Mathematically a conformal map has the property that the Jacobian of the transformation is a scaled version of a rotation matrix. The transformation involved in a spherical Lambert conformal conic map projection doesn't have that property. In what sense is the LCC conformal?


Map projection in pseudocode (x,y are the map coordinates); I'm using the notation as described on pg. 104 of Map Projections - a Working Manual


Jacobian (phi,lam refer to latitude, longitude respectively):

J =   (drho/dphi)*sin(theta)     rho*cos(theta)*(dtheta/dlam)
     -(drho/dhpi)*cos(theta)     rho*sin(theta)*(dtheta/dlam)          

This would be a scaled rotation matrix if (drho/dphi)==rho*(dtheta/dlam) however, since

 rho=R*F*cot( phi/2+pi/4)**n

(R,F,n, lam0 are constants derived from the parameters defining the projection.) we have

drho/dphi=-n*R*F*cot( phi/2+pi/4)**(n-1)*csc( phi/2+pi/4)*0.5


Thus, the Jacobian is not a scaled version of a rotation matrix, therefore the LCC is not a conformal map.

  • 2
    You have computed a Jacobian as if the geographic coordinates (phi, theta) were Euclidean--but they are not. Please read the introductory sections of the Manual, especially "Distortion for Projections of the Sphere." The check for conformality is that h = k in formulae (4-4) and (4-6), p. 23.
    – whuber
    May 6 '15 at 15:11
  • @whuber - penetrating comment again. Are you hinting that the geographic coordinates(phi,theta) are Riemannian ?
    – gansub
    Jan 21 '16 at 9:24
  • @gansub Could you explain what you mean by "Riemannian coordinates," if it's intended to be more specific than any set of coordinate functions for a patch on a Riemannian manifold?
    – whuber
    Jan 21 '16 at 14:06
  • @whuber - No I mean you had written that the geographic coordinates(phi, theta) were not Euclidean. So I was wondering whether that was a reference to an alternate space either Riemannian or some arbitrary space.
    – gansub
    Jan 21 '16 at 14:54
  • @gansub There is a Riemannian metric (determined by the ellipsoid) but it is not Euclidean, because the ellipsoid is nowhere flat.
    – whuber
    Jan 21 '16 at 14:58

As per whuber's comment: differentiating with respect to the angular coordinates is incorrect, i.e. it is not a conformal map between (lat,lon) and (x,y) map coordinates; instead it is a local mapping from (local) easterly, northerly displacement onto map displacements. Thus, you need to differentiate w.r.t. easterly and northerly displacements in meters (or whatever length unit):

J =  (dx/de)     (dx/dn)   
     (dy/de)     (dy/dn)

  =   rho*cos(theta)*(dtheta/dlam)*(dlam/de)    (drho/dphi)*(dphi/dn)*sin(theta)
      rho*sin(theta)*dtheta/dlam)*(dlam/de)    -(drho/dphi)*(dphi/dn)*(dlam/de)*cos(theta)

with a bit of algebra you end up with:

J = (n*rho)/R/cos(phi)* |  cos(theta)    -sin(theta) |
                        |  sin(theta)     cos(theta) |

It's pretty straightforward to verify that k==h==(n*rho)/R/cos(phi) as a check.

  • 1
    +1 Great self-answer. It's nice to see some rigorous analysis of projections appear on this site.
    – whuber
    May 6 '15 at 17:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.