This paper discusses the necessary and sufficient condition for the existence of a non-negative definite quadratic unbiased estimator (MINQE(U,NND)) of variance components, and explains theoretically why some negative estimators arise in MINQE (I,U) at times. From this conclusion variance components ma2
of distance observations cannot be determined simultaneously and accurately by the MINQE in the direction-distance net adjustment. Furthermore, the analysis of some examples in the paper gives the same result. Therefore, considering the ease that ma2
is known generally, an improved variance component estimation model is provided for the direction-distance net adjustment, according to which we can obtain unbiased estimates of both direction and distance variance components and arrive at a reasonable solution to the above examples. In orther words, we can determine more appropriate and reasonable weights in the direction-distance net adjustment.