The NPs of all Germanic languages are head-initial. Headedness is determined on the basis of the distribution of complements, not modifiers; the distribution of modifiers only correlates with the distribution of complements.
(1) destroy the city
(2) the destruction of the city
The complement obligatorily follows the head. The contrast can be seen more nicely in German, where VP is head-final (3) while NP is head-initial (4), such that the position of the complement changes on behalf of the category of the head.
(3) die Stadt zerstören
the city destroy
(4) die Zerstörung der Stadt
the destruction of-the city
For further discussion I'd suggest looking into Haider (2004) on the matter. He's far from uncontroversial in general, but pages 782ff. illustrate a nice further contrast between German NPs and VPs which was previously discussed as the 'head-final filter': ungrammatical: die schneller(e) als Bolt Frau ('the faster than Bolt woman') vs grammatical: schneller als Bolt rennen ('faster than Bolt run'). These examples the English translations are supposed to show that the head-initial phrases do not allow, that the head of a modifying phrase is non-adjacent to the modified head. That is, it is not a general property of German to allow for non-head-adjacent modifying phrases, but it a property of its head-final phrases (VP and AP) only. This is just to illustrate one further putative difference between these phrases. Here's the doi-link, so it should be stable:
Apart from Haider being controversial, I'm not aware of any researcher who claims that the German NP is head-final. Also note that the head-initiality of the majority of phrases among the languages of Europe was surely one of the main reasons for Kayne (1994) to claim that every phrase is basically head-initial.