Environmental light undergoes intensity and spectral changes over a 24-hour period of the day.
The available power distribution of solar spectrum is low. -
Dimensional models, such as CIE（
Clara International Committee)
Solar model, but the performance of this model-
The daytime system has no characteristics.
We measured the multi-day downflow spectral irradiance in two regions of North America (a rural area). (
Cherry Springs National Park, Pennsylvania)
Minimum Artificial Light Sources and a City Location（
We describe variations in spectra, intensity and color, and extend the existing daylight CIE model to capture low-light components and night sky spectra.
We measured, which is defined as light collected from the entire hemisphere of the measured surface.
The light arriving at the detector has the greatest weight in the surface direction.
The light is integrated according to the cosine of the angle of incident light.
The radiation spectrum of ground surge in northern Pennsylvania was measured with the least light pollution. (
Cherry Springs National Park, Pennsylvania)
Differences in the lunar cycle between June and July 2014.
On the night of June 30 and the morning of July 4, measurements were made under the new moon.
Full moon measurements were made on 11 and 12 July.
Finally, 60% and 49% of lunar illumination were measured between July 18 and 20.
In the same month, we also collected spectra from a four-storey roof under urban light pollution conditions. -
High-rise buildings near central Philadelphia, Pennsylvania.
This paper summarizes the date and range of the solar height, lunar height and the part of the irradiated moon.
Spectral radiation measurements of downward illumination spectra are performed every 60 seconds.
In order to capture the dynamic range of illumination, the integral time of the spectral radiometer is adjusted so that the maximum power of all wavelengths does not exceed about 85% of the maximum allowable intensity reading of the instrument, so as to avoid the saturation of spectral measurement under fast changing illumination conditions. S.
These were manually altered by the experimenter.
Because the sensitivity of the two spectrometers is different, the integration time used is also different, up to 60 s. -
Night Sensitivity Spectrometer.
The sky spectral power distribution was measured in Porter County, Pennsylvania, USA. (41. 6646° N, −77. 8125° W;
Elevation 710 m, NED point query service, National Elevation Data Set of USGS)
A certified International Dark Sky Park (IDA)).
The measuring position is located on a high ground in the forests of Susquehannock State, so there is no direct or indirect artificial light source.
During the measurements, special attention is paid to covering any stray light from the notebook computer that controls the spectrometer with a black cloth.
The Pennsylvania Department of Conservation and Natural Resources issued a permit to establish research camps.
We also measured the spectral power distribution of downwelling irradiance in urban environments. ()Beginning on the fifth-
Roof of David Ritterhouse Laboratory, Department of Physics and Astronomy, University of Pennsylvania（39. 952237° N, −75. 188734° W; elevation 12u2009m).
Because the electric light source used in the urban environment has a wide range of usability, the field measurements include natural light, street light and other artificial light sources, as well as the reflection of these light sources in the building environment.
The roof we measured was higher than the other roofs near it, providing a detector. -
The whole hemispherical sky, the roof is not directly illuminated by its own artificial light source.
For these reasons, the measuring position is sufficient to obtain the measurements of downwelling irradiance produced by natural and man-made sources. These downwelling irradiances are averaged over large distances in the city.
The ephemeris is generated by software, and the solar height, lunar height and lunar phase of the given terrain coordinates are extracted. (MICA Version 2. 2. 2).
The ephemeris is interpolated linearly to determine the solar height, lunar height and lunar phase of each measurement time and position.
Therefore, the reported true altitude does not represent a significant altitude taking into account local atmospheric conditions and refraction.
The method of converting the elevation in ephemeris into approximate apparent elevation can be used.
All measurements were performed using two custom USB spectrometers. (
USB2000+, Ocean Optics Company; Dunedin, FL)
Henceforth known as "A" and "B" spectrometers combine a habit. -
A downwelling irradiance detector was established.
"A" spectrometer is a manufacturer-
Optimized for high sensitivity measurement（
Sony ILX511B Linear Silicon CCD Array)
Between 180 and 875 nm.
This is used to measure dusk and night.
The sensitivity of "B" spectrometer is low, and the wavelength range is 340-1025 nm.
This is used to measure sunlight.
We used a combination of two spectrophotometers with a span factor of 10 in the daytime irradiance between noon and night.
Processing a large number of possible scene directions in human vision, while still able to detect low-brightness measurable light. -
Under low-light dusk and night conditions, we estimate the downward irradiance or upward impact of the light source for human vision in a given sky. -
Facing the plane, from the sky.
Sweeney and his colleagues, Johnsen and his colleagues described the technology to construct the downwelling irradiance detector. A single-mode, 3-
Diameter 1000 um, 5.
8 mm nominal external diameter, numerical aperture 0. 22u2009±u20090.
02, acceptance angle theta 12.
7 degrees, full angle 25. 4°;
Ocean Optics Corporation, Dunedin, FL)
It is connected to the spectrometer and fed through a tube in the downwelling irradiance probe.
In order to construct the probe, the measuring end of the optical fiber cable is pointed at the position of 10 to 45 degrees. 16u2009cm (4″)
Diameter Plexiglass Disc Coated with Bird Paint-
B White Reflective Coating（
Bird technology; New London, NH), a water-
Barium basic sulfate（BaSO)coating.
Ultimately, the 45 degree angle is arbitrary, and all the physical equivalent measurement needs is the unclosed patch of the scattering disc in the same direction for each measurement.
In fact, when the cable is at an angle close to 0 degrees from the normal condition, it is possible to see the difference of its own shadow on the disk. When the cable is at an angle close to 90 degrees from the normal condition, the light reflected by the disk may be lost when the cable moves slightly on the installation hardware. ISKAn O-
The environmental protection fixed around the cable maintains the distance between the cable and the reflector, so the distance between the cable and the reflector is constant.
The distance between the cable tip and the center of the reflector disk is 7. 74u2009cm.
We confirm that optical fibre cables can only detect the light reflected by disk by emitting halogen tungsten lamp. (LS-1;
Ocean Optics Corporation, Dunedin, FL)
Cable in a dark room.
When connected to a lamp instead of a spectrophotometer, only the reflector and other parts of the measurement module are illuminated.
This shows that the alignment and positioning of the optical fiber probe is good.
Lenovo ThinkPad X240 laptop running the Linux distribution xubuntu 14.
04 is used to control the spectrometer and record the measurement results.
Laptop computers are equipped with heat-swappable 9-
Batteries allow uninterrupted power supply during measurement.
Ocean Optics Corporation; Dunedin, FL)
And the custom software for MATLAB（Mathworks Inc. ; Natick, MA)
It is used to read and save the spectral power distribution of the spectrometer on the hard disk of the notebook computer.
There is a metal cover at the entrance of the spectrometer.
Two spectrophotometers were placed in a MyTemp miniature digital incubator. (
Scientific benchmarks; Edison, NJ).
The initial temperature of incubator is 24 C.
The incubator temperature was set to 10 degrees Celsius, and continuous darkness measurements were made.
Then, the incubator temperature was set to 60 C, and the measurement was continued.
According to the report of the internal sensor, the plate temperature in the spectrometer is measured with the dark spectrum.
This enables us to build a dark spectral database parameterized by integration time and temperature of the measuring board. We find that this is a significant parameter in a given dark spectral noise. ().
The dark spectrum most similar to the given measurement conditions in this library is used to process all individual spectra in the data set.
USB Spectrometer Exit Factory-
Wavelength calibration, each pixel on the sensor corresponds to a wavelength.
The calibration is verified by independent wavelength measurements of two line sources. (AS-361 Mercury [Hg]
Spectral calibration lamp; AS-364 Argon [Ar]
Spectral calibration lamp;
Spectral Products, Putnam, CT).
We calibrate factory calibration by comparing measured values with samples of known spectral lines. (404. 7, 435. 8, 546.
1. Mercury source 579 nm; 696. 5, 706. 7, 727. 3, 738. 4, and 763.
Argon source 5 nm).
These corrections are good. -
Approximate to an additive shift, and very small（
Both spectrometers are less than 1 nm.
It shows that the wavelength calibration of the factory is good.
The displacement required for the two spectrometers is_0. 79u2009±u20090. 57u2009nm (
"A" Spectrometerand −0. 98u2009±u20090. 37u2009nm (
"B" Spectrometeron average (±1SD)
The cross-spectral average（).
Although spectrometers are calibrated when they leave the factory, they are not calibrated in terms of power readings for each wavelength.
In order to incorporate the measurements into the absolute radiation calibration, we need to link the absolute spectral irradiance measurements reaching the measurement module disk with the original readings at each spectrometer wavelength.
Since the wavelength ranges of various calibration sources and measuring instruments available to us overlap only partially with the wavelength ranges of the two spectrophotometers, we use the following methods:（
A little involved)procedure.
We measured it with two spectrophotometers. (‘A’ and ‘B’)
Three Calibration Sources（
The relative spectra are shown below.
We studied the measurement signal. -to-
The noise ratio of these three noise sources (natick, ma)
The wavelength is corrected, and the average dark signal suitable for the integration time and temperature is subtracted. The "ground truth" and the measured spectrum are interpolated to 1. -
Relative correction coefficient a 2)
We divide the result by the integration time and get a noise. -
Correction measurements in units of uncalibrated power per second; 3)
We are from the wrong side. -
The calibration wavelength is sampled from 280 to 840 nm. (
And 360 and 840 nm（
Correct the wavelength calibration of the above factories. 4)
Then multiply by the wavelength. -
Using the correlation radiation correction coefficients found by the above program, the spectral radiation of illumination is obtained. -
Sample spectra were measured in WMSRNM. 5)
We multiply the emissivity by PI (the projected stereo angle on the hemisphere) and convert it into spectral irradiance in wmnm.
The saturated spectrum at any wavelength is removed from the analysis.
Sometimes, dark noise subtraction results in negative values, which we set to zero for further analysis.
We give up any spectrum that makes the spectrum zero at all wavelengths.
We use eight-order one-
Dimensional median filter using MATLAB medfilter 1 function（Mathworks Inc. ; Natick, MA).
Two independent additional data sets and two alternative models are considered in this paper.
From the authors, we obtained a data set of 10756 solar spectra collected by DiCarlo and Wandel in Stanford, California. (
Data Set of 2600 Solar Spectra（
Henceforth known as the Granada Spectrum.
Andr_s on the author's website（;
Date of visit: 14 December 2015.
From Wyszecki and Stiles and-
Component Model Derived from Granada Spectrum（
Henceforth known as the Granada Model.
From the author's website（;
Date of visit: 14 December 2015.
We used Webplotdigitzer to digitize the reference rural night sky spectra of Zabriske Point, CA and Cronin, Boston, Massachusetts. (
Date of visit: 15 December 2015.
We fit various models with spectral data, as shown below.
Let's first use spline to base functions. -
The nanometer spacing is then normalized by the vector norm. (L-norm).
Then we normalize each spectrum with vector norm, and find that-
Fitting Weight of Base Function with Minimum Value-
We calculated the ratio of variance. (R)
The model is used to explain the measured spectrum and fit it into the cross-wavelength square linear pair correlation coefficient.
The process described above produces the linear model weights needed to fit the normalized spectrum.
It is these standardized weights that are drawn in the component load diagram. (e. g. , ).
In the supplementary data, we provide the average normalized weights grouped by solar height. (and )
And the average scale factor needed to map the normalized spectrum of each elevation group back to the absolute spectrum.
The normalization, fitting and calculation of R ranges from 360 to 830 nm in our dataset and 380 to 780 nm in Granada, DiCarlo and Wandel datasets. (
For more information, see headings and, s2, S3 and s4).
In order to capture components in our data set that are not characterized by the CIE solar model, we use the following iteration method.
After fitting the CIE datum function, we extract all the average residuals and the spectrum of sunlight. (≥u20090)
The average of these residuals produces the first additional basis function, which we call CIE + basis function.
We refer to a model consisting of CIE sunlight model and CIE+1 basic function, which is regarded as CIE+1 model.
Then we fit all the spectra and spectra into the civil dusk. (−6°u2009