Remote-Sensed Evidence of Fire Alleviating Forest Canopy Water Stress Under a Drying Climate

Abstract

Fire is a distinctive factor in forest ecosystems. While uncontrolled wildfires can cause significant damage, prescribed burning is widely used as a management tool. However, despite the growing threat of forest water stress under climate change, there is a lack of concrete evidence on the impact of fire on water stress in forest ecosystems. This study utilized Landsat time-series remote sensing data combined with the Infrared Canopy Dryness Index (ICDI) to monitor changes in canopy dryness patterns across the eucalyptus-dominated Northern Jarrah Forest of southwestern Australia. The forest was chosen due to its exposure to a changing climate characterized by decreasing rainfall and more frequent droughts, signs of water stress in otherwise drought-resilient trees, and its well-documented fire management history. Analysis of ICDI patterns over the period from 1988 to 2024 revealed a clear overall trend of increasing water stress, coinciding with a small overall decline in annual rainfall in the 10,000 km² study area. Furthermore, by examining five prescribed burns and five wildfires, we found that NDVI-assessed canopy cover recovered rapidly in fire-affected areas, typically within one to three years, depending on fire severity. However, ICDI water stress levels were reduced for approximately 7–8 years following low-severity prescribed burns and more than 20 years after high-severity wildfires. These findings suggest the potential of prescribed burning as a tool to mitigate water stress in vulnerable forest landscapes, particularly in regions prone to drought and climate change. Additionally, the study underscores the effectiveness of the ICDI in monitoring forest water stress and its potential for broader applications in forest management and climate adaptation strategies.

1. Introduction

Fire is a periodic factor in forest ecosystems but can have profound impacts on other components, such as water resources [1,2,3]. Rising global temperatures, prolonged droughts, and shifting weather patterns create drier conditions, making forests more susceptible to wildfires (WF) [4]. Warmer temperatures dry out vegetation, while reduced rainfall and prolonged droughts increase the flammability of forests and thus the extent and severity of fire following ignition [5]. Stronger winds also further accelerate fire spread [6]. Although natural causes, like lightning, and human activities, such as deforestation, contribute to WF, climate change acts as a multiplier, making fires more severe and harder to extinguish [4,7]. On the other hand, prescribed burning (PB) has been widely implemented in many countries to mitigate the risk of WF and minimize the damage they can cause [8,9,10].

Wildfires are unmanaged fires that spread rapidly across forests and grasslands, often causing significant environmental and economic damage. In contrast, PB is a fire management technique where forest managers deliberately set small, low-intensity fires under safe conditions to reduce fuel loads to prevent larger, more destructive wildfires. This method reduces the build-up of dry fuel that could otherwise lead to damaging wildfires [10,11].

The widespread use of PB also has a noticeable impact on forest-watershed hydrology [2,12,13,14]. The reduction in natural forest cover directly affects the water balance and water movement within the ecosystem [15]. Key effects include increased runoff and peak water flows, decreased soil permeability, and reduced interception and evapotranspiration [2,13,15,16,17,18]. In addition to its immediate and direct impacts, the effects of fire on hydrological processes become particularly evident within one to two years after the event [18,19]. However, research suggests that these effects can remain significant for up to seven years [18,19,20], depending on the severity of the fire. The recovery of hydrological processes to pre-fire conditions is closely tied to the restoration of vegetation cover to its undisturbed state [3,21,22]. Other critical factors influencing recovery include soil properties, erosion characteristics, and topographical features, particularly slope [20,23].

In southwestern Australia, fire has played a crucial role in shaping species composition. Indeed, species such as jarrah (Eucalyptus marginata), which is fire-resistant or capable of recovering well after fire, have a competitive advantage in dominating the forest ecosystem [1,24,25]. The Northern Jarrah Forest (NJF), which is mainly covered by E. marginata, experiences a Mediterranean climate, characterized by a stark contrast between warm dry and cool wet seasons. Prolonged dry spells, coupled with low rainfall and a hot, dry season lasting up to six months, create highly favourable conditions for forest fires [26,27,28].

In the context of climate change, there is growing evidence in this region of reduced rainfall and stream flow [29], more frequent and intense droughts, and severe heatwaves [30,31,32,33,34]. Prescribed burning has been widely practiced in jarrah forests since the early 1960s with one of the world’s most extensive and systematic prescribed burning programmes, aiming to reduce fuel loads and prevent catastrophic wildfires [11]. This approach was influenced by both indigenous fire management practices and more recent scientific research, demonstrating the benefits of PBs in maintaining forest health and biodiversity [35,36]. Despite its success in reducing WF severity, the use of PBs remains a topic of debate due to concerns about smoke pollution, carbon emissions, and impacts on wildlife [11]. Nevertheless, the effectiveness of targeted burning is undeniable, and it has become a widely used, multi-purpose management tool in other regions of the world, including Europe and North America [37,38].

Jarrah has an extensive root system, which enables it to access deep soil water reserves in the regolithic soils throughout the summer months [39,40,41]. Although jarrah trees are known for their resilience to drought and fire, natural jarrah forests are increasingly experiencing water stress, which can result in canopy damage and tree mortality [27,34,42,43,44]. Jarrah is well-equipped to withstand and recover from forest fires, which are a natural feature of its habitat. Its thick, fibrous bark acts as a protective barrier, shielding internal tissues from heat damage, while its underground lignotubers store energy and nutrients, enabling regeneration even if the tree’s above-ground structure is destroyed [24]. Jarrah also relies on epicormic buds hidden beneath its bark, which can rapidly sprout new leaves and branches following a fire. Fire also plays a crucial role in stimulating seed germination by breaking seed dormancy and providing nutrient-rich ash beds for young seedlings to establish. These traits enable jarrah not only to survive but to flourish in fire-regulated ecosystems, securing its dominance in the Northern Jarrah Forest (NJF) [24]. However, its resilience may be threatened by the increasing frequency and intensity of drought events driven by climate change due to weak physiological drought avoidance mechanisms and increasingly limited access to decreasing deep soil water reserves.

Numerous studies have highlighted the strong connection between water stress and forest canopy cover [45,46,47], especially leaf area index (LAI) as it directly reflects tree transpiration [48,49] and interception of rainfall [50]. Additionally, we previously observed that fire events leading to a decrease in canopy cover may have a positive impact on canopy dryness [51]. Obviously, to mitigate the adverse impacts of water stress on forest ecosystems, management interventions are essential. Among these, ecological thinning is considered a priority for jarrah forests [52,53]. However, this approach faces significant challenges due to its high demands for financial and time resources. In contrast, PB is a long-established and scalable practice with sustainable impacts on forest canopy health and understory vegetation [28,53]. The 70-year history of PB in jarrah forests thus offers a deeper and more nuanced understanding of the complex interactions between fire and forest hydrology, particularly in relation to canopy water stress under a drying climate.

Remote sensing has been used to monitor post-fire forest recovery since the 1990s [54,55,56], primarily by assessing changes in canopy cover through popular vegetation indices such as Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Normalized Burn Ratio (NBR), which rely on visible and infrared wavelength bands [56,57]. Radar and LiDAR (Light Detection and Ranging) data have also been used to assess the recovery of vertical forest structure [58,59,60]. However, there have been limited applications of remote sensing to monitor the recovery of forest ecological functions, particularly in terms of water balance. This aspect is especially crucial for regions frequently threatened by drought and heatwaves, such as jarrah forests.

In our previous study, we observed that prescribed burnt forest appeared to be less stressed than unburnt forest, based on a single temporal snapshot. This observation is important because the forest is under increasing water stress and damage from wildfire under climate change. However, we did not investigate this in any detail. Therefore, this study, utilized remote sensing indices [51] derived from multi-spectral and multi-temporal satellite imagery, to examine effects of PB and WF on canopy water stress. The findings in this study will provide insights for the conservation of natural forest ecosystems in southwestern Australia and other similar regions, especially in the face of changes in climate conditions.

2. Materials and Methods

2.1. Study Location

The Northern Jarrah Forest, an expansive broad-leaved evergreen eucalypt forest covering 11,276 km² and situated at an average elevation of 300 m above sea level [61], is a key ecological system in southwestern Australia (Figure 1A). Much of the forest has developed over deep and strongly weathered profiles which have poor fertility but provide significant soil water storage that allows the forest to survive during the annual summer drought [62]. For over five decades, the forest has been managed to support various activities, such as timber production, bauxite mining, biodiversity preservation, water catchment protection, WF prevention through PB activities, and recreational use.

The NJF experiences a typical Mediterranean climate characterized by two distinct seasons: a cool, wet season and a long, hot, dry season lasting from 4 to 7 months. The forest exhibits a pronounced west-to-east rainfall gradient, with annual precipitation exceeding 1100 mm along its western edge and decreasing to around 700 mm in the eastern and northern areas [41].

2.2. Data Sources

2.2.1. Remotely Sensed Data

This study utilized satellite imagery from Landsat Collection 2 Level-2 Science Products, which includes scene-based global surface reflectance and surface temperature data [63,64,65], spanning from 1987 to the present. A total of 70 Landsat scenes from path/row 112/082 were collected, covering approximately 63% of the northern portion of the Northern Jarrah Forest (Figure 1A). These scenes were selected due to the high frequency of both WF and PB in the area, as well as the increased likelihood of obtaining cloud-free images compared to those covering the southern part of the forest.

Based on data availability, Landsat images from three sensor types, namely Landsat 4-5 TM, Landsat 7 ETM+, and Landsat 8-9 OLI/TIRS, were selected, capturing the beginning and end of each dry season from 1987/1988 to 2023/2024. Exceptions were the end of the 1988/1989 season, the entire 2011/2012 season, and the beginning of the 2012/2013 season due to the absence of data. Three spectral bands, namely Red (RED), Near Infrared (NIR), and Short-Wave Infrared (SWIR), were used for remote sensing index calculations.

2.2.2. Fire History and Control Unburnt Data

The study used the fire history dataset (DBCA-060) provided by the Department of Biodiversity, Conservation, and Attractions, Western Australia [66]. This dataset contains records of fire events, including PB, WF, and other fire events, on government-managed lands, dating back to 1937. For this study, 3761 fire occurrences of the two main types (PB and WF) from 1987 onward within the study area were extracted for further analysis (

Table 1. Remote sensing indices used for detecting plant water stress.

Fire Type	Number of Events	Minimum Area (ha)	Maximum Area (ha)
Prescribed burning	1068	0.13	25,392.77
Wildfire	2793	0.09	28,289.89
Other fires	1308	0.03	3956.03

). Each record is represented as a single polygon delineating the fire-affected area and includes various attributes such as the calendar year of the fire, fire season, affected area, fire type, and other relevant details.

Random points were generated using ESRI ArcGIS 10.8.2 at a density of one point per 10 ha within the fire-affected area (sample points) and the adjacent unburnt area (control points) (see example in Figure 1B). The control area was selected to have similar forest cover to the burnt area prior to the fire, as represented by NDVI feature values. Where possible, high-quality time-series images from Google Earth, utilizing a time-lapse tool, were used to verify the similarity of forest cover between the sample and control areas. Additionally, the locations of both sample and control points were required to have NDVI values above the first quartile (Q1) of all pixel data within the area (see Figure 2 for an example of NDVI Q1 determination). This condition helps minimize the influence of non-forested areas and fragmented shrublands, ensuring greater consistency for further analyses.

2.3. Analysis

The overall methodology, where Landsat imagery time series was used in association with actual fire history records to compare remote sensing indices of burnt and unburnt forest canopies is illustrated in Figure 3.

2.3.1. Data Pre-Processing

The fire records, along with the Landsat data, were projected to a common coordinate system, the World Geodetic System 84 (WGS 84), Zone 50 South. All satellite images were clipped to the boundary of the NJF to optimize memory usage and reduce processing time. Additionally, the digital number values from the satellite images were converted to surface reflectance values for each band using the scaling factor of 0.0000275 + (–0.2) [63,64,65].

2.3.2. Data Processing

Normalized Difference Vegetation Index (NDVI), Normalized Difference Infrared Index (NDII), and then Infrared Canopy Dryness Index (ICDI) [51] were calculated for each Landsat image. The ICDI is an index based on the relationship between NDVI and NDII, designed to express differences in canopy water content relative to the water content observed under the maximum water stress condition, while controlling for similar levels of canopy greenness.

The NDVI utilizes NIR and RED wavelengths to estimate fractional vegetation cover, which is determined as follows:

$$NDVI={\displaystyle \frac{NIR-RED}{NIR+RED}}$$

(1)

The NDII [67], which employs NIR and SWIR1 bands to estimate the dynamics of water content in the vegetation canopy, is calculated as:

$$NDII={\displaystyle \frac{NIR-SWIR1}{NIR+SWIR1}}$$

(2)

Meanwhile, ICDI [51] which effectively depicts forest dryness utilizing the differences in NDII/NDVI relationship, is determined as below:

$$ICDI=\left(aNDV{I_i}^2+bNDV{I}_i+c\right)-NDI{I}_i$$

(3)

where a, b and c are the parameters from the identified dry edge function, which is the region indicative of significantly low water content in the NDII/NDVI feature space (see Le et al. [51] for more details). NDII_i and NDVI_i are the values at the pixel i.

2.3.3. Changes in Forest Canopy Dryness over Time

To analyze temporal changes in overall forest water stress across the NJF, ICDI maps were generated at the beginning and end of each dry season (subject to cloud-free imagery availability) for the period 1987–2024. A visual comparison was conducted by overlaying these maps to observe changes over time.

We assessed whether water stress had increased over time, particularly in the context of climate change, by classifying ICDI values into three categories representing increasing canopy dryness (low, medium, and high) using an equal interval classification. The proportion of the NJF area falling within the high dryness class was calculated annually at the end of each dry season when water stress is expected to peak. These data were graphed alongside annual rainfall records, with trendlines derived from linear regression, to observe any long-term trend. Rainfall data were from Mundaring (Station No. 9030) (31.898°S;16.158°E) at the centre of the study area, provided by Australia Bureau of Meteorology (http://www.bom.gov.au/ (accessed on 15 November 2024)).

2.3.4. Exploring Prolonged Effects of Fire on Canopy Water Stress

We randomly selected 10 fire events (5 WF and 5 PB), to examine how the dryness of the forest canopy responded to fire by analyzing changes in the ICDI (

Table 2. Summary of the ten selected prescribed fire (PB) and wildfire (WF) events. Coordinates based on World Geodetic System 84 (WGS 84), Zone 50 South.

No.	Fire Event	Coordinates (X;Y)	Fire Type	Date of Fire	Area (ha)	Number of Sample Points	Number of Control Points
1	PB_19951024	422,045; −3,553,428	PB	16 September 1995	1911.9	191	130
2	PB_19990501	441,336; −3,576,979	PB	1 April 1999	1468.5	147	125
3	PB_20071014	437,977; −3,562,773	PB	14 October 2007	1959.7	196	128
4	PB_20141015	434,078; −3,469,660	PB	15 October 2014	1465.7	147	116
5	PB_20170606	446,960; −3,571,018	PB	6 June 2017	2835.8	284	148
6	WF_19991223	446,240; −3,581,614	WF	1 April 2000	1960.4	196	127
7	WF_20020202	465,461; −3,600,251	WF	1 November 2001	1358.7	136	125
8	WF_20041220	435,376; −3,570,263	WF	20 December 2004	1773.2	177	120
9	WF_20080108	440,016; −3,581,642	WF	8 January 2008	606.9	61	72
10	WF_20180114	429,971; −3,537,654	WF	14 January 2018	3259.5	326	167

). To ensure enough sample points, the events were chosen from those with a burned area > 300 ha, which included 404 PB and 69 WF events. Additionally, all selected areas were required to have been fire-free for at least 10 years prior to the event to minimize the impact of previous fires on the analysis. Each sample was checked against the Landsat image database to confirm that the fire date recorded on DBCA-060 matched the actual date.

After generating the sample and control points, NDVI and ICDI values from 1987 to 2024 were calculated. These values were then plotted over the 37-year period. Additionally, the data were analyzed using the Mann–Whitney U Test built in Microsoft Excel software to determine if there were significant differences between the two datasets after fire events and to assess how long it took for the burnt forest to recover, in terms of both canopy cover and dryness, compared with the unburnt forest.

To further highlight the impact of fire, the study conducted a fire severity classification for each selected event using the dNBR (Differenced Normalized Burn Ratio) index following the method proposed by USGS [68]. The indices were calculated as below:

$$NBR={\displaystyle \frac{NIR-SWIR2}{NIR+SWIR2}}$$

(4)

$$dNBR=preNBR-postNBR$$

(5)

Fire severity was classified into five levels as in

Table 3. Burn severity levels obtained calculating dNBR (Source: USGS [68]).

Severity Level	dNBR Range
Enhanced Regrowth, high (post-fire)	−0.500 to −0.251
Enhanced Regrowth, low (post-fire)	−0.250 to −0.101
Unburned	−0.100 to +0.099
Low Severity	+0.100 to +0.269
Moderate–low Severity	+0.270 to 0.439
Moderate–high Severity	+0.440 to +0.659
High Severity	+0.660 to +1.300

.

3. Results

3.1. Canopy Dryness in the Northern NJF from 1987 to 2024

The study produced ICDI maps for each processed image, showcasing the patterns of canopy dryness in the target area over time. Figure 4 presents ICDI maps from late in the dry season at regular intervals over the period 1987 to 2024. Overall, dryness appears to increase gradually, as evidenced by the visible expansion of orange and red areas on the maps, indicating higher dryness levels. In certain periods, some areas exhibit unusual changes in dryness levels, clearly visible as patches (usually in light blue) on the maps. These fluctuations are likely driven by environmental factors, particularly pests, forest fires, forest harvesting and mining rehabilitation, which can alter the structure and function of the forest canopy.

The marked increase in water stress over time in the study area is evident in the increasing proportion of high dryness areas (Figure 5) which rose from 20% in the late 1990s to approximately 48% in the early 2020s (r² = 0.708). An overall decline in annual rainfall is associated with the increasing trend in the proportion of high dryness areas over time, especially in the period from 2010 onwards (Figure 5).

3.2. Cover and Dryness of Burnt and Unburnt Forests

Figure 6 shows the changes in the NDVI and ICDI remote sensing indices for the corresponding control areas and PB and WF areas. This is for 2–3 years before and up to around 10 years after each individual fire. Before each fire, the indices for the control (unburnt) and burnt areas had similar values (NDVI in the range of 0.5–0.6 and ICDI values around −0.1). The trends and rates of change were consistent over the later years. This indicates a similarity in the vegetation greenness of the forests between the control and burnt areas. However, immediately after the fire, both the NDVI and ICDI indices for the burnt samples decreased significantly, while the corresponding values for the control areas remained stable.

Over the following years, the changes in the two indices became notably differentiated (Figure 6). For PB and low-intensity WF, the NDVI recovered rapidly from average NDVIs of 0.4 immediately after the fires. After a short period (1–2 years for low-intensity PBs, 2–3 years for moderate-intensity WFs), the NDVI values for the burnt areas increased markedly, and matched the values in the control areas (

Table 4. Severity categories and time for NDVI and ICDI values to recover to adjacent unburnt forest values for 10 selected fire events in the Northern Jarrah Forest.

No.	Fire Event	dNBR	Severity Level	Time for NDVI Catch-Up (Years)	Time for ICDI Catch-Up (Years)
1	PB_19951024	0.1898	Low	0.5	6.5
2	PB_19990501	0.1670	Low	1	7
3	PB_20071014	0.1508	Low	1	7
4	PB_20141015	0.3506	Moderate–low	9	>10 (on going)
5	PB_20170606	0.1364	Low	1.5	6.5
6	WF_19991223	0.3812	Moderate–low	1.5	12
7	WF_20020202	0.6588	Moderate–high	2.5	14
8	WF_20041220	0.7026	High	2	>19 (on going)
9	WF_20080108	0.9464	High	1.5	8.5
10	WF_20180114	0.8310	High	6	>6 (on going)

). In some cases, the burnt forest canopy recovered so strongly that the NDVI values slightly surpassed those of the unburnt areas.

However, the dryness of the forest canopy, as indicated by ICDI values after each fire, remained at lower levels for a considerably longer period. In the case of PBs, the ICDI values in the burnt areas remained lower for 6.5–7 years compared to the unburnt areas (Table 4). For wildfires, the NDVI values usually decreased to below 0.2 due to tree crown loss. Here, the ICDI values in the burnt areas took approximately 10–20 years to catch up with the unburnt areas. An example of the NDVI and ICDI pattern changes before and after a fire is shown in Figure 7.

Examining the selected sample sites, we found that site WF_20041220 experienced two fires during the 37-year analysis period, PB in late 1989 (low severity; dNBR = 0.1842) and then a high-severity WF in late 2004 (Figure 8). Notably, the control area remained fire-free throughout the entire period, and the second fire occurred at the site when the forest had recovered to the same state as the control area. The recovery time was greater for ICDI than for NDVI. However, with different fire intensities, the specific catch-up time corresponding to each fire event is different (Table 4).

The sample area was prescribed burnt in 1989 and within 5 months, the canopy had recovered to the level of the unburnt forest with an NDVI value of nearly 0.6. However, it was not until late 1996, 7 years later, that the post-fire ICDI values had caught up with the control forest with values of approximately −0.1 (Figure 8).

The similar conditions between the prescribed burnt and unburnt areas were maintained from 1996 until 2004. In nearly one year before the WF in 2004, they both experienced a decline in canopy (i.e., a decrease in NDVI) and a rapid increase in ICDI indicating that the forest was under water stress. This could also be considered a factor that increases the risk of fire. After the fire, the forest canopy recovered after 2 years, however by the time this analysis was performed (June 2024), nearly 20 years after the fire, the ICDI values in the burnt area were still considerably lower than the control area.

4. Discussion

4.1. ICDI Application for Forest Water Stress Monitoring

Forest water stress management has become increasingly important under climate change, as drought events become more frequent and severe. Satellite-based remote sensing, particularly using visible and infrared reflectance data, has emerged as the dominant method for monitoring water stress impacts and forest recovery over large areas. The integration of multispectral and hyperspectral satellite imagery, such as that from Landsat time-series, Sentinel-2 and MODIS [69], has facilitated the development and application of general vegetation indices and more specialized indices like the Temperature Vegetation Water Stress Index (TVWSI) [70] and the IASI Water Deficit Index (IASI-WDI) [71], which incorporate canopy temperature, water content, and vegetation structure to detect stress conditions with greater sensitivity [48]. General vegetation indices derived from canopy spectral reflectance, such as NDVI and EVI [72,73], are widely used due to their compatibility with satellite data, though they are limited in sensitivity and effectiveness for early detection. The need for more accurate assessment tools has driven research toward developing new indices using narrower spectral bands, but such approaches are currently restricted by the coarse spatial and spectral resolution of most satellite platforms. Developing new, more sensitive indices in association with the use of higher-quality satellite imageries, as well as refining and combining existing indicators, presents promising directions for future applications for forest water stress monitoring.

The ICDI can clearly detect changes in canopy dryness, thus providing valuable insights into patterns of forest water stress following prescribed burns and wildfire. The results indicate that fire plays a role in maintaining lower levels of canopy dryness, thereby reducing water stress intensity for several years following a fire event. Since low-intensity fires in jarrah forests primarily burn the understorey canopy and ground litter, without significantly hindering post-fire regeneration, these findings align with a previous study that showed that the understory species in the jarrah forest accounted for 51% of the evaporation [74]. It also aligns with the results of a study in this region that demonstrated an increase in soil water storage and recharge following a prescribed fire in a Banksia woodland [75]. The longer time to recovery after wildfire can be ascribed to the removal of both the understory and overstory canopies. Previous studies have established a strong relationship between leaf area (i.e., LAI) and plant water stress [76,77].

The success of this application is particularly significant in the context of climate change, which is increasing the risk of water stress in forests worldwide [32,78] and the jarrah forest in particular [34,42,44]. Popular ground-based methods for assessing water stress, which include predawn leaf water potential, sap flow measurement, leaf chlorophyll fluorescence, pigment concentrations and leaf water content [79,80], are not only time-consuming and labour-intensive but also most rely on destructive sampling. Employing ground-based methods is impractical for large-scale applications [79]. In contrast, ICDI enables the use of multi-temporal remote sensing data, making it possible to assess water stress even in inaccessible forest areas or for extensive fire events. The availability of continuous historical remote sensing data further enhances the ability to track fire-related changes over time. When sufficient data are available, calculating ICDI for a given study area is straightforward.

This study utilized Landsat imagery with a spatial resolution of 30 m. Theoretically, fires with a minimum area of 900 m² (corresponding to a single Landsat pixel) could be examined. However, in practice, relying on a single pixel is unreliable, especially if it is located at a boundary between different land cover types or contains heterogeneous vegetation cover. To ensure statistically representative sample points, this study focused on fires larger than 300 ha. While no specific research has yet determined the minimum fire area applicable to ICDI, fire patterns tend to mirror the distribution and composition of surrounding forest vegetation. Based on previous studies, particularly Lu et al. [81], we recommend applying ICDI to fires with a minimum size equivalent to a 9×9 pixel window, approximately 7.3 ha, when using Landsat data.

Beyond the success in wildfire monitoring, ICDI also holds potential for assessing the impacts of other forest disturbances, both natural (e.g., drought, pest infestations) and anthropogenic (e.g., thinning, pruning, and other silvicultural interventions), on forest canopy dryness. Furthermore, given the widespread availability of multispectral remote sensing data, we expect that the ICDI has broad applicability across diverse forest types globally, though validation will be necessary for each specific forest type. In parts of the world where PB is being used in forests experiencing increasing water stress and the threat of wildfire, there is the opportunity to explore the use of this index to prioritize areas for intervention. Examples could include the pine forests in Sierra Nevada, USA where PB is recommended for reducing tree mortality from drought [82], and for increasing landscape heterogeneity and ecosystem resilience to wildfire and drought [83]. This is also supported by various studies in the Mediterranean region, such as in Spain [84,85] and Portugal [86], where PB has been effectively used as a forest management tool.

4.2. Prescribed Burning—Water Stress Relief Under Climate Change

Fire is a unique ecological factor in forest ecosystems, capable of both destruction and promoting biotic growth. When unmanaged, as in the case of wildfires, the consequences are almost always severe. However, forest managers have long used fire as a strategic tool to minimize the risk and intensity of wildfires. Through prescribed burns, they reduce combustible materials in the forest while avoiding significant adverse effects [10,11,58].

This study focuses on jarrah, a tree species known for its resilience to fire [24,39]. The NDVI results show that when fires are low-intensity or occur under unfavourable conditions for burning, the forest canopy can recover within a year. However, beyond canopy recovery, this study also investigates an often-overlooked aspect of forest health: canopy dryness. This reflects the lack of available water in the canopy to support optimal growth [51] with the IDCI remote sensing index considered as an indicator of water stress. Notably, in forests experiencing high level of water stress, drier canopies increase the likelihood of fire spread and intensity following ignition [87].

Recently, Camarero et al. [88] also explored the long-term impact of fire on forest growth, particularly in the context of drought and water stress. The authors confirmed a link between fire and reduced water stress in Mediterranean pine forests. The study was limited to using growth indicators, specifically tree-ring measurement, to compare individual trees on a small scale. By monitoring ICDI across different fire events, we found that forest fires significantly reduced signs of water stress for extended periods (from 7 to more than 20 years), even after the canopy had fully recovered (Figure 6, Table 4). The results of the ICDI analysis are also consistent with our observation in Le et al. [51]. Previous studies have examined the effects of fire on hydrological behaviours based on calculating water balance at the watershed scale, which affirmed that these effects had typically been observed to last between 5 and 7 years [18,19,20].

Several factors may explain this phenomenon. First, the reduction in both overstorey and understorey vegetation after a fire leads to lower overall leaf area, thereby reducing water demand through transpiration. Various studies have confirmed the relationship between defoliation and the reduction in water use in forest after fire [89,90]. Additionally, water that would have otherwise been used by the shrub layer becomes available for overstorey trees. Second, beyond altering canopy structure, fire also clears understorey vegetation and litter, significantly reducing interception [50]. As a result, more rainfall infiltrates the soil, increasing soil water stores and thus increasing subsequent water availability for trees through the dry season. Numerous studies have found that wildfires can enhance water availability and even raise groundwater levels [2,13,91]. This effect is particularly relevant for jarrah, a deep-rooted species with an extensive root system capable of accessing groundwater. Fire also temporarily increases nutrient availability and forest growth in the NJF [92].

In practice, the influence of spatial clustering and local heterogeneity on forest canopy may exist. However, there have been no studies on these factors in the NJF. In selecting 5 random sites for each of PB and WF, we ensured that the control and sample areas were paired with the same environmental conditions. In addition to random sampling, we tried to minimize the influence of spatial clustering and local heterogeneity by applying the condition “NDVI values above the first quartile (Q1)” as a filter. The homogeneity of sample and control points is reflected in the similarity of the indices, in terms of both figure and trend, for a long period before each fire event. In the future, a larger number of forest sites could be used to examine the effect of forest structure and heterogeneity on forest canopy response after fire.

Given the effect of PB in alleviating water stress reported here, a key question arises: which areas should be prioritized for PB implementation? Since applying this approach across an entire forest is impractical due to time and resource constraints, strategic planning is essential. In this context, the ICDI presents itself as a valuable decision-making tool. Mapping ICDI patterns can help identify areas with high canopy dryness, allowing forest managers to prioritize interventions where they are needed most. Although ecological thinning has been advocated in this region as a means of reducing forest water stress [53]; this involves mechanical intervention, is costly and has only been applied to small areas. Prescribed fire can be surmised to be a relatively cheap alternative.

There are concerns in the literature regarding the effect of forest fire on other ecosystem services. The effects of prescribed fire on biodiversity are contentious; with direct effects on biodiversity reported [93,94]. These have to be contrasted with the catastrophic effects of wildfire on biodiversity, which prescribed fire alleviates. Water stress will also result in biodiversity loss in the NJF [28], and alleviating this can potentially reduce these losses. In addition, carbon emissions from fire in the NJF come from the combustion of litter and under- and overstorey canopies [95]. In Australian Government National Greenhouse Gas Inventory reports to the United Nations Framework Convention on Climate Change (UNFCCC), these are regarded as neutral, as the forest regrows after fire [96]. Carbon losses from drought in this forest have also been quantified [44,97]; thus relieving water stress will likely reduce drought deaths and carbon loss via this route.

5. Conclusions

This study utilizes multi-temporal Landsat remote sensing data combined with historical fire records to investigate the effects of fire, including both prescribed burning (PB) and wildfire (WF), on water stress in the Northern Jarrah Forest ecosystem. The ICDI is employed as an indicator of canopy water stress.

Temporal analysis of ICDI data reveals a clear trend of increasing high-canopy-dryness areas within the study region. Analysis on burnt areas further shows that post-fire canopy dryness remains at lower levels for an extended period, even after full canopy recovery. We observed that, depending on fire severity, the reduction in water stress persisted for 7–8 years following low-severity prescribed burns and more than 20 years after high-severity wildfires. These findings suggest a role for PB in alleviating water stress, and that PB can serve as a strategic tool for water stress mitigation particularly with the advent of climate change.

Additionally, the ICDI proves to be a valuable decision-making tool, as its spatial patterns can help identify areas with high canopy dryness, enabling forest managers to prioritize interventions where they are most needed. Beyond its success in monitoring the impacts of wildfire, the ICDI also holds significant potential for assessing the impacts of other forest disturbances—both natural (e.g., drought, pest infestations) and anthropogenic (e.g., thinning and other silvicultural interventions)—on forest physiological responses.