*2.1. Study Area*

Qeshm Island is located a few kilometers o ff the southern coast of Iran, opposite the port cities of Bandar Abbas and Bandar Khamir. It is the largest island in the Persian Gulf and covers an area of 1491 km<sup>2</sup> (Figure 1). Most of the mangrove forests of Qeshm are located in the northern part of the island in the Hara Protected Area, a biosphere reserve that covers an area of approximately 20 by 20 km and is characterized by numerous tidal channels [42]. The mangroves are rooted in the saltwater of the Persian Gulf, but the special pores within their leaves extract the salt from the water. The whole forest area is a ffected by frequent boat trips, fishing and a small amount of leaf-cutting for livestock feed. The forests are the habitat for migratory birds, hooked turtles and venomous aquatic snakes, all of which are indigenous species.

**Figure 1.** *Cont.*

**Figure 1.** (**A**) Left: location of Qeshm Island and the mangrove ecosystem, shown as a false-color Sentinel-2B image (2017, Combination of Bands 8-4-3); right: Worldview-2 image data used for the upscaling approach. (**B**) Aerial roots (pneumatophores) growing in a wide radius around the mangrove (*Avicenna marina*) are highlighted by the red polygon.

## *2.2. Field Data*

The field survey revealed that *Avicenna marina* was the dominant mangrove species on Qeshm Island. Visual analysis of high-resolution images made it possible to distinguish between eight target classes of mangrove ecosystem, including three types of mangrove spatial pattern: closed canopy mangrove, open canopy mangrove, and individual mangrove trees (found in a small patch on the island). The remaining target classes in the study area were mudflat (either with or without aerial roots), tidal zone (sand, beaches or unvegetated area), shallow water (rivers or ponds), and deep (open) water.

During the field survey, a total of 170 GPS reference points (Garmin 629sc with spatial accuracy between 1 and 5 m) were collected and used for validation of the classification of the eight land cover classes. In order to minimize and avoid the negative impacts on the vulnerable ecosystem, the collection of field samples was restricted to easily accessible parts. In order to increase the number of samples for three types of mangrove and two types of mudflat, 53 points were additionally selected from Spot 6/7 data using image interpretation. Figure 2 shows the distribution of the samples for the eight land cover classes. The set of reference points collected from both GPS and from the Spot images are depicted for each class separately in Table 1.


**Table 1.** Overview of the two different sets of reference points collected from the GPS survey and the Spot 6/7 image interpretation.

**Figure 2.** Distribution of the samples obtained from the field survey and from Spot 6/7 image interpretation of the whole study area.

#### *2.3. Remote Sensing Data and Pre-Processing*

Technical specifications of the Worldview-2 and Sentinel-2 imagery are given in Table 2. Images were cloud-free over coastal areas. The multispectral bands of Worldview-2 consist of four standard bands (red, green, blue and near-infrared 1) and four additional bands (coastal, yellow, red edge and near-infrared 2), which facilitated spatial and spectral analysis, mapping and monitoring of large areas at a more detailed level [43]. Sentinel-2 bands consist of four bands at a 10-m spatial resolution (blue, green, red and near-infrared), six bands at a 20-m spatial resolution (four narrow bands near the red edge and two wider SWIR), and three bands at a 60-m spatial resolution (aerosols, water vapor and cirrus) [44]. The obtained data were pre-georeferenced to the UTM zone 40 North projection using the WGS-84 datum. Sentinel-2 data were radiometrically calibrated to apparent surface reflectance by the FLAASH (Fast Line-of-sight Atmospheric Analysis of Hypercubes) atmospheric corrected algorithm [45] in ENVI 5.4 software. Fusion of panchromatic with multispectral images of Worldview-2 data resulted in an image with a 0.5-m spatial resolution. In the present study, the Gram Schmidt pan-sharpening algorithm was applied [46] because it preserves the primary spectral value of the objects and has successfully been applied to multispectral images. In this study, a Sentinel-2 level 1C product image was applied, acquired on a clear day and under the lowest tide condition over Qeshm Island.
