Can I simulate the effects of changing patterns of land use on hydrographs using simple unit hydrograph models ?

Dependently from what you are really interested you should consider different type of models. If you are looking to investigate influence of land-use changes on hydrological responses for producing a sound scientific paper you will never publish a paper on a major journal using UH and CN curves.

If you want a quick and dirty solution for an engineering/management problem,  you can certainly gain some rough idea with a modeling like this (UH+CN is something better than the complete ignorance) but is mandatory to move to a semi-distributed model, i.e. using a (Geomorphologic Unit Hydrograph) GIUH  instead a UH. In theory a GIUH model contains the possibility to include land use for sub-basins, and therefore to perform the investigations you require. Land use, in all of these models will impact infiltration (according to some rule), but not evapotranspiration (unless you add a module for it), and the CN part will take care of it.

These lumped models usually work fine (for rainfall-runoff prediction) after a calibration of the parameters. So, what they guarantee is that you can eventually reproduce  with some degree of reliability the same variable which you calibrated (discharge).  In other words, the parameters of these models are always "ex-post" and never "ex-ante".

The only rational way to obtain information about land use from this procedure and type of models is therefore to perform calibrations before and after a land use change happened. If you are bold and brave, you can try to extends the results to other basins.  Therefore GIUH+CN can work where you have records of discharges, and, at the same time, a record of land-use changes   for so long time in a basin that some land use change happened, and you registered it. 

I guess this is not your case.

However, I have some concern about CNs. Many distinguished hydrologists suggest to use CN for describing the role of land use. But this is largely "a conspiracy": since engineers (agro-meteorologists, and hydrologists) "must do it", they have mandatorily to come out with some number, many agreed that using CNs could have been e way to get it. Actually, as I already told, when CNs works is because the curve numbers were calibrated before (and not because the tables in literature magically work), and possibly they can reproduce good results, just for those basins and situations where they were calibrated.  Finally, CNs cover just infiltration aspects and do not even recognise the existence of evapotraspiration which is clearly a main issue when talking about land-use. Certainly you could try, at your risk, a type of GIUH+CN+ET modelling.

People that have really to concern with land use/land cover, i.e. those working in agro-meteorology, in fact directed themselves to use more complex models, which recognise a very detailed partition of the landscape, and a variety of processes. Among these models, those that came to may attention are MHYDAS by Morel-Seytou/Roger Moussa and AGEs, by ARS/Jim Ascough.  Also our JGrass-NewAGE can possibly be used  to try to give some answer.
The latter are possibly the most simple models that can be used successfully to try to assess the role of land-use, by parameterising - a priori - its influence. I remark however, that this would require a painstaking verification of the modelling of sub-processes, that was seldom pursued in literature (hydrologists like shortcuts). 

The best choice, would be, however, to use process-based models like our GEOtop, or Paniconi-Putti CATHY, or Garrote-Ivanov-Vivoni-Bras tRibs, or Wigmosta-Lettenmaier's DHSVM, or Zehe's CATFLOW where the right processes are explicitly modeled. 

You can also find useful this old post, always in this same blog.