = ['data/input/']
filename_in = ['guv_data_all_f.txt'];
filename_out = ['cs_data_uv_min_days_met1.txt']
filename_out_thres = ['threshold_in_met2.txt']
input years and location section
years_total = [2018,2019,2020]
lon_s = [-75.30]
lat_s = [-12.04]
elv_s = [3314.0]
zone_s = [-5]
conditions select day section
rsquare_in = [0.982]
rmse_in = [0.025]
len_in = [600]
normalized_method
dir work section
dir_out = ['data/output/']
dir_graph = ['graphics/'];
dir_in = ['data/input/']
filename_in = ['guv_data_all_f.txt'];
filename_out = ['cs_data_uv_min_met2.txt']
input years and location section
years_total = [2018,2019,2020]
lon_s = [-75.30]
lat_s = [-12.04]
elv_s = [3314.0]
zone_s = [-5]
conditions select short period section
bcoef = [1.30]
sup_lim = [0.82]
inf_lim01 = [0.62]
inf_lim02 = [0.58]
plot_clear_sky_days
dir work section
dir_out = ['data/output/']
dir_graph = ['graphics/']
dir_in = ['data/input/']
filename_in = ['guv_data_all_f.txt']
filename_in1 = ['cs_data_uv_min_days_met1.txt']
filename_in2 = ['cs_data_uv_min_met2.txt']
filename_out = ['cs_data_uv_min_days_met2.txt']
input years and location section
years_total = [2018,2019,2020]
lon_s = [-75.30]
lat_s = [-12.04]
elv_s = [3314.0]
zone_s = [-5]
conditions select short period section
len_in = [600]
After the previous settings run wavelet_transform_method.mat, normalized_method.met and plot_clear_sky_days.mat
Results
Figure graphics\select_days_mets.png shows the clear sky days identified by both methods. The wavelet method identified 48 days of clear sky (graphics\select_days_mets.png a). While the normalization method identified 53 days (graphics\select_days_mets.png a). In total 33 days overlapped between the two methods.
The final values for the coefficients of the normalized method, obtained after the automatic process, are 1.44 for the coefficient b and 0.68 W m-2 nm-1 for the coefficient a, with a good level of accuracy (fitted line with R2 = 0.994) (graphics\regression_global_cosz.jpg).
Djafer, D., Irbah, A., and Zaiani, M. (2017). Identification of clear days from solar irradiance observations using a new method based on the wavelet transform. Renewable Energy, 101, 347-355. https://doi.org/10.1016/J.RENENE.2016.08.038
Long, C. N., and Ackerman, T. P. (2000). Identification of clear skies from broadband pyranometer measurements and calculation of downwelling shortwave cloud effects. Journal of Geophysical Research: Atmospheres, 105(D12), 15609-15626. https://doi.org/10.1029/2000JD900077
Suárez Salas, L. F., Flores Rojas, J. L., Pereira Filho, A. J., and Karam, H. A. (2017). Ultraviolet solar radiation in the tropical central Andes (12.0°S). Photochemical & Photobiological Sciences, 16(6), 954-971. https://doi.org/10.1039/C6PP00161K