The appellees were awarded judgment for $55,120.25 against appellants, based on appellees’ claim of ownership of thirty-five acres of land in Knott County from which appellant Knott Coal Corporation (as lessee of appellant Kentucky River Coal Corporation) had mined 220,481 tons of coal. The appellants present numerous grounds upon which they deem themselves entitled to reversal of the adverse judgment, and the appellees have perfected a cross-appeal asserting entitlement to interest on the judgment from the date of the filing of their suit.
The land in controversy (from which the coal was mined) lies on the north side of the Kelly Fork of Lotts Creek. It is a mountainous region. The description in the deed under which appellees claim is so confusing (about which there will be further discussion in this opinion) that ap-pellees were unable to establish that the disputed land lay within the boundaries recited in the deed. However, it appears tacitly agreed that appellees’ 35-acre tract lies within the boundary of a 500-acre patent issued to John Cornett in 1851. If appellees could satisfactorily show that the disputed land is within the latter boundary, then their claim must be upheld. On the other hand, appellants sought to prove that the land in dispute is embraced within the boundary of a junior, 200-acre patent issued to William Kelly in 1888. The two patents do not overlap and the latter patent specifically refers to a beginning corner of the John Cornett survey. The opening lines of the description in the William Kelly patent are:
“Lying and being in the County of Knott on both sides of Lotts Creek and is bounded as follows; Beginning at two white oaks in the head of the Deadening Hollow S. 66 E. 22 Poles from a beginning corner of a survey made in the name of John Cornett * * (Emphasis added.)
Thus, it is seen that the two patents relate to each other, and the proper location of the starting point of either will fix the first call of the other. In substance, the evidence and maps offered for the appellees locate the starting points of these two patents about 2,400 feet east of their positions as asserted for appellants. One of the principal controversies in this case involves the location of Deadening Hollow.
*255The Chancellor’s Findings of Fact are somewhat confusing. The first finding is in the following language:
“The Court is of the opinion that the property in dispute is possibly a part of a tract of land referred to in the evidence as the 500-acre John Cornett patent made July 1, 1851, but since the court finds that the lands herein lie within a well marked boundary claimed adversely by plaintiffs for more than 15 years that would not be necessary.” (Emphasis added.)
This indicates rather clearly that the Chancellor was not determining the issue involving the correct location of the Cornett and Kelly patents, and his adjudication appears to have been based on the finding of adverse possession rather than paper title. However, near the conclusion of the Findings the following appears:
“The Court concludes that the property in dispute belongs to the plaintiffs, and that the property comes within the confines and boundaries of the 500-acre John Cornett patent, since Deadening Hollow is located above Walnut Log or Road Fork Hollow.”
These alternative findings are somewhat inconsistent and tend to weaken the basis of the Chancellor’s adjudication. Apparently both parties assume that the crux of this case is the proper location of the two related patents. But first we must consider the question of adverse possession.
In our view the record falls short of supporting the trial court’s finding that the appellees acquired title by adverse possession to the surface, much less the minerals. Actual adverse possession of the land in dispute was simply not proven. This land is mountainous, rugged and wooded. While one of the appellees occupied a home on the south side of Kelly Fork, this was not on the property in dispute and there was no actual open and notorious dominion over or possession of the land under which coal was mined. That appellees were not in actual possession of this tract is demonstrated by the fact that while mining penetrated through the area from February 1943 to April 1945, appellees did not discover this until several years later.
Nor can it be held that appellees’ alleged adverse holding was based on constructive possession. The deed under which they claim, dated February 16, 1919, was not recorded until 1948, just two years before the suit was filed. This may have been sufficient as “color of title,” even though unrecorded. Cf. Culton v. Simpson, 265 Ky. 343, 96 S.W.2d 856. However, the description in the deed is so vague as to bring it within the rule proscribing “extension of possession” to the limits of the description because not described with reasonable certainty. See Lipps v. Turner, 164 Ky. 626, 176 S.W. 42; Fordson Coal Co. v. Roark, 214 Ky. 247, 283 S.W. 106; Ball v. Moss, 218 Ky. 617, 291 S.W. 1042; Holcomb v. Swift Coal & Timber Co., 251 Ky. 642, 65 S.W.2d 741. The actual description in the deed is as follows:
“Lying and being on the right hand fork of Lotts Creek, a fork of the North Fork of the Kentucky River, and bounded as follows, to-wit: Beginning at a conditional line between John Kelly, W. Me. Kelly in the gap of the mountain; thence-running down the top of the ridge with a marked line to a marked rock; thence down the ridge to a 50 acre survey in the name of John Kelly and John Kelly’s fence; thence running with the line of a 100 acre survey made in the name of John Cornett; thence with Lotts Creek to the beginning so as to include all the land held by John Kelly and others inside of this 100 acre survey made in the name of John Cornett. This is to make all those parties rights and titles back to the grantor which he deeded to them at the above date mentioned, containing 35 acres more or less.”
Though there is abundant reference in the record and briefs to the “conditional *256line” between John Kelly and W. Me. Kelly, which is the beginning point of this description, we cannot find that either party was able accurately to locate it. (That is one of the reasons the parties resorted to the more easily understood descriptions in the Cornett and William Kelly patents.) It is practically impossible from this description to determine the boundary lines of appellees’ 35-acre tract. Therefore, there is no basis for claiming constructive possession of the whole tract.
Even if there had been sufficient proof of adverse possession of the surface, this would not constitute adverse possession of the minerals. The mineral deeds created a severance of the surface and mineral estates before the date of the surface deed under which the appellees claim. The surface deed for the 35-acre tract was dated February 16, 1919, but not recorded until October 12, 1948. The mineral deeds, creating the severance, were recorded before 1919. It is well settled that adverse possession of the surface alone will not create title to the underlying minerals if there has been a severance of the surface and mineral estates. KRS 381.430; Griffith v. J. C. Miller Oil Co., Ky., 349 S.W.2d 833; Inland Steel Co. v. Isaacs, Ky., 291 S.W.2d 522-; 13A Ky.Digest, Mines and Minerals, ®^49. There is not even a contention by appellees that they adversely held the mineral as such. Cf. Diederich v. Ware, Ky., 288 S.W.2d 643. In these circumstances, the asserted adverse possession of the surface had no effect upon the title to the minerals. The Chancellor’s finding of adverse possession is clearly erroneous.
We now come to the major issue in this litigation. It is whether the land in controversy lies within the John Cornett 500-acre patent under which appellees claim, or the William Kelly 200-acre patent under which appellants claim. As heretofore noted, while the Chancellor specifically stated “it would not be necessary” to determine this question, he did, in the manner of an afterthought, reach a conclusion on this point. It is appellees’ contention that in view of this finding the Chancellor’s judgment must be upheld unless the finding was clearly erroneous.
It is fundamental that the burden rested upon the appellees (as plaintiffs below) to affirmatively establish their own title; they could not rely upon the weakness of appellants’ title. Coleman Mining Company v. McClanahan, Ky., 237 S.W.2d 543; 18 Ky.Digest, Trespass, ®^>44. In Bryant v. Perry, 284 Ky. 698, 145 S.W.2d 1055, 1057, it is said:
“The burden was on the plaintiffs to locate the boundaries and to show that the land in dispute was embraced within the lines claimed by them. White v. Philpot, supra; Potter v. Tolliver, 219 Ky. 522, 293 S.W. 1074; Kentweva Coal & Lumber Company v. Helton, 170 Ky. 211, 185 S.W. 838. In Combs v. Valentine, 144 Ky. 184, 137 S.W. 1080, the court quoted with approval the following from Pearson v. Baker, 4 Dana 321, 323:
‘That construction is to prevail, which is most against the party claiming under an uncertain survey. It is his duty to show, and establish his corners. Preston’s Heirs v. Bowmar, 2 Bibb 493. From which it will follow that he who sets up and relies on an outstanding claim, must show that it embraces the land in contest, and should not succeed by using it, when it is uncertain whether it embraces it or not. * * *
Appellees had two expert witnesses. Much of their testimony was devoted to locating the beginning point of the William Kelly patent “in the head of the Deadening Hollow”. The inherent weakness of their testimony is that they made their survey and constructed their maps by taking as beginning points locations which were pointed out to them by one of the appellees, Martha Kelly. She, of course, was a very interested party. As one of these expert witnesses testified, by taking a given begin*257ning point, he “could put that William Kelly patent most anywhere”. This is obvious.
It is apparent that appellees’ experts did not independently attempt to fit their location of the William Kelly patent boundary into the panorama of other patent and deed locations. They seem to have placed it in an area owned by one William Young, whose property lines had not been disputed. There are other indications that their location of this patent boundary was more or less arbitrary and created many discrepancies.
Appellees contend that the begining point of the William Kelly survey in “Deadening Hollow” was adequately established by other witnesses. The dispute seems to be whether this point is the first or second hollow east of “Enoch Hollow”, the location of which is not in dispute. If Deadening Hollow is the same as “Road Fork”, it is the first hollow east of “Enoch Hollow” and appellants should prevail. If it is the second hollow east of “Enoch Hollow”, ap-pellees should prevail.
Appellees did have witnesses who testified that Deadening Hollow was not the same as Road Fork Hollow, although it is not clear that these witnesses knew where Road Fork Hollow was. The credibility of this testimony is seriously shaken by other evidence. John H. Lewis, an expert witness introduced by appellants, had surveyed this area in 1931. Whether Deadening Hollow and Road Fork were the same or not, his independent survey, made many years before, showed the beginning point where appellants claim it to be (i. e., the first hollow east of Enoch Hollow). Perhaps of more significance is a specific finding by a federal judge in 1932 in the case of Kycoga Land Co. v. Kentucky River Coal Corporation et al. (not reported), affirmed 110 F.2d 894. In the opinion the court expressly found: “Road Fork is also known as Deadening Hollow”. In that opinion Judge Cochran also located the patents here involved substantially as claimed by the present appellants.
It seems to us that the independent survey of Lewis and the opinion of Judge Cochran, made and rendered over 35 years ago when the monuments would have been more readily recognizable and the witnesses’ memories would have been fresher, substantially impaired the testimony of ap-pellees’ witnesses.
In addition, appellants’ principal expert witness, Howard Lusk, made an independent survey of this area. He started with deeds and maps and examined the opinion in the Kycoga lawsuit. From these sources, rather than from some arbitrary location pointed out to him by an interested party, he completed a map which supports appellants’ position. He located monuments on the ground which fitted with deed and patent descriptions. His boundary locations are consistent with adjoining parcels and do not show the discrepancies apparent in the testimony of appellees’ experts.
After careful consideration of all the evidence, it is our opinion that appellees failed to carry the burden of establishing the boundaries of their land so as to include the property in dispute. If the Chancellor’s findings of fact may be construed to require such a determination, they were clearly erroneous.
What we have said disposes of the cross-appeal.
The judgment is reversed on the original appeal, with directions to enter a new judgment dismissing the claim of the ap-pellees; the judgment is affirmed on the cross-appeal.
WILLIAMS, C. J., and STEINFELD, MONTGOMERY, MILLIKEN, and PAL-MORE, JJ., concur. HILL, J., not sitting.