Since you've detailed the girls entirely as if they were on a perfectly flat and horizontal ground, it's probably too late to add uneven ground or grass.

In the earlier version, an issue was that the far trees had 'leaves' which were drawn as big as the closest detailed tree: the far tree was made of 5 "leaves" vertically, while the closer one was 14 leaves high.

Hi I'm yrizoud;
I'm a software developer (for industries, nothing very fun), I had an Amstrad CPC and an Amiga when I was a kid, and I'm still very fond of games and graphics from the 80s-90s. I've always programmed rudimentary games, doing code and graphics. I spent two years drawing huge sprites for my own fighting game on the Amiga, the project failed and the sprites were mediocre, but at least I improved. The experience helped me program a successful remake of the arcade cabinet game "Vigilante", using the original's graphics. I tend to have good theoric knowledge of pixel art, but I really lack imagination and originality, which makes me more useful for final pixelling than concept art.
At the moment my graphic-related activity is mostly to maintain the painting program Grafx2, which is probably the best I can contribute to the pixel world 

About the tile space, the screen you showed was 80x200, not 160x200. You would have 2x more space available than you think.
However, you should make sure if you have that much space for *each* level, or if it's common scenery for the whole game.
In the former case there will be disk loading on level change, in the latter case you'd better re-use some tiles from former levels.

Do you have the necessary tools for good tile work ? I made my own "map to tiles+map" converter, but even a standard tool like the Mappy tile map editor can read a map image and count the unique tiles. It's handy if you have a modification to perform on one tile and want to propagate everywhere the tile was used.

Layers are unique and independent, almost all tools and effects can only read and write in the current layer.
I can't view your screen captures from here, but:

> when i create a new layer, then grab it, it draws a black square that covers the object

It sounds like you used the right mouse button to grab. It replaces the grabbed part by pixels of your current BG color. If this BG color is the same number as your image's transparent color, it really erases (becomes transparent, you see other layers below). But if it's a different color, it draws solid pixels in its place. It can be confusing if the image's transparent color is "black", and you also paint in "black" with an other color index.

> but does not select it and also grabs the object from layer1.

The brush-grabbing sees only one layer, so if you don't capture the brush that you expected, it means you grabbed from the wrong layer, or you had drawn the shape in the wrong layer.
Note that when you right-click the button of the currently active layer, it hides all others (or shows them back). It's very handy to check the content of a layer.

Here's a test image, it can help experiment with the tools:
http://grafx2.googlecode.com/issues/attachment?aid=6958252391428804566&name=12345678.gif&token=8b83c394aa744636d6150b1e670e153f
(Attachment from page http://code.google.com/p/grafx2/issues/detail?id=7#c26  , use button "download" because button "view" fetches a flat PNG copy.)
The "1" is written in layer 1, "2" in layer 2 etc. All the rest is transparent.