**4. Discussion**

#### *4.1. The NTL Distribution Pattern in Di*ff*erent Types of PAs*

The primary objective of type Ia PAs is to conserve regionally, nationally, and globally outstanding ecosystems, species (occurrences or aggregations), and/or geodiversity features: these attributes will have been formed mostly or entirely by non-human forces [55]. However, the global NTL level inside type Ia PAs was not the lowest. The NTL value in type Ia PAs was 2.5 times higher than that of the type Ib PAs. Among seven types of PAs, the NTL growth rate inside type Ia was very high from 1992 to 2013, reaching 378%, which was far greater than the others. The NTL for 0–1 km and 1–5 km buffer zones had growth rates that were ranked highly, with values of 441% and 372%, respectively. The area with the highest NTL level around Ia consistently approached the boundary of the PAs. The brightest NTL areas between 1992 and 1999 was in the 25–50 km buffer zone, which shifted to the 5–10 km buffer zone between 2001 and 2003. According to the growth rate and trend of each buffer zone, the brightest NTL area would be encroaching in the 0–5 km area quickly. Human development in the 0–1 km outside PAs could have direct impacts [56,57]. Because the baselines of the nightlight level of different types of protected areas are very different, the actual change values corresponding to the same growth rate are very different even though the growth rates are the same. For example, the NTL growth rate of type Ia PAs exceeded the others, but the absolute increase of the NTL DN value inside and around the Ia PAs was still low in comparison to other types because its baseline of the nightlight level was far less than other types. It is noteworthy that, according to the guidelines, Ia will be degraded or destroyed when subjected to all but very limited human impact [55]. In addition, the area of category Ia was generally small, and human-induced impacts contributed more in small PAs than to stochastic processes [58].

For type Ib PAs, the average NTL DN value was 0.04, which was the lowest among all types. Also, the NTL levels in and out of the PAs boundaries was lower than type Ia PAs. The 50–100 km buffer zone was the area with the most concentrated human population around Ib PAs. The NTL distribution for type II PAs were similar to type VI and both were lower among the types of PAs. The NTL levels of types II and VI were not significantly higher than those of Ia and Ib.

The main function of type IV PAs is to protect certain species and habitats. The NTL levels inside type IV PAs and buffer zones were significantly higher than all other types except V. NTL levels within and outside type III PAs were high, second only to type V. NTL levels within and outside of type V PAs were much higher than other types.

The result suggested that the NTL levels were significantly lower within the PAs than that of the surrounding 0–100 km buffer zones. In particular, there was a big difference between inside and outside the boundary of types Ia and III PAs. For most PAs, the surrounding areas with the highest DN value were in the 1–25 km buffer zone.

#### *4.2. Skyglow as a Biodiversity Threat*

Skyglow occurs when artificial light is scattered by water, atmospheric molecules, or aerosols and returned to Earth. In this study, we used the NTL data to show the impact of urbanization on protected areas. However, we did not apply any light propagation models to integrate the skyglow effect due to indirect lights. Skyglow is of increasing concern since it is able to multiply and extend the light pollution effect and affect those areas with no direct light pollution. Because of skyglow, the biological impacts of artificial light are not limited to the vicinity of the light source and may spread over much larger extents via several mechanisms. Individual lights may be visible kilometers away from their source, and the addition of artificial skyglow can extinguish such lunar light cycles and permanently remove dark nights from a landscape [59,60]. Therefore, artificial lighting may have an effect on natural ecosystems even when the source of lighting lies kilometers away. Most organisms have evolved molecular circadian clocks controlled by natural day–night cycles. These clocks play key roles in metabolism, growth, and behavior [61]. As the world grows ever-more illuminated, many light-sensitive species will be lost, especially in or near highly illuminated urban areas [62]. Light pollution threatens biodiversity through changed night habits, including organismal movements [63,64]; foraging [65–68]; interspecific interactions [69]; communication [70,71]; reproduction [72] of insects, amphibians, fish, birds, bats, and other animals; and it can disrupt plants resulting in phenological changes by distorting their natural day–night cycle [63].

Therefore, measures are necessary to prevent or reduce the ecological impact of night-time light pollution. Maintaining and increasing natural unlit areas is likely to be the most effective option for reducing the ecological effects of lighting [73]. From the lighting differences inside and outside the PAs, it can be seen that in the process of human development, PAs have greatly reduced human interference, so PAs are undoubtedly the best place to maintain darkness. More stringent control measures should be implemented within and around PAs, such as limiting the duration of lighting, reducing the trespass of lighting, changing the intensity of lighting, and changing the spectrum of lighting [73,74].

Setting an area surrounding PAs is an ideal option to provide a buffer against the light pollution released by human development. However, it cannot be solved by using only remote sensing data. Different types of PAs are in different natural and socio-economic conditions, and as such, the appropriate buffer radius should also be different. The success of planners in reducing the ecological impacts of light pollution will ultimately depend on an assessment of the critical mechanisms and thresholds that determine those impacts in a particular environment [73]. One possible way is to combine remote sensing data with biodiversity data to explore how biodiversity in protected areas with similar natural conditions respond to different lighting patterns. This may be of grea<sup>t</sup> significance for solving light pollution near the PAs and may also contribute to determining how much of a buffer distance should be set around the PAs.

#### *4.3. Limitations of Night-Time Lighting Data*

The 1992–2013 NTL data used in this study were derived from six di fferent sensors. Therefore, it is di fficult to discriminate whether the small di fference between the light data of di fferent years was caused by the sensor or by changes in the field light. Moreover, we are not sure how much data error was caused by the sensor. In this study, we compared the nightlight levels between 1992 and 2013. The time span and the di fference in the NTL data were su fficiently large, such that the data error caused by the di fferent sensor could be ignored.

In addition, the light data could not detect negative light growth (i.e., reductions in light) in each PA. To reduce the error caused by the sensor when processing the light data, we assumed that all regions were developing positively, and the DN value was increasing; thus, we assigned the larger DN value of the previous year to the darker pixels corresponding to the following year. If the actual nightlight of a certain area was dark each year, the DN value would remain unchanged. On the global scale, the nightlight within and around each type of PA was increasing each year, and there was no negative growth in the brightness of the PA. However, when assessing individual PAs, there may be negative growth due to increasingly strict regulations and improved awareness of protection, e.g., lighting tools could be replaced by those with a lower brightness and lighting time could be reduced. Therefore, the lighting data may not be suitable for the study of PAs on a small spatial scale, especially in developed countries where the managemen<sup>t</sup> of PAs is more stringent.

**Author Contributions:** All authors contributed to this manuscript. Specific contributions include conceptualization, L.F., J.Z., and Y.W.; methodology, J.Z. and L.F.; software, L.F.; validation, L.F. and J.Z.; formal analysis, L.F.; data curation, L.F. and J.Z.; writing—original draft preparation, L.F. and J.Z.; writing—review and editing, L.F., J.Z., Y.W., Z.R., H.Z., and X.G.; visualization, J.Z. and L.F.; supervision, J.Z.; funding acquisition, J.Z.

**Funding:** This research was funded by the National Natural Science Foundation of China (grant nos. 41771450, 41871330, and 41630749), the Fundamental Research Funds for the Central Universities (grant no. 2412019BJ001), the Foundation of the Education Department of Jilin Province in the 13th Five-Year project (grant no. JJKH20190282KJ), and the Science and Technology Development Project of Jilin Province (grant no. 20190802024ZG).

**Conflicts of Interest:** The authors declare no conflict of interest.
