Tracking the Extent and Impacts of a Southern Pine Beetle (Dendroctonus frontalis) Outbreak in the Bienville National Forest

Abstract

The Bienville National Forest (BNF) in central Mississippi experienced an outbreak of southern pine beetle (SPB) beginning in 2015 and continuing through 2019. To assess the extent of the outbreak and subsequent treatments of impacted areas, high-resolution imagery was obtained from various sources and interpreted to determine the feasibility of this imagery for detecting SPB spots and tracking their spread and treatments. A negative binomial regression model then described the relationship between spot detection, year, and status (i.e., infestation/treatment) and then incidence rate ratios were calculated. The peak active infestation acreage occurred in 2017 with treatment and mitigation measures peaking in 2018. In total, over 4450 hectares (ha) were treated through 2019 in the BNF. Overall, it is possible to discern small areas of active SPB infestation and treatments. If used as a monitoring method, consistently available high-resolution imagery (e.g., from a satellite) provides an effective means of detecting, evaluating, and tracking infestations and related treatments.

1. Introduction

The pine-type forests of the southern United States South are susceptible to a variety of non-harvest disturbances that can significantly reduce standing timber inventories [1,2,3]. Some disturbances, such as fire and severe weather, are short-lived, discrete events. Infestations and outbreaks driven by insects and diseases, however, can occur over multi-year cycles. The most impactful native insect to southern pine forests is the southern pine beetle (SPB, Dendroctonus frontalis Zimmerman). The region’s last widespread epidemic surpassed four million hectares and $1 billion [USD] in economic damages [4].

The SPB attacks all pine species; its targets are not limited to weakened or dying trees (e.g., from storm damage). It can also kill otherwise healthy trees and spread throughout a forest using pheromone communication signals that enable successful mass attacks and infestation (‘spot’) growth [5]. This typically occurs in denser stands with overstocked conditions. SPB prevention and control involves silvicultural methods (e.g., thinning and prescribed fire) to optimize stand density in a way that disrupts communication signaling between attacking beetles [6]. The Southern Pine Beetle Prevention and Restoration Program was established by the USDA Forest Service and Southern Group of State Foresters (in 2003) to assist communities through education, research, and stand-level forest operations [4]. Kushla et al. [7] detailed the programmatic impacts in Mississippi. Overall, benefits exceeded cost by 13 to 1.

This program has benefited both privately owned timberlands and the national forests. However, given the abundance and concentration of a high-hazard stand conditions on the national forests, the pace and scale of thinning and other harvesting on national forest lands have not alleviated the overall risk of SPB outbreaks occurring on the national forests in the south [8]. Harvests on national forests have declined substantially from 12 billion board feet in fiscal year 1998 to less than 3 billion in fiscal year 2018 nationally [9]. Timber receipts from Mississippi’s national forests totaled $21 million in 1988 and $1.02 million in 2018 [9]. Growth now exceeds removals on national forest timberlands in Mississippi by a ratio of 5.44 to 1.00; this varies by region from 0.92 to 1.00 in the Mississippi delta (a region predominately devoted to crop production with little timberland) to 10.56 to 1.00 in south Mississippi [10]. Across the South, SPB activity and subsequent loss in timber value frequently sum upwards of $40 million annually [11].

The National Forests (NF) in Mississippi experienced a brief outbreak in 2012 on the BNF and to a greater degree on the Homochitto NF [12], with less managed stands experiencing the majority of the infestations [6]. The same two forests experienced a longer-duration and more severe outbreak from 2015–2019; both events align with predicted suitability for SPB activity [3]. The U.S. Forest Service reported nearly 700 SPB spots in the year 2016 alone, 361 in the Homochitto NF and 317 in the BNF (see https://www.fs.usda.gov/Internet/FSE_DOCUMENTS/fseprd562900.pdf, accessed on 17 December 2022). Damage detection includes using aerial imagery (from aircraft and Unmanned Aerial Vehicles) and satellite data that encompass manual and automated classification techniques [13,14,15,16,17]. Predictive models then use a variety of indices, spatial trajectories, and manual interpretations to predict the risk of beetle infestation [18]. While the scale ranges from single tree to stands to forests, operationalizing remotely-sensed data into assessing insect disturbances at greater spatio-temporal scales has been suggested [19]. It is difficult to obtain imagery at a consistent temporal and spatial resolution that allows for adequate mapping of insect spots within a forest [20]. Accomplishing this may necessitate incorporating and integrating data from multiple spatial scales. At an adequate resolution tracking and mapping outbreaks and responses (treatments) become more feasible.

Mississippi is a rural state. Timber provides the greatest gross value added among the state’s agricultural products [21], and Mississippi’s bioeconomy was the most industrially concentrated in the United States [22]. The state is comprised of approximately 7.7 million hectares of forest with more than 40 percent of that (3.2 million ha) classified as pine forest [23]. The forestry sector in Mississippi accounts for more than 69,000 (over 4% of all jobs in the state) and has an industry output of over $12 billion [24]. Biological shocks like SPB outbreaks can therefore inflict serious damage on the forestry supply chain and economic well-being of businesses and communities. The present study seeks to determine the spatial and temporal extent of the 2015–2019 outbreak by annually quantifying the infestations and treatment responses within the BNF in Mississippi.

2. Materials and Methods

The BNF is located in central Mississippi between Jackson and Meridian (Figure 1). The proclamation area of the BNF is 157,150 ha with approximately 72,000 ha managed by the United States Forest Service. It is intersected by multiple state highways and county roads as well as U.S. Highway 80 and Interstate 20; within the proclamation boundary there are multiple land uses (farms, home sites, etc.) in privately owned lands. Pine (Pinus spp.) species dominate the forested areas of the BNF. The region is classified as Humid-subtropical with an average annual high temperature of 24.3 °C, average annual low temperature of 10.9 °C, and averages 157.2 cm of precipitation per year (mostly in the form of rain).

Given the temporal extent of the outbreak (four years), it was not possible to obtain aerial imagery with a consistent spatial resolution. For 2016, data were acquired from the National Agricultural Imaging Program (NAIP) as county-level orthomosaics, with a 1 m resolution [25]. The NAIP imagery was acquired between late August and mid-September 2016 and processed according to NAIP standards [26]. Eight-band WorldView-2 imagery was obtained in October 2017, with a resolution of ~2 m (https://earth.esa.int/eogateway/missions/worldview-2, accessed on 17 December 2022); there were a total of 12 scenes covering the whole of the BNF proclamation area. In June-July 2018 and 2019, aerial imagery acquisition was contracted for the BNF and had a spatial resolution of 0.28 and 0.27 m, respectively. Even though the image resolutions vary, these high-resolution datasets were still adequate to identify the SPB impacts and treatments in a given year.

The imagery was reprojected to a common coordinate system (WGS84-WMAS) and displayed year-by-year in GIS software (ArcGIS Pro Version 2.7). The images were visually inspected and polygons manually created/digitized corresponding to SPB activity and/or treatment within the US Forest Service lands within the BNF; the minimum extent of a polygon was 10 trees/approximately 0.01 ha. Because this work seeks to classify active infestation, mortality, and various treatments, it was decided that a manual approach would work better than on using classification or machine learning techniques. Ultimately, five categories were utilized to classify the digitized polygons in order to quantify the extent of the outbreak and corresponding treatment/management activities by year. The polygons were classified as (1) active infestation—groups of trees meeting the size threshold that contained evidence of insect activity (i.e., yellow/red crowned trees known as ‘faders’) as determined from visual inspections; (2) standing dead—trees that appeared to have brown/gray needles or were needleless; (3) cut-and-leave—a treatment where trees were directionally felled into a central location that included a buffer cut of 30.5–45.7 m around an active SPB spot; (4) cut-and-remove—timber that was allowed to be cut and physically removed (i.e., harvested); (5) hazard tree mitigation—dead or infested trees within a right-of-way of a road, path, powerline, etc. where additional impacts/damage would be incurred if and as the dead trees succumbed to wind and/or rot.

The images were systematically assessed, generally from the northwest to southeast corners. When a spot was identified, a new polygon was created and immediately classified as belonging to one of the groups outlined previously. This process was repeated for imagery acquired in each year. Once each individual year was mapped, spots and subsequent treatments were evaluated spatio-temporally (via Intersect tool) to ensure correct classification and spatial extents (e.g., that a cut-and-leave in one year was not also delineated in a subsequent year, etc.). For example, if an area was delineated as ‘standing dead’ in one year, it would not be double counted in the subsequent year for accounting acreage impacts. It could, however, be treated in a subsequent year (e.g., ‘hazard mitigation’).

A Poisson regression model was originally constructed to describe the relationship between SPB spots, their year of detection along with the status identified on the spot

$$\ln \left({\mu }_{SPOT}\right)={\beta }_0+{\beta }_{1}2016+{\beta }_{2}2017+{\beta }_{3}2018+{\beta }_{4}2019+{\beta }_{5}ACTIV{E}_{INF}+{\beta }_{6}S{T}_{DEAD}+{\beta }_{7}CU{T}_{LEAVE}+{\beta }_{8}CU{T}_{REMOVE}+{\beta }_{9}HAZAR{D}_{MIT}+ϵ$$

(1)

where $\mu$_SPOT was the expected mean number of SPB spots detected; 2016, 2017, 2018, and 2019 were the years under study. Status included ACTIVE_INF, which was an active infestation; ST_DEAD was a spot occupied with standing dead trees; CUT_LEAVE was a spot treated by cut and leave, where trees were felled and left in the field; CUT_REMOVE was a spot treated by cut and remove, where trees were felled and hauled to local processors; HAZARD_MIT was a spot where beetle-killed trees were felled for hazard mitigation, such as those located along roadways; model parameters were ${\beta }_i$; and $ϵ$ was the error term. Testing between years at alpha = 0.05 evaluated whether ${\beta }_1={\beta }_2={\beta }_3={\beta }_4=0$ versus the alternative of at least one coefficient being different from zero. Analyzing the effect of status involved assessing whether ${\beta }_5={\beta }_6={\beta }_7={\beta }_8=0$ versus the alternative of at least one coefficient being different from zero (also at alpha= 0.05).

A Poisson distribution was initially considered, but overdispersion was present. The occurrence of spots and their status was considered a Poisson process with its own parameter µ_i, but these were not jointly identical. The Negative Binomial distribution allowed modeling heterogeneity in a population. It did so by relating the variance to the mean by

$${\sigma }_{SPO{T}_i}^2=\phi (\mu +k{\mu }^2)$$

(2)

The variable k provided flexibility for a negative binomial model to account for overdispersion due to its intuitive augmenting of the variance versus the Poisson model that does not contain k. The scale parameter, $\phi$, controlled overdisperson of residuals. Thus, the negative binomial model was the preferred analytical tool for the present assessment.

Least squares means were calculated for Years or Status if significant overall differences were identified through the likelihood ratio statistic, calculated as a type 3 chi-square value, for each variable grouping (year and status). From these, incident rate ratios (IRR) were computed to gauge incident rates for one year or status versus a benchmark year or status

$$IRR=\frac{e^{u_i}}{e^{u_{Reference}}}$$

(3)

where $e$ is the base of the natural logarithm, $u_i$ was a least squares mean for a year or status, and $u_{Reference}$ was the least squares mean for the reference year or status. The reference year was set to 2016, and the reference status was set to ACTIVE_INF. Equations (1) and (2) were simultaneously estimated in SAS 9.4 using the GENMOD procedure [27,28]. Parameters were estimated by maximum likelihood where the year 2016 and status ACTIVE_INF were designated reference groups and assigned zero degrees of freedom. Least squares means were calculated post-modeling using the PLM procedure.

3. Results

The SPB spots (infestation and treatments) unexpectedly increased in number and extent within the USFS-managed lands within the BNF through 2018 before decreasing as the infestation waned in 2019 (Figure 2). The status of spots detected by year track the active infestations and responses throughout the outbreak period (

Table 1. Contingency table describing southern pine beetle spots on a year (row) by status (column) basis.

Year	Active Infestation	Standing Dead	Cut and Leave	Cut and Remove	Hazard Mitigation	Yearly Totals
2016	607	476	63	2	0	1148
2017	485	574	176	7	1	1243
2018	487	1087	60	166	164	1964
2019	607	700	4	41	15	1367
Spot Status Totals	2186	2837	303	216	180	5722

). Of the 5700 spots identified over the four-year period, those with standing dead trees comprised nearly half the total, 49.6%. Active infestations made up another 38.2%. Thus, 87.8% of the total SPB spots had yet to be treated by forest operations crews. Total SPB spots by year were more uniformly distributed.

The difference observed between Years (p = 0.0519) could be best described as moderate, where 2018 was greater than the others (Figure 3). Highly significant differences were concluded to be present between statuses (p = 0.0002). Actively infested spots and those occupied by standing dead trees were significantly more common than sites where cut-and-remove or hazard mitigation had occurred (Figure 4). Diagnostically, the negative binomial regression model provided good fit. Deviance was 23.33 with 12 degrees of freedom, and overdispersion was minimal ($\phi$ = 1.94). The Akaike Information Criterion was 249.70, nearly five times below that produced by the Poisson regression model, where lower is better [29].

Table 2. Least squares means (natural logged) for southern pine beetle spots by status, and incident rate ratios for the status referenced to spots containing standing dead trees.

Year	Estimate	Difference from 2016	Incident Rate Ratio
2016	3.9912	----	1.0000
2017	4.4457	0.4545	1.5736
2018	6.0981	2.1069	8.2231
2019	4.6118	0.6206	1.8601
Status	Estimate	Difference from ACTIVEINF	Incident Rate Ratio
ACTIVEINF	6.5790	----	1.0000
CUTLEAVE	4.6211	−1.9579	0.1412
CUTREMOVE	3.3095	−3.2695	0.0380
HAZARDMIT	2.8080	−3.771	0.0230
STDEAD	6.6159	0.0368	1.0375

provides IRRs contrasting the spots detected in 2016 to the other study years. Additionally, included are comparisons of spots that were active infestations at the time to others with standing dead trees, or where a harvest operation had occurred (either cut-and-leave, cut-and-remove, or hazard mitigation). Table 2 illustrates that the incident rate for SPB spot detection in 2018 was 822% versus 2016 levels. Thus, we can conclude that 2018 had 8.2231 times the number of spots detected in 2016. This finding highlighted the rate of progression of the epidemic on the BNF as it ramped upwards from 2016 to 2017, rapidly expanded through 2018, before then lessening in 2019. Detecting standing dead trees was 103% that of detecting an active infestation, nearly a 1:1 ratio. Incident rates for detecting harvest operations were much lower. The incident rate for detecting a cut-and-leave operation was 14.1% that of detecting standing dead trees, meaning for every 100 active infestations detected 14 would be detected as being treated by cut-and-leave. Cut-and-leave was more likely to be detected than cut-and-remove by a factor of 3.72 (0.1412/0.0380). This allowed the recent trend in forest health management on national forests [28]. Hazard tree mitigation was the least observed; this treatment occurred where safety concerns weighted heavily (roadsides, utility lines, etc.).

The manual interpretation of imagery obtained covering the BNF from 2016–2019 led to the delineation of 5722 discrete spots over the four-year period. The total impact of the outbreak (including all categories) over the four year period was 8955 ha (approximately 12% of the BNF). The total area treated (i.e., cut-and-leave, cut-and-remove, or hazard tree mitigation) over the course of the outbreak was 4488.1 ha. At the onset of the outbreak, through the first year of assessment (2016), active infestation acreage was greater relative to treatment as the BNF began to suppress some of the numerous and enlarging infestations. Based on the interpretation of the imagery, from 2016 into 2017, the acreage actively infested nearly tripled from 337.1 ha to 952.6 ha. Coincident with the infestation increasing and peaking in 2017 (in terms of area actively infested), was the acreage treated. Areas actively infested and those occupied by standing dead timber totaled approximately 405 ha in 2016. This increased to an estimated 1618.7 ha in 2017. However, the area treated by the cut-and-leave method was 1149.3 ha. Combined with limited cut-and-remove operations, the total area treated in 2017 is approximately 1295 ha. This roughly matched the total area actively infested in 2016 and 2017 (1290 ha actively infested between the two years vs. 1303.5 ha treated in 2017).

Intersecting the polygons for each assessed year, it is possible to determine if infested areas were treated in subsequent years. In 2016, the BNF experienced some 240 ha actively infested; of these, approximately 45.7 ha were treated in 2017. Approximately 10% of the infested trees in 2016 were standing dead in 2017 with just over 20.2 ha being treated, predominantly via cut-and-leave. This does not necessarily mean that the majority of the infested areas were left untreated but indicates the areas that were directly visited the following year. Some areas were subject to buffer cuts with directionally felled trees so that the treatment polygon would not necessarily overlap in the actively infested polygons from the prior year.

The 2017 analysis indicated approximately 951 ha were actively infested. By 2018, nearly half of the area (522 ha) were standing dead. A further 136 ha were treated, with a majority in cut-and-remove, followed by hazard tree mitigation (cuts in rights-of-ways), and cut-and-leave. There were also nearly 121 ha of standing dead treated between 2017 and 2018, most of which being hazard tree mitigation. As previously mentioned, the suppression treatments began ramping up in 2017 and nearly equaled the sum of infested acreage in 2016 and 2017. This indicates a more proactive approach for mitigation and further validates the areas that appear untreated from 2016 to 2017 with the increase in cut-and-leave operations that included felling of uninfested buffers ahead of active spot heads, in an attempt to suppress active infestations.

Into 2018, the main mode of treating the outbreak shifted from cut-and-leave to cut-and-remove as approximately 2023 ha were cut within the BNF. This treatment had the added benefit of removing felled trees (containing beetles) from the forest and reducing the fuel load for potential wildfires, as well as providing a cleared area facilitating regeneration/restoration efforts going forward. There was a drastic reduction in the actively infested areas. Where 2017 was considered the peak of the outbreak, 2018 was the peak for the treatment protocols and preparation for the future. In 2018, actively infested acreage reduced by approximately 1/3 from 2017. Of the area actively infested (355.5 ha), 66.8 ha were either standing dead or treated in 2019 (

Table 3. Total acreage impacted by year and type.

Year	Class/Treatment	# Events	Area (ha)	Total/Year
2016	Active Infestation	607	337.10	547.37
	Standing Dead	476	66.36
	Cut & Leave	63	75.38
	Cut & Remove	2	68.53
2017	Active Infestation	485	952.79	2921.86
	Standing Dead	574	665.66
	Cut & Leave	176	1149.26
	Cut & Remove	7	149.27
	Hazard Tree Mitigation	1	4.88
2018	Active Infestation	487	355.52	4422.47
	Standing Dead	1087	1546.97
	Cut & Leave	60	173.26
	Cut & Remove	166	2027.96
	Hazard Tree Mitigation	164	318.77
2019	Active Infestation	607	315.53	1063.67
	Standing Dead	700	227.35
	Cut & Leave	4	8.62
	Cut & Remove	41	467.70
	Hazard Tree Mitigation	15	44.48
Total				8955.38

).

The standing dead timber presented problems within rights-of-way adjacent to public highways and navigable roads within the BNF, necessitating hazard tree mitigation to fell the trees that threatened these areas within the forest. This treatment resulted in 318.9 ha of treatment in 2018 and 44.5 in 2019. Active infestation continued to wane into 2019 but still had approximately 315.7 ha infested. There was an estimated 467.4 ha cut-and-remove performed. The course of the outbreak and shifting treatment protocols show a peak in acreage infested with subsequent peaks in shifting treatments (Figure 5).

4. Discussion

Using multiple image sources, this study tracked an outbreak of SPB and treatment response in the BNF in central Mississippi. Utilizing high-resolution imagery allowed identifying infested groups of trees and the consequent areas of standing dead timber and/or treatments and removals. While manual classification can be utilized to identify areas for mechanical treatment and suppression efforts in the future [30], this study tracked the progression of the outbreak and treatment activities by incorporating satellite imagery into SPB outbreak assessment and risk projection. Assessing high-resolution datasets identified spot heads with fading trees (active infestation). Although this is not new per se [15], routine employment would allow for proactive treatment (as opposed to the post-event assessment done in the present study). Tracking treatments also provides an archive of disturbance response and can be used to plan resource allocations to combat future outbreaks in other areas. Most studies that incorporate imagery into assessing insect outbreaks or disturbances generally are interested primarily in live or impacted trees. Many studies have used image classification detect infestation. In Austria [31], high-resolution imagery was employed to classify dead, early infestation, and healthy trees. Image indices are often used globally to detect outbreaks of similar magnitudes to that presently studied with an outbreak of spruce bark beetle being tracked in Italy using such a methodology [32]. In the western U.S., imagery is used to classify damage severity from infestation but not treatments [33]. Future studies should consider working to classify treatments so that methods of rapid classification of imagery can be leveraged to more efficiently prioritize areas of impact and treatment. Image indices and ancillary weather or climate data may also be useful to enhance spectral responses to SPB infestation and project spread in upcoming growing seasons [34]. Regular acquisition of high-resolution imagery like WorldView and automated classification processes could be developed to support the monitoring process.

The reproductive cycle of SPB is an important factor in tracking and treating an outbreak. There are typically 7–9 overlapping generations per year in the southern United States with parent adult beetles reemerging 10–14 days after an attack, and egg laying, before moving to additional trees. Brood adults can emerge in 28 days during the summer (and 56 or more days in winter) and continue to propagate [see 5]. This ability to reproduce and navigate in stands as dense as those found in the BNF make it difficult to contain an infestation and force continuing efforts to treat SPB spots. Other studies treat insect outbreaks similarly, attempting to remove down and infested trees to prevent further spread [35] with removal of downed trees, especially near roads serving as a check on continued spread [36]. Cut-and-remove operations require a sales contract on National Forests while cut-and-leave operations do not and can thus be employed faster in affected areas to mitigate further spread with the added benefit of minimizing spread of SPB outside of the containment zone [28]. Future research should consider a more dense time series of imagery [37] to detect fresh attacks, perhaps monthly. This would allow for tracking the effectiveness in treatments for preventing additional attacks and may necessitate the use of automated classification [32,33] combined with manually classifying and tracking treatments if they cannot be classified and separated from bare ground/low vegetation.

Significant differences were found based upon count data of SPB spots and the status of those spots. Among years, 2018 was the more dynamic for detecting active, exhausted (where trees were vacated by beetles and standing dead), and professionally treated sites. The more active status among the SPB spots were those that had yet to be treated by forest operations. Hazard tree mitigation is usually conducted on small strips of rights-of-way on dead deteriorating trees, and would generally not produce the needed volume to contract with local wood-using mills. Cut-and-leave would be more common for small-sized spots, again too small to attract private harvest operators, or those found well into the interior where access is limited. Cut-and-remove can be performed on larger areas, but with added delays and associated costs due to sale preparation, environmental and cultural resource analysis, and harvest monitoring [38]. In addition to these added costs, the wood quickly loses value to mills in this region due to deterioration from high heat and humidity that is exacerbated by moisture and specific gravity losses resulting from the symbiotic relationship the SPB has with a blue-stain fungus (Ophiostoma minus (Hedgc.) Syd. & P. Syd.) [39]. In much of the BNF, management activities such as thinning, have not been regularly conducted. This provides a ready supply of wood for attack from emerging broods. If non-intervention in public forests is to be carried out effectively, some buffered management zone and low-impact salvage (cut-and-remove) will need to be considered and has been recommended as a means of maintaining forests with conservation vs. production aims [40]. In the BNF, however, the USFS faces private lands within the proclamation boundary as well as adjacent private lands that could be managed for economic production (i.e., timber) which may necessitate more invasive means of mitigation of beetle spread.

Image sources used in this study vary slightly but were all considered high-resolution images. While overlapping imagery from multiple systems within a year were not obtained for comparison, it would be an interesting exercise to determine if multiple high-resolution sensors vary in their ability to detect differences in spectral characteristics that might connote SPB impacts. The differences in a 2 m resolution WorldView image and approximately 0.25 m aerial imagery may not be significant enough to warrant yearly aerial flights. Still, WorldView or a similar platform might guide aerial reconnaissance from a plane or UAV [41]. WorldView and/or UAV supplemental flights would also be useful in determining treatment success/completions. Some discrepancies occur between contractor payments for treatment and detection with imagery owing to the acquisition dates of the imagery relative to when treatments occurred. This would help with validating treatment and timber sales within the boundaries of the national forests.

This outbreak followed another, shorter duration, outbreak in 2012. Then, unthinned, unburned, stands on the Homochitto NF were more susceptible to SPB than their managed counterparts [6]. We did not assess stand-level/management differences in infested vs. non-infested areas. The geographic distribution of infested areas on the BNF, though, underscored that the more recent (2016–2019) outbreak originated in the same area as the 2012 outbreak and expanded outward. This is similar to an outbreak in southwestern France [42] that found downed log piles to be related to facilitating outbreak. While that study did not assess methods for combating the outbreak, it does point to existing downed material as a culprit. In the BNF, if the outbreak sprang from downed dead trees, there existed a ready supply of healthy trees for the beetles to attack. This points to a need for active management in natural and previously planted stands. The forest has experienced minimal, if any, harvest and has old, overstocked stands that allowed for the ready movement of SPB.

It is obvious and easy to recommend that public lands be managed to minimize risk of forest health outbreaks. However, forest operations like those prescribed to mitigate southern pine beetle typically occur in an open and competitive setting. Here, buyers strategically offer timber bids for stumpage and subsequently expect a return on their investment. Sales restrictions and administrative complexities have been identified as negative factors impacting national forest timber sales’ values versus those on private lands, and limits the timeliness with which cut and remove operations can be implemented against dynamic SPB infestations [43,44]. No-bid sales and those that are not sold occur with varying levels of frequency [45,46,47], although some sales can still be executed through later negotiation [46]. Those that are not must wait to be re-offered at a future date. The absence of competition could foster timber sales inefficiencies [44] and place no-bid stands susceptible to SPB attack at elevated risk. The potential spread into neighboring private forests could impact timber production and/or investment over many years.

The risk of SPB activities rising to outbreak levels in both Mississippi and throughout the southeast looks to increase through 2100 based on temperature and precipitation modelling [3]. High-resolution imagery and manual classification will provide useful information if that eventuality occurs. The tradeoff would be exhausting the resource of time involved in classification. Pairing this imagery with automated classification may provide a more efficient means of directing mitigation measures to minimize spread if classification and/or machine learning methodologies can differentiate treatments to SPB impacts similar to those employed on classifying beetle infestation [16,17,18,19,20,32,33].

5. Conclusions

Our study demonstrated the utility of high-resolution imagery in tracking and quantifying a forest disturbance event. Delineating areas treated subsequent to the onset of the outbreak in 2015 aligns well with the areas contracted for treatment from 2016–2019. High-resolution imagery can play a significant role in the early detection of insect disturbance in large dense forests. Additional research will focus on spectral assessment of detecting active infestations which could lead to automated processes for early indication of SPB activity. This would allow for better resource allocation and could potentially mitigate further impacts/losses due to SPB. Ultimately, management activities (thinning, fire, etc.) need to be planned and executed in these dense stands as a mitigation effort against future outbreaks. The present results suggest that acquiring high-resolution satellite data may reduce reliance on aerial flight and speed detection time and efficiency related to image acquisition and processing. The efficiency and cost of such imagery vs. aerial flight and field-based assessments warrants study. Satellite imagery also standardizes spectral and spatial acquisition which aids in automating detection. In the absence of regular in situ visits and intensive management activities, incorporating imagery such as WorldView enables rapid detection and tracking of events within a forest.