Permafrost distribution in the Qinghai-Tibet Engineering Corridor (QTEC) is of growing interest due to the increase in infrastructure development in this remote area. Empirical models of mountain permafrost distribution have been established based on field sampled data, as a tool for regional-scale assessments of its distribution. This kind of model approach has never been applied for a large portion of this engineering corridor. In the present study, this methodology is applied to map

more »

... t distribution throughout the QTEC. After spatial modelling of the mean annual air temperature distribution from MODIS-LST and DEM, using high-resolution satellite image to interpret land surface type, a permafrost probability index was obtained. The evaluation results indicate that the model has an acceptable performance. Conditions highly favorable to permafrost presence (≥70%) are predicted for 60.3% of the study area, declaring a discontinuous permafrost distribution in the QTEC. This map is useful for the infrastructure development along the QTEC. In the future, local ground-truth observations will be required to confirm permafrost presence in favorable areas and to monitor permafrost evolution under the influence of climate change. on the QTP. Therefore, improved methods for mapping permafrost distribution are essential to designing road and pipelines and to understanding the dynamics of alpine ecosystems. The construction of a new expressway from Golmud to Lhasa has been proposed along the QTEC. Since the permafrost condition at a site would affect the engineering design and cost of road construction, a detailed and present-day knowledge of permafrost distribution and its relationships with geomorphology were essential in the corridor. Present available permafrost maps were mainly at low resolutions, which varied from scales of 1: 600,000 to 1: 10,000,000, although many permafrost maps have been compiled since the early 1960s [17] . At more local scales, factors that affect local microclimate and surface energy balance e.g., slope, aspect, local hydrology, vegetation cover, geology, and snow cover strongly influenced the permafrost features [18, 19] . A good understanding of these effects on permafrost occurrence was significant, as it may provide some hints on the techniques and measures we can use to artificially simulate similar effects [20] [21] [22] . Thus, a detailed and up-to-date permafrost map of the QTEC was required for mitigating potential engineering problems associated with permafrost-affected terrain. It will also help to identify areas for further investigation so that permafrost areas can be avoided or, if necessary, engineering solutions can be designed to maintain the physical and thermal state of permafrost. Many models already exist for estimating the spatial distribution of permafrost in regions of the European Alps [23] [24] [25] , and the Arctic [26, 27] . These models may be of the equilibrium, empirical-statistical, or process-based types and have been widely used at regional and local scales [3, 28] . Permafrost mapping based on geophysical techniques or process-based types was expensive, time consuming, and spatially restrictive due to the difficult detection and monitoring of permafrost on the QTP, although these techniques can provide a detailed and robust transient thermal state of permafrost. Furthermore, the lack of sufficient and reliable data for calibration and validation probably was one of the most important limitations for permafrost modeling [25] . Empirical-statistical models describing the distribution of mountain permafrost based on geomorphological permafrost indicators and topographic and climatic predictors were a simple yet effective approach toward a first assessment of its distribution at a regional scale [29] [30] [31] [32] . Furthermore, remote sensing as a permafrost monitoring tool was under continuous development with fine spatial and temporal resolution data from satellites such as Landsat-8, SPOT-5, and GF-1 and 2 (Gaofen-1, 2 satellite). This technique has the potential to provide a valuable and cost-effective mean for mapping and monitoring near-surface permafrost conditions, as well as seasonally frozen ground [33] . High-resolution image and elevation data acquired by satellite can be used to interpret the geomorphological features such as slope failure, thermokarst [16, 34] , and biophysical features such as vegetation, topography, and surface hydrology [26, 35] . A number of studies have demonstrated the usefulness of this approach for permafrost mapping [29, [36] [37] [38] [39] . Therefore, the target of this study was to map the potential permafrost distribution in the QTEC (91 • E-95 • E, 32 • N-36 • N) based on a logistic regression model linking the permafrost existence probability to surface variables that included vegetation type and climatic conditions at a high resolution of 30 m. For this purpose, we investigated the surface characteristics (e.g., vegetation, soil) and permafrost conditions along the QTEC based on the permafrost survey position and boreholes that were carried out when the QTR and Qinghai-Tibet highway (QTH) were built. Considering the potential incoming solar radiation, vegetation type, and the mean annual air temperature as potential predictors, they were compiled and developed from remote sensing data. Study Area The study area was located in the central QTP and encompassed a 550 km long and 40 km wide section (22,351 km 2 ) of the QTP (Figure 1a,b) , extending from Xidatan to Anduo, which were the northern and southern boundaries of permafrost occurrence (Figure 1b) , respectively, bounded by latitude 32 • -36 • N and longitude 91 • -95 • E (Figure 1c ). This corridor is likely to become a locus of many developmental projects because constructions of a gas pipeline, QTP, QTR, and electric Remote Sens. 2018, 10, 215 3 of 18 transmission line were along this route. Most parts of the terrain were located above 4500 m a.s.l., with alternating distribution of mountains, valleys, and basins. A recent field investigation and literature indicated that near surface deposits were dominated by coarse materials such as gravel and sandy soils [10, 40] . These specific geomorphologic and sedimentary patterns resulted in significant differentiations in permafrost features [10] . Climatically, it was located in an extremely continental climate zone, favoring clear skies and high solar radiation. The mean annual air temperatures (MAAT) were commonly between −6.5 • C and −2.0 • C, with the annual total precipitation ranging from 250 mm to 450 mm, which occurred mostly as rainfall between May and August [6, 8, 40] . The majority of the plateau had a free snow cover in winter [40] . Vegetation type was characterized as alpine meadow and steppe with the coverage ranging from 0.3 to 0.9. The detailed information of the vegetation type was discussed in Section 3.3. Remote Sens. 2018, 10, x FOR PEER REVIEW 3 of 19 literature indicated that near surface deposits were dominated by coarse materials such as gravel and sandy soils [10, 40] . These specific geomorphologic and sedimentary patterns resulted in significant differentiations in permafrost features [10] . Climatically, it was located in an extremely continental climate zone, favoring clear skies and high solar radiation. The mean annual air temperatures (MAAT) were commonly between −6.5 °C and −2.0 °C, with the annual total precipitation ranging from 250 mm to 450 mm, which occurred mostly as rainfall between May and August [6, 8, 40] . The majority of the plateau had a free snow cover in winter [40] . Vegetation type was characterized as alpine meadow and steppe with the coverage ranging from 0.3 to 0.9. The detailed information of the vegetation type was discussed in Section 3.3.