Forest dynamics where typhoon winds blow

Abstract

Introduction

Tropical cyclones (TCs), also known as typhoons or hurricanes, are rotating storm systems that bring strong winds and heavy rainfall, often causing substantial damage to natural ecosystems. Even TCs graded 1–2 on the five-point Saffir–Simpson scale bring sustained wind speeds > 125 km h⁻¹, leading to defoliation, branch breakage, bole snapping, and uprooting of forest trees (Tanner et al., 1991; Everham & Brokaw, 1996; Negrón-Juárez et al., 2014; Lin et al., 2020). TCs cause substantial loss of aboveground forest biomass (AGB), with West Mexican and Puerto Rican forests reportedly losing 34% (Parker et al., 2018) and 23% (Hall et al., 2020) of ABG after category 3–4 TC events, respectively. TCs change forest structure, not only by damaging trees but also by remodelling tree architecture amongst survivors (Bonnesoeur et al., 2016; Ankori-Karlinsky et al., 2024). Regions that frequently experience strong TCs have shorter forests with higher stem densities (De Gouvenain & Silander, 2003; Ibanez et al., 2019; Lin et al., 2020), with trees investing into larger basal areas relative to their heights (Ibanez et al., 2019). Under climate change, TCs are becoming less frequent but more intense (Kossin et al., 2020; Chand et al., 2022) and are shifting towards higher latitudes (Murakami et al., 2020; Chand et al., 2022). To predict how these changes might affect forests in the future, it is critical that we have a comprehensive understanding of wind-forest dynamics at various spatiotemporal scales (Ennos, 1997; Lin et al., 2020).

We currently have limited knowledge on how wind, topography, and forest structure affect forest resistance to TCs at a landscape scale. Previous studies have shown that canopy height, soil type, stock density, and management action (e.g. thinning) could all affect forest resistance to strong winds (Cremer et al., 1982; Martin & Ogden, 2006; Gardiner, 2021). However, most of these studies were carried out in coniferous monocultures on flat terrain. We now know that the most valuable forests from biodiversity, carbon, and ecosystem services stand points are those with complex canopy structures (Bohn & Huth, 2016; Jucker et al., 2018; Zhu et al., 2023). Much of these forests also grow on rugged landscapes, where sites a mere few hundred meters apart could have vastly different wind regimes (Finnigan et al., 2020). Only a handful of studies have investigated the factors affecting TC-resistance in these more complex systems (Boucher, 1990; Tanner et al., 1991; Martin & Ogden, 2006; Lin et al., 2020; Ni et al., 2021). Most of these studies are based on field observations with small sample sizes and none have explicitly modelled wind (either long term or during TCs) across the landscape. Thus, the relationship between site-level exposure to wind and the patterns of damage remains poorly resolved. We also have very little understanding of how wind damage during TCs shape forest structure over longer time scales. At a regional level, Gorgens et al. (2021) found that wind affects the distribution of giant trees in the Amazon basin. Chi et al. (2015) suggested that typhoons reversed the elevation-tree height gradient in Taiwan by disproportionally impacting lowland vegetation. However, to our knowledge, no studies have explored whether long-term effects of TCs on forest height operate on finer spatial scales. In particular, it is unclear whether these wind-effects are more important than other environmental variables, such as wetness or aspect, in shaping local forest structures.

Monitoring forest damage after TCs is no trivial task. Many existing studies are based on field measurements in established forest inventory plots, which provide detailed measurements of tree damage and mortality but only over limited spatial scales (Tanner et al., 1991; Everham & Brokaw, 1996). A few recent studies have turned to analysing changes in satellite multispectral imagery, but changes in vegetation indices such as the normalised difference vegetation index or enhanced vegetation index primarily reflect defoliation and are only indirectly linked to structural damage (Rossi et al., 2013; Abbas et al., 2020; Hall et al., 2020; Xu et al., 2021). The development of repeated airborne laser scanning provides a solution to this. By generating point clouds from millions of returns, light detection and ranging (LiDAR) datasets can produce detailed maps of both canopy structure and background topography across large spatial scales. Comparing repeated LiDAR scans provides unparalleled information on forest structural responses against wind. The main constraint of LiDAR is that it is expensive to collect and we cannot predict the arrival of extreme TCs. Hence, datasets rarely capture forest conditions both before and after devastating TCs.

Similarly, measuring and modelling wind across a forested, mountainous site is notoriously difficult (Finnigan et al., 2020). Fundamental models of wind flow across flat terrain assume that wind speeds exhibit a logarithmic height profile, depending on the roughness of the surface (Wieringa, 1986), but these models fail to capture how wind interacts with complex terrain. Wind speeds increase significantly on windward slopes but are sheltered on leeward slopes (Lemelin et al., 1988; Miller & Davenport, 1998; Belcher et al., 2011; Finnigan et al., 2020). The position of the wind shadow cast by mountains depends on wind direction, while the size of the wind shadow depends on wind speed and the associated deflection of wind (Belcher et al., 2011; Finnigan et al., 2020). On steeper hills, separation bubbles could form on leeward slopes, which cause wind near the boundary layer to reverse direction (Kaimal & Finnigan, 1994; Belcher et al., 2011; Finnigan et al., 2020). In narrow valleys, the Venturi effect can speed up incoming wind (Mikkola et al., 2023). Modelling these effects is challenging, and anemometer measurements for training and validation are often unavailable (Shah et al., 2023). As a result, most studies on TCs avoid modelling local wind speeds and rather resort to proxies of wind exposure, such as rainfall, aspect, elevation, or topographical exposure (TOPEX) (Wilson, 1984; Albrecht et al., 2019; Morimoto et al., 2019; Araujo et al., 2021; Gardiner, 2021). To our knowledge, no study has combined wind modelling with repeated LiDAR surveys to assess forest resistance to strong winds.

New datasets available for the mountainous countryside of Hong Kong provide a unique opportunity wind modelling and TC damage assessment. In September of 2018, subtropical rainforests on the rugged landscape were hit by Typhoon Mangkhut. The typhoon was the strongest TC to affect Hong Kong in over four decades, bringing 10-min average wind speeds of >190 km h⁻¹ in exposed areas (category 3 on the Saffir–Simpson scale) (Hong Kong Observatory, 2023). Remarkably, the whole area was surveyed by airborne LiDAR scans in 2010, 2017, and 2020. These LiDAR scans captured structural changes of forests through time and provide a rare opportunity to study both pretyphoon growth and post-typhoon damage across large areas. Furthermore, hourly wind data are available from 28 non-urban automatic weather stations scattered across the rugged terrain (Hong Kong Observatory, 2023). This allowed us to properly validate wind maps generated by computational fluid dynamics (CFD) modelling software, which estimates near-surface wind speeds from a given digital surface model. In this study, we utilised the rare availability of repeated LiDAR and wind data to advance our understanding of how TCs affect forests on rugged terrains. In particular, we addressed five research questions:

1. How important was wind compared to other environmental variables in limiting local forest height?
2. Does the long-term effect of strong TCs impose limits on local forest height?
3. How do forest resistance to strong TCs affect forest rugosity and structure?
4. How did the interactions between forest height, local wind regime, and background topography affect forest resistance to extreme TCs?
5. Were natural forests more resistant to extreme TCs than plantations?