The comment on “Fast 3D inversion of gravity data using solution space priorconditioned Lanczos bidiagonalization” by Meng and Li is focused on Wk ∈ Rn×k matrix that is multiplied by kernel matrix G ∈ Rm×n for reducing the number of columns in the matrix from n to k (k bb n). Here, n is the number of model parameters. They assumed Wk × Wk −1 must equals identity matrix, then they proposed a new formulation for Wk ∈ Rk×k. Here, I will explain the misunderstanding of the assumption and the new formulation in the comment by Meng and Li.
