How to measure the variance of the random heterogeneity component needed in the generalized-weights meta-estimator?

Question

I am trying to do a metaanalysis of regression estimates, using the generalized-weights meta-estimator, because I have overlapping samples and thus need to take into account that the estimates are not independant. I use the method described in Bom & Rachinger (2020) https://doi.org/10.1002/jrsm.1441 , paragraph 2.5. I need to create a variance covariance matrix of the error term, and in the diagonal the values are : variance(error term) = variance(sampling error component) + variance(random heterogeneity component). The authors give a formula to calculate the variance(sampling error component) (=variance(regression error term)/(N*variance(x)), but they don't explain how to calculate the variance of the random heterogeneity component. If anyone has an answer or tips it would be amazing!