Sounds like you need a high-pass filter. Specifying the cut-off as a percentage and normalising the values should allow for variable s/n ratios.

#! perl -slw
use strict;
use List::Util qw[ max ];

sub islands {
    my( $percent, $dataRef ) = @_;
    my $max = max @$dataRef;
    my $filtered = join '', map{
        ( $_ / $max ) > $percent ? 1 : 0
    } @$dataRef;
    printf "%4.2f : %s\n", $percent, $filtered;
}

my @data = split ' ', do{ local $/; <DATA> };
islands $_, \@data for map{ $_ / 20 } 1 .. 20;

=comment
c:\test>545901
0.05 : 0000000111111111100001
0.10 : 0000000011111111000000
0.15 : 0000000001111111000000
0.20 : 0000000001111110000000
0.25 : 0000000000111110000000
0.30 : 0000000000111110000000
0.35 : 0000000000111110000000
0.40 : 0000000000111100000000
0.45 : 0000000000111100000000
0.50 : 0000000000111100000000
0.55 : 0000000000111100000000
0.60 : 0000000000111100000000
0.65 : 0000000000011000000000
0.70 : 0000000000011000000000
0.75 : 0000000000011000000000
0.80 : 0000000000011000000000
0.85 : 0000000000011000000000
0.90 : 0000000000011000000000
0.95 : 0000000000010000000000
1.00 : 0000000000000000000000
=cut

__DATA__
101 203 321 45 67 156 203 502 899 2003 5007 8020
7301 5030 3045 1243 567 321 234 45 123 453
[download]

You could enhance that by adding a minimum width to exclude transient spikes like the one shown at the end of the 5% line. Turning the 0s & 1s back into either groups of values or index ranges is fairly easy depending upon what you need to do with the islands once isolated.