Hi and welcome. That certainly forms a pretty pattern if nothing else
On a more serious note, I would guess that it is caused by a combination of some form of spectral shaping performed by the service provider and the bit-load algorithm they use. Are you a fair distance from the cab? Is your cab far from the exchange?
Most of any spectral shaping occurs in tones up to 512 to protect adsl/2/2+. Upstream tones U1 usually centered at around tone 1050 are usually heavily shaped to give protection to long lines from the cab. D2 can also experience shaping. Tones in U2 and any above tone 2000 are less likely to be shaped.
There's many different masks used for spectral shaping, the mask allocated to your line will depend on certain factors such as how close you are to the exchange, how close you are to the cabinet, your atten/power, noise from neighbouring lines, performance of neighbouring lines. Modern spectral shaping is dynamic and can change based upon line factors at each sync.
Im therefore not at all surprised to see lines shaped below tone 1984. Hlog shows a nice smooth transition across the spectrum of your usable tones.
If I understand you correctly, you are specifically enquiring about the visible wave pattern across all the useable tones. Due to the fact that the waves are very regular and smooth is what at first steered me towards method of bitload. In the UK, Openreach uses the waterfill method for bitloading, this causes a more blocky image similar to that on top of a castle, yours are more rounded but extremely regular. It's possible your SP is using one of the different algorithms for bit load.
However, there is something else that cannot be completely ruled out because I have very limited facts about your line and if it is under performing. The hlog & qln graphs can show line impairment such as a bridge tap or cross talk respectively. Using the distance between valleys in the wave its possible to calculate things such as tap length. There are special engineer tools that can analyse the frequency of the first dip and width of repeating wave then use a formula to pin point the tap length. I've never before seen such a narrow and shallow wave. If you look closely on most line's QLN, then the SNR is constantly showing small up and down variances of a few dB from the neighbouring tone. If it is some sort of line impairment, then an engineer with a JDSU would immediately be able to identify.
