Solar Declination Angle Calculator

Introduction to solar declination and the Sun's yearly north-south swing

Solar declination, written $\delta$, is the angular distance of the Sun's centre north or south of the celestial equator — the projection of Earth's equator onto the sky. It is one half of the Sun's geocentric equatorial coordinate pair; the other half is right ascension. Because Earth's rotation axis is tilted with respect to the plane of its orbit and keeps a nearly fixed direction in space over a single year, the Sun appears to migrate from roughly 23.44° south of the equator in late December to roughly 23.44° north in late June, and back again. That single angle drives the seasons, the length of the day at every latitude, and the geometry every fixed-tilt photovoltaic array is designed around.

The number this page reports is the apparent geocentric declination of the Sun at the instant you specify, computed with the algorithm published by the NOAA Global Monitoring Laboratory for its Solar Calculator, which is itself a transcription of the low-precision solar-coordinates method in Jean Meeus, Astronomical Algorithms. It is not the crude single-sine approximation that most declination pages use. That distinction matters: the classic $23.45°$ day-number formula is off by more than a degree on some dates, which is enough to move a computed sunrise time by several minutes and to mis-state a solar noon elevation by a visible amount.

Two things declination is not. It is not the Sun's elevation above your horizon, and it is not the solar zenith angle. Declination depends only on the moment in time; it is the same number for an observer in Reykjavik and one in Nairobi. Elevation and zenith angle additionally require your latitude and the hour angle, and this calculator derives them for you as separate, clearly labelled outputs so the three quantities are never conflated.

How to use the date, UTC time and latitude fields

Enter three things. Date (UTC) is the civil Gregorian calendar date. The engine converts it to a Julian Day number directly from the year, month and day, so leap years — including the century rule that makes 1900 an ordinary year and 2000 a leap year — are handled exactly, and no day-of-year lookup table is involved. Time of day (UTC) defaults to 12:00 because most published declination tables are quoted at noon; change it if you need a specific instant. Note that this field is Coordinated Universal Time, not your local clock: if you are in New York in summer you must add four hours to your wall clock, and if you are in Berlin in summer you must subtract two.

Observer latitude is optional to the declination itself but required for the derived quantities. Give it in decimal degrees, positive north and negative south, so Sydney is $-33.87$ and Anchorage is $+61.22$. Latitude never enters the declination calculation; it only feeds the solar-noon elevation, the zenith angle and the day-length figures. If you type a latitude outside ±90° or a non-numeric string the calculator refuses to run and tells you why rather than producing a silent $\mathrm{NaN}$.

Press Calculate and you get the declination in decimal degrees and in degrees-minutes-seconds, the day-of-year number, the Julian Day, the equation of time in minutes, the Cooper and Spencer approximations with their signed deviation from the rigorous value, and the latitude-dependent block. The annual curve below the results redraws for the selected year with your date marked, and the two approximate models can be overlaid so you can see exactly where and by how much they go wrong.

The declination formula chain, from Julian day to apparent solar longitude

The rigorous route to declination never uses a day number at all. It starts from the Julian Day $\mathrm{JD}$, converts that to Julian centuries since the epoch J2000.0, and builds the Sun's ecliptic longitude from a mean term plus an equation-of-centre correction. The time argument is

Formula: T = (JD − 2451545.0) / 36525

$$T=\frac{\mathrm{JD}-2451545.0}{36525}$$

The geometric mean longitude and geometric mean anomaly of the Sun, in degrees, follow as third- and second-order polynomials in $T$:

Formula: L_0 = 280.46646 + 36000.76983 ⁢ T + 0.0003032 ⁢ T^2

$$L_0=280.46646+36000.76983T+0.0003032T^2$$

Formula: M = 357.52911 + 35999.05029 ⁢ T − 0.0001537 ⁢ T^2

$$M=357.52911+35999.05029T-0.0001537T^2$$

Earth's orbit is an ellipse, so the Sun does not move along the ecliptic at a constant rate. The equation of centre $C$ converts the mean anomaly into the true one; it is the term the naive sine formula throws away.

Formula: C = (1.914602 − 0.004817 ⁢ T − 0.000014 ⁢ T^2) ⁢ sin ⁡ M + (0.019993 − 0.000101 ⁢ T) ⁢ sin ⁡ 2 M + 0.000289 ⁢ sin ⁡ 3 M

$$C=(1.914602-0.004817T-0.000014T^2)\sin M+(0.019993-0.000101T)\sin 2M+0.000289\sin 3M$$

Adding $C$ to $L_0$ gives the true longitude, and a small aberration-and-nutation correction keyed to the ascending node of the Moon's orbit gives the apparent longitude $\lambda$:

Formula: λ = L_0 + C − 0.00569 − 0.00478 ⁢ sin ⁡(125.04 − 1934.136 ⁢ T)

$$\lambda =L_0+C-0.00569-0.00478\sin (125.04-1934.136T)$$

The second ingredient is the obliquity of the ecliptic. It is emphatically not a constant. The mean obliquity ${\epsilon }_0$ shrinks by about 46.815 arcseconds — roughly 0.013° — per Julian century, and a nutation term of amplitude 0.00256° gives the true obliquity $\epsilon$:

Formula: ε_0 = 23 ° 26 ′ 21.448 ″ − 46.8150 ″ ⁢ T − 0.00059 ″ ⁢ T^2 + 0.001813 ″ ⁢ T^3

$${\epsilon }_0=23°26\prime 21.448″-46.8150″T-0.00059″T^2+0.001813″T^3$$

Formula: ε = ε_0 + 0.00256 ⁢ cos ⁡(125.04 − 1934.136 ⁢ T)

$$\epsilon ={\epsilon }_0+0.00256\cos (125.04-1934.136T)$$

Declination is then the single line that turns an ecliptic longitude into an equatorial one:

Formula: δ = arcsin ⁡(sin ⁡ ε ⁢ sin ⁡ λ)

$$\delta =\arcsin (\sin \epsilon \sin \lambda )$$

In mid-2026 that chain gives a mean obliquity of 23.4358° and a true obliquity of 23.4381°, so the maximum declination reached at the June solstice is 23.4381°, not 23.45°. Anyone quoting 23.45° is using a value rounded from a mid-twentieth-century epoch.

How the three published declination models compare

Three formulas dominate the literature. Cooper (1969) is a single sine in the day number $n$. Spencer (1971) is a three-harmonic Fourier series in the day angle $\Gamma =2\pi (n-1)/365$. The NOAA/Meeus chain above is the rigorous one.

Formula: δ_Cooper = 23.45 ° ⁢ sin ⁡ ((360 ⁢(284 + n)) / 365)

$${\delta }_{\text{Cooper}}=23.45°\sin \left(\frac{360(284+n)}{365}\right)$$

Formula: δ_Spencer = 0.006918 − 0.399912 ⁢ cos ⁡ Γ + 0.070257 ⁢ sin ⁡ Γ − 0.006758 ⁢ cos ⁡ 2 Γ + 0.000907 ⁢ sin ⁡ 2 Γ − 0.002697 ⁢ cos ⁡ 3 Γ + 0.00148 ⁢ sin ⁡ 3 Γ

$${\delta }_{\text{Spencer}}=0.006918-0.399912\cos \Gamma +0.070257\sin \Gamma -0.006758\cos 2\Gamma +0.000907\sin 2\Gamma -0.002697\cos 3\Gamma +0.00148\sin 3\Gamma$$

Spencer's result is in radians and is converted to degrees for display. The table below gives the three models for five instants in 2026, each at 12:00 UTC, with the signed deviation of the approximations from the NOAA/Meeus value. Every figure was produced by the same code that runs in this page.

The pattern is the one you would predict. Near the solstices the declination curve is flat, so a phase error in the day number costs almost nothing and every model agrees to about a hundredth of a degree. Near the equinoxes the declination sweeps through 0.395° per day and the same phase error becomes a large angular error. Sampled day by day at 12:00 UTC across 2020–2030, Cooper's peak deviation from the NOAA/Meeus value is 1.18° with an RMS of 0.53°, and Spencer's peak is 0.55° with an RMS of 0.29°. Spencer is roughly twice as good as Cooper, but both are an order of magnitude worse than the rigorous chain.

Turning declination into elevation, zenith angle and day length

Declination alone tells you nothing about the sky over your head. Combine it with latitude $\phi$ and hour angle $H$ and you get the solar zenith angle $\theta$:

Formula: cos ⁡ θ = sin ⁡ φ ⁢ sin ⁡ δ + cos ⁡ φ ⁢ cos ⁡ δ ⁢ cos ⁡ H

$$\cos \theta =\sin \phi \sin \delta +\cos \phi \cos \delta \cos H$$

Elevation is $90°-\theta$. At solar noon the hour angle is zero and the expression collapses to the familiar

Formula: α_noon = 90 ° − | φ − δ |

$${\alpha }_{\text{noon}}=90°-|\phi -\delta |$$

For the day length, NOAA solves the zenith equation for the hour angle at which the Sun's upper limb sits on the horizon, taking the standard 0.833° allowance for refraction plus the solar semi-diameter, so the zenith angle at sunrise is 90.833°:

Formula: H_0 = arccos ⁡ ((cos ⁡ 90.833 °) / (cos ⁡ φ ⁢ cos ⁡ δ) − tan ⁡ φ ⁢ tan ⁡ δ)

$$H_0=\arccos (\frac{\cos 90.833°}{\cos \phi \cos \delta }-\tan \phi \tan \delta )$$

Day length in hours is $2H_0/15$, because the Earth turns 15° of hour angle per hour. When the argument of the arccosine falls below −1 the Sun never sets and you have midnight sun; when it exceeds +1 the Sun never rises and you have polar night. This calculator detects both cases explicitly instead of returning $\mathrm{NaN}$.

Worked example: declination at 12:00 UTC on 21 June 2026 for a site at 40° N

Take the June solstice of 2026 and an observer at latitude 40.00° N. The Gregorian date 21 June 2026 at 12:00 UTC converts to Julian Day $\mathrm{JD}=2461213.0$, which puts the time argument at $T=0.264695$ Julian centuries after J2000.0.

Feeding $T$ through the polynomials gives a geometric mean longitude $L_0=89.7052°$ and a mean anomaly of 9886.3126°, which reduces to 166.3126° after removing whole turns. The equation of centre evaluates to $C=+0.4437°$ — already twenty times larger than the whole error budget of the method, which is why dropping it is not an option. True longitude is therefore 90.1489°, and after the aberration and nutation correction the apparent longitude is $\lambda =90.1454°$.

The obliquity polynomial returns a mean obliquity of 23.4358° and, with the 0.00256° nutation term, a true obliquity of $\epsilon =23.4381°$. The declination follows immediately:

Formula: δ = arcsin ⁡(sin ⁡ 23.4381 ° ⁢ sin ⁡ 90.1454 °) = + 23.4381 °

$$\delta =\arcsin (\sin 23.4381°\sin 90.1454°)=+23.4381°$$

Because the apparent longitude is within 0.15° of 90°, the sine is essentially 1 and the declination has all but reached the obliquity — the definition of a solstice. In degrees, minutes and seconds that is 23° 26′ 17″ N. The equation of time at the same instant is −1.82 minutes, meaning true solar noon runs about 1 minute 49 seconds later than mean solar noon at the observer's meridian.

Now bring in the latitude. Solar noon elevation is $90°-|40.00°-23.4381°|=73.44°$, so the zenith angle at noon is 16.56° and the Sun passes 16.56° to the south of the observer's zenith — close, but never overhead, because 40° N lies outside the tropics. The sunrise hour angle works out to 112.61°, giving a day length of 15.01 hours, or 15 h 01 min. Contrast this with the Cooper value for the same day, +23.4498°: only 0.012° adrift here at the solstice, but the same formula is 0.82° adrift on 23 September, which would shift the computed sunrise hour angle at 40° N by more than a degree, or roughly four minutes of clock time.

Interpreting the result and where the number actually gets used

Read the sign first. Positive declination means the subsolar point — the one place on Earth where the Sun is exactly overhead — lies in the northern hemisphere; negative means it lies in the southern. The magnitude tells you the latitude of that subsolar point. On 21 June 2026 the Sun is directly overhead somewhere along the 23.44° N parallel, which is the modern position of the Tropic of Cancer. Because the obliquity is shrinking by 0.013° per century, the tropics are creeping toward the equator by about 14 metres a year, and published tropic latitudes drift accordingly.

For fixed photovoltaic arrays, the tilt that puts a south-facing panel normal to the beam at solar noon is $|\phi -\delta |$, not the latitude alone. At 40° N that is 16.6° at the June solstice and 63.4° at the December solstice — a 47° annual swing, which is exactly why seasonally adjusted racking exists and why an annual-average fixed tilt close to the latitude is a compromise rather than an optimum. For tracker control, shading studies, daylight autonomy modelling and skylight sizing, the declination is the input that sets the entire seasonal envelope.

The equation of time reported alongside declination is a different quantity that is easily confused with it. It is the difference between apparent solar time and mean solar time, in minutes, and it comes from the same eccentricity and obliquity that shape the declination curve. Plotting declination against the equation of time over a year traces the analemma — declination on the vertical axis, equation of time on the horizontal.

Limitations of this declination model and the assumptions built into the page

Several assumptions are baked in, and being explicit about them is more useful than claiming a precision the method does not have.

• Accuracy is roughly 0.01°, not arcsecond-level. The implementation reproduces every intermediate column of the NOAA Solar Calculations spreadsheet to better than 10⁻⁹ degrees, but the underlying Meeus low-precision method is quoted at about 0.01° in solar longitude. Checked against the worked example in NREL/TP-560-34302 — whose stated uncertainty is ±0.0003° — this page's declination for 17 October 2003 at 19:30:30 UT differs by 0.0015°. If you need better, use the full NREL Solar Position Algorithm or JPL ephemerides.
• Terrestrial Time is not applied. The calculator treats your input as UT and does not add ΔT (about 69 seconds in 2026). That omission moves the declination by at most 0.0003° near an equinox and far less elsewhere, which is inside the model's own noise, but it means this page is not suitable for eclipse timing.
• The result is geocentric, not topocentric. No parallax, no observer elevation, and no atmospheric refraction is applied to the declination itself. Solar parallax is under 9 arcseconds (0.0025°), so the geocentric-to-topocentric difference is small, but refraction at the horizon is not: it is the reason the day-length calculation uses a 90.833° zenith and the reason apparent sunrise beats geometric sunrise by minutes.
• The polynomials are fitted, not exact. Meeus's series are truncated expansions valid over a limited span. NOAA warns that its own spreadsheet is only valid for 1901–2099 because of an approximation in the Julian Day step; the Julian Day routine used here is the full Meeus algorithm and works across the Gregorian era, but the accuracy claim should still be treated as applying to roughly 1800–2100. This page rejects years outside 1600–2400.
• Nutation is modelled by a single term. The 0.00478° longitude and 0.00256° obliquity corrections capture only the dominant 18.6-year lunar nodal term. The full nutation series has 63 terms.
• Cooper and Spencer are reported for comparison only. They are day-number formulas and therefore cannot know where in the leap-year cycle a date falls; a substantial part of their error is that phase slip rather than a defect in the fit. Do not treat their agreement near a solstice as evidence that they are accurate.
• Dates and times are UTC throughout. There is no time-zone or daylight-saving handling; converting local civil time to UTC is your responsibility. Day length is computed for the declination at the instant you entered, not integrated over the day, so it is a very good approximation rather than an exact sunrise-to-sunset interval.

Frequently asked questions about solar declination

Sources

The declination, obliquity and equation-of-time chain implemented here was transcribed cell by cell from the primary sources below and checked numerically against them.

• NOAA Global Monitoring Laboratory, Solar Calculation Details and the NOAA Solar Calculations (day) spreadsheet — the source of every polynomial used on this page. gml.noaa.gov/grad/solcalc/calcdetails.html. NOAA notes the calculator is no longer actively maintained and that the spreadsheets are valid for 1901–2099.
• Meeus, J., Astronomical Algorithms, Willmann-Bell — the book NOAA credits as the basis of the calculator; chapters on Julian Day, solar coordinates, nutation and the obliquity of the ecliptic.
• Reda, I. and Andreas, A., Solar Position Algorithm for Solar Radiation Applications, NREL/TP-560-34302, National Renewable Energy Laboratory, June 2003 (revised January 2008) — states uncertainties of ±0.0003° for the years −2000 to 6000; its Appendix A.5 worked example was used to check this page.
• United States Naval Observatory and HM Nautical Almanac Office, The Astronomical Almanac — the reference series against which the NREL SPA report benchmarks its own output.
• Cooper, P. I. (1969), "The absorption of radiation in solar stills", Solar Energy 12(3), 333–346 — the origin of the 23.45° sine approximation, shown here for comparison only.
• Spencer, J. W. (1971), "Fourier series representation of the position of the Sun", Search 2(5), 172; the coefficients as reproduced in Iqbal, M., An Introduction to Solar Radiation, Academic Press, 1983.

Deviation statistics quoted on this page (Cooper 1.18° peak, Spencer 0.55° peak over 2020–2030 at 12:00 UTC) were measured directly by evaluating all three models day by day against the NOAA/Meeus chain; they are our own measurements, not figures quoted from a third party.