The two strands of double- stranded DNA are held together by a number of weak interactions such as hydrogen bonds, stacking interactions, and hydrophobic effects [The Three- Dimensional Structure of DNA].
Of these, the stacking interactions between base pairs are the most significant. The strength of base stacking interactions depends on the bases. It is strongest for stacks of G/C base pairs and weakest for stacks of A/T base pairs and that's why it's easier to melt A/T rich DNA at high temperature. (It is often incorrectly assumed that this is due to having only two hydrogen bonds between A/T base pairs and three between G/C base pairs.)
The figure below shows a melting curve of various DNAs. The curve shows the conversion of double-stranded DNA to denatured single strands by following the change in absorbance as the temperature is increased from left to right. When the double helix is unzipped the absorbance increases. Note that poly(AT) "melts" at a lower temperature (TM = melting temperature) than poly(GC). This is because the average stacking interactions of G/C base pairs are two or three times stronger than A/T base pairs so more thermal energy is need to disrupt them.
The base stacking interactions have been measured in several different ways but most of these measurements are indirect and all of them have been with double-stranded DNA. Of the single-stranded polynucleotides, only polyA has a helical structure in solution and that's because of the stacking interactions between single adenylate resides in the polynucleotide. PolyT is somewhat unstructured and polyG and polyC have complex three-dimensional structures that are difficult to interpret.
Assuming that the stacking interactions of the adenylate residues is the only significant force maintaining the polyA helix, it's possible to measure the stacking interaction directly by pulling both ends to see how much pressure it takes to disrupt the helix. This can be done by fixing single-stranded polyA to a substrate and grabbing the other end with a molecular probe. The elasticity of the DNA can be measured by single-molecule atomic-force spectroscopy (Ke et al. 2007).
As the molecule is stretched, it resists up to the point were the bases become unstacked and the helix is disrupted. The force required can be used to directly calculate the stacking interactions between the adenylate residues. The value turns out to be 3.6 ± 0.2 kcal/mol per base (15 kJ/mol). This is very close to the stacking energies calculated for A/T base pairs in earlier experiments. (The stacking energies for G/C base pairs in DNA are about 61 kJ/mol.)
The experiment is independent, and direct, confirmation of the literature values for stacking interactions. The energies of these stacking interactions turn out to be significantly larger than the energies of the other weak interactions involved in holding double-stranded DNA together (hydrogen bonds, "normal" van der Waals interactions, and hydrophobic interactions).
Changhong Ke, Michael Humeniuk, Hanna S-Gracz, and Piotr E. Marszalek (2007) Direct Measurements of Base Stacking Interactions in DNA by Single-Molecule Atomic-Force Spectroscopy. Phys. Rev. Lett. 99:018302[The top figure is from Ke et al., 2007]