**1. Introduction**

Wildfire has been well recognized as a crucial process governing the dynamics of forest structure, composition and function in boreal forests [1,2]. Severe burns can cause abrupt changes to forest ecosystem by killing living plants, consuming organic matter, and altering biophysical environments [3–5]. In general, forest ecosystems are resilient and able to recover key functional and compositional attributes after disturbances [6–8]. However, the resilience of boreal forests is decreasing due to a misalignment between changing fire regimes and vulnerable ecosystem states [9–11]. Biome transitions in boreal forests, including shifts from forests to treeless steppes and an increase in deciduous trees, are thought to be exacerbated by changing fire regimes [12–15]. In addition, both fire frequency and severity in boreal ecosystems is predicted to increase in the future [16–19]. Research on the spatial controls of forest restoration post-fire is essential for understanding how boreal ecosystems respond to changing fire regimes and advancing ecologically sound policies in boreal fire managemen<sup>t</sup> and restoration.

Tree sapling regeneration is strongly correlated to the future successional trajectory of burned forests and thus can be used to anticipate future compositional and structural dynamics [20,21]. Structural attributes of regeneration observed in the field, such as abundance, biomass, and species composition, can be used to determine the strength of post-fire forest recovery [21–24]. Field-based inventory approaches can provide first-hand information, but this is a labor-intensive and time-consuming process with limited utility for long-term monitoring over a broad spatial scale. On the other hand, remote sensing is a novel and cost-effective technique for monitoring forest ecosystems [25], as well as for evaluating fire-related characteristics such as burn severity [26,27], burned area [28], and monitoring post-fire vegetation recovery [29–31].

Satellite imagery for monitoring post-fire vegetation dynamics assumes that the remotely sensed surface reflectance can capture spectral signal variations that correspond to vegetation recovery. Vegetation indices were designed to reflect vegetation photosynthesis and were widely accepted as surrogates of canopy attributes, biomass, or tree coverage across global forest ecosystems [30,32]. However, spectral variation as exhibited in satellite imagery is not a biophysical parameter that can precisely characterize structure or function of forest ecosystems [33]. Forest recovery is typically described ambiguously in terms of changes in vegetation index or greenness in remote sensing literatures, however little attention has been paid to investigate whether such remotely sensed spectral information can reflect key forest structures in newly established boreal forest stands [34,35]. On the other hand, structural attributes rather than spectral variations are the critical indicators of forest ecosystem functions (i.e., carbon sequestration and water budgets) [36,37], and are thus more important for assessing forest recovery for managemen<sup>t</sup> and scientific perspectives.

Tree sapling abundance (TSA), the number of tree saplings per unit area, is a crucial attribute of the post-fire forest stands as it determines the future successional trajectory of forests [38,39]. Burned areas with high TSA have a higher possibility of developing as forests through succession and recovery processes, while burned areas with few tree saplings may take a longer time to approach canopy closure. The spatial distribution of TSA is thus a critical reference for directing artificial managemen<sup>t</sup> measures that aim to promote forest restoration post-fire. Leaf area index (LAI) is another structural attribute with critical importance to forest productivity and biomass accumulation through influencing forest canopy photosynthesis [40,41]. LAI can be measured rapidly in the field and is one of the few forest structures that is well-linked to optical remote sensing at various spatial scales [42,43].

Spatially monitoring these two structural attributes is essential for understanding the functional dynamics of post-fire forest ecology but presents a challenge for traditional remote sensing approaches. In post-fire landscapes, tree sapling distribution usually exhibits a high degree of heterogeneity as a result of interactions among seed availability (legacy effects), filter effects from environmental factors, inter- and intraspecific competition and edge effects [23,44]. Previous studies aimed at disentangling these effects were dependent on field investigations but very high-resolution (VHR) satellite imagery, which can provide fine details of post-fire landscapes (<1 m), is a promising approach that has not previously been utilized for tree saplings. The high resolution of VHR imagery is significant in allowing for accurate assessment of tree sapling crown size, but very few studies have investigated the utility of VHR imagery for monitoring forest structural attributes in newly reestablished stands.

Based on this, we sought to address the following questions: (1) How do the performances of VHR and Landsat imagery differed for delineating the spatial distribution of TSA and LAI in the early post-fire landscape; (2) which spatial controls (interspecific competition, legacy effects, environmental filters or edge effects) are most important in the recovery of LAI and TSA post-fire; and (3) how do these spatial control factors influence (inhibiting or facilitating) spatial distribution of LAI and TSA post-fire?

#### **2. Materials and Methods**
