The question piqued my interest, but after hunting through the literature for a bit, I hadn't found any direct answers. Then I went back and read the mouse study you cited a bit more carefully. The mouse study only made a reference to mice being affected at 4 lux, ~100x more sensitive than humans. However, for that number it cited a paper in Science that has a direct answer.
Melatonin concentrations decreased 10 to 20 minutes after the subjects were exposed to 2500-lux incandescent light and reached near-daytime levels within 1 hour (Fig. 1). After the subjects resumed sleeping in the dark, the melatonin concentrations increased immediately and within 40 minutes were at the levels measured before exposure. The fluorescent light (500 lux) did not reduce melatonin, and there was no change after the return to darkness. In the two subjects who were exposed to 1500-lux incandescent light, melatonin concentrations decreased to levels intermediate between those measured during exposure to 500 and 2500 lux (Fig. 2). The return to normal nighttime concentrations after subjects were exposed to 1500 lux was similar to that occurring after their exposure to 2500 lux. The concentration of melatonin in subjects awakened and exposed to 500-lux fluorescent light did not differ significantly from that measured while they were asleep in the dark.
Since that was the granddaddy study of the subject, I just checked the recent citation list on that page for modern articles.
Using pure blue LED lights at 446-477 nm wavelengths, West et al. measured light intensities necessary to induce melatonin suppression. Table 1 of that study converts LED irradiance/lux into retinal irradiance (uW/cm2) based on mean pupil size. Figure 2 shows plasma melatonin falling (p<0.05) at 20 uW/cm2 of corneal irradiance (then curve fit to 14.19 uW/cm2). It's at least twice as powerful as white light, which didn't show significant suppression at 40 uW/cm2, but was "numerically similar" to 10 uW/cm2 exposure. If I've got the conversions right, 20 uW/cm2 is about 136 lux, which is about the brightness of an overcast day or half the brightness of typical office lighting.
As for the discrepancy between the two studies (1500 lux vs. 136 lux), I would blame it mostly on technological advances since 1980. The original study used gas chromatography. You can see the huge error bars and noisy data in the Figure. The modern study uses a radioimmuno assay using antiserum, and is presumably far more sensitive.
I was also going to mention a nice review paper that summarizes more findings, but apparently I can only post 2 links as a new user. So I'll just paste the abstract and citation.
Light is a potent stimulus for regulating the pineal gland's production of melatonin and the broader circadian system in humans. It initially was thought that only very bright photic stimuli (≥ 2500 lux) could suppress nocturnal melatonin secretion and induce other circadian responses. It is now known that markedly lower illuminances (≤ 200 lux) can acutely suppress melatonin or entrain and phase shift melatonin rhythms when exposure conditions are optimized. The elements for physical/biological stimulus processing that regulate photic influences on melatonin secretion include the physics of the light source, gaze behavior relative to the light source, and the transduction of light energy through the pupil and ocular media. Elements for sensory/neural signal processing become involved as photons are absorbed by retinal photopigments and neural signals are generated in the retinohypothalamic tract. Aspects of this physiology include the ability of the circadian system to integrate photic stimuli spatially and temporally as well as the wavelength sensitivity of the operative photoreceptors. Acute, light-induced suppression of melatonin is proving to be a powerful tool for clarifying how these elements of ocular and neural physiology influence the interaction between light and the secretion of melatonin from the human pineal gland.
Photic Regulation of Melatonin in Humans: Ocular and Neural Signal Transduction
Brainard, Rollag, and Hanifin
J Biol Rhythms December 1997 vol. 12 no. 6 537-546
If you need more information, I'd start by checking the papers cited by the second paper. Alternately, there's probably more historical information available by looking at papers which have cited the first one.