Consideration of anthropogenic factors in boreal forest fire regime changes during rapid socio-economic development: case study of forestry districts with increasing burnt area in the Sakha Republic, Russia

This paper presents an original approach to characterizing historical fire regimes for regions with limited fire data. Fire variables were derived from satellite datasets and regional fire occurrence statistics. They defined the integral elements of a fire regime such as historical trends, spatiotemporal evolution, fire seasonality and causes. Temporal evolution was investigated based on a regime shift detection method developed by Rodionov while changes in the fire regime were analyzed for statistical significance using the Mann–Kendall trend test and Sen’s slope estimator. A descriptive analysis was performed to assess fire seasonality, causes, and together formed the basis for this methodology. We validated the proposed approach by assessing historical fire activity in the Sakha Republic (Yakutia), which is one of the most fire-prone regions of Russia. The assessment was conducted with data from the period of 1996–2018. We detected increases in historical fire activity as well as thresholds of change in the fire regime. Changes during the analysis period included lengthening of fire season, increased burned area extent, and extension of peak fire period. Overall, significant changes in the fire regime were detected in the regions strongly affected by warming and increasing anthropogenic alteration.


Introduction
The increase of forest fires is potentially a major hazard across forest ecosystems worldwide and has been linked to climate change and increasing human impacts (Zumbrunnen et al 2011). Fire activity is a complex natural phenomenon (Wastl et al 2012), because each fire has its own spatiotemporal characteristics and fire effects. To capture this diversity, ecologists in the early 1960s formulated the term 'fire regime' (Bowman et al 2009) to indicate the characteristics, effects, and variability of fire disturbance patterns (Schoennagel et al 2004). Specifically, 'fire regime' refers to the different time windows, spatial units, fire characteristics, and conditions determining fire occurrence and its impacts (Conedera et al 2009). However, these characterizations may possess limitations.
Human-caused changes can modify fire regimes and drive them outside of their natural historical range (Torres 2013). Many studies (Chuvieco et al 2008, Pechony andShindell 2010) report shifts from natural fire regimes due to human impacts. From this point of view, we advocate for inclusion of socioeconomic aspects into the fire regime characterization, specifically in regions with forest-based industries and developing economies, such as the Sakha Republic in the Russian Far East.
The other challenge in fire regime characterization is where there is a lack of documented historical fire data. This is crucial for the regions having a long history of persistent fire events. Eastern Siberia and the Far East of Russia possess these characteristics as the area lacks adequate data on past fire occurrences (Vivchar 2011). The region is quite critical as the Republic of Sakha (Yakutia) is one of the areas with the highest annual burned area. In certain periods, the Republic alone accounts for up to 75%-93% of burning forest fires based in the Russian Far East (Rosstat 2019).
At the same time, the Sakha Republic was undergoing an economic transformation through the State economic development policy. As a result, since 1996 the Republic has experienced rapid industrialization, primarily in mining exploration, exploitation, and refining, along with large agro-industries (Investment guide book of the Sakha Republic (Yakutia) 2011, InvestYakutia 2019, Nikiforova 2019). Since 1996 the new wave of industrialization in the Sakha Republic is due in part to the recent changes in forestry legislation oriented to increasing the economic use of forested lands. It brought forestry legislation, which was supposed to protect forests, into accord with civil legislation, managing businesses and services; converted some forested lands into lands for industrial use; and granted easy access and long-term leases of forested lands to large industries (Torniainen and Saastamoinen 2007, Karjalainen et al 2013). The highest number of incidents of forest land conversion are those for exploration and mining operations, particularly in the Far East of the country, of which the Sakha Republic is the largest part. As a result, economic uses of forest and forested lands in the Republic have risen rapidly leading to further deforestation, which increases heat, dryness, and risk of forest fires. The yearly changes in the law that resulted in increased economic activity since 1996 are shown in the Russian original of the Forest code (see articles 7-9, 12, 21, 24, 43, 70-71, 79-80, 87, 91 of the Forest Code of Russia, Russian Federation 2006) and its English version (Russian Federation, in English 2006). These changes, coupled with other anthropogenic interventions such as deforestation, can alter forests and can cause changes in natural fire regimes. The deforestation can make the regional climate even drier and hotter (Bonan 2008, Lejeune et al 2018, and hence more vulnerable to forest fires during peak fire period.
Sakha Republic has a long history of persistent fire events. Published research on fire phenomena began from the 1960s when various scientific surveys were conducted (see Utkin 1965; also the overviews by Scherbakov et al 1979on surveys starting from 1960and by Tsvetkov and Buryak 2014. Later surveys covered a wide range of fire research topics such as the 'development of local fire danger scales in accordance with regional forest composition' (Protopopova and Gabysheva 2016), 'analysis of spatiotemporal distribution of lightning and lightning-ignited forest fires' (Kozlov et al 2014, Tarabukina et al 2018, and 'fire-climate studies' (Hayasaka 2011, Kirillina et al 2016. Despite this extensive research, fire studies in Yakutia can be argued to contain three main limitations. First is a lack of regional fire studies, as most of the research was focused in Central Yakutia which historically was a fire hot spot area, whereas now fires occur on the whole territory of the Republic. Secondly, these studies did not address how the fire regime in the republic might change due to a strong warming trend which began in the 1960s (Skachkov 2005, Fedorov et al 2014. This is because the weather station data for the 96 meteorological stations in the Sakha Republic dates back to 1953 and up to 2013. The record shows a trend of gradual warming, and a cumulative trend of warming of over 1°C since the year 2004. Particularly, from 2011 we see an obvious and continual warming trend (see appendix), occurring at the same time as new industrial projects began, including: • Launching of timber logging and processing facility, Amginsky forestry district (2011-2012).
This warming trend is especially strong in the transitioning spring season (Kirillina and Lobanov 2015), which might affect the length of fire season. The third limitation is the omission of a human-induced or anthropogenic aspect to the fire regime characterization. Regional climate change combined with intensive human alteration of forests and conversion of forested land into lands for industrial use have the potential to shift the regional fire regime.
To build on the previous research and account for the identified limitations, this research analyzes historical trends of fire activity in the Sakha Republic. It assesses trends in the fire activity for Sakha Republic from 1996 to 2018 and assesses the spatiotemporal evolution of fire regime, the underlying causes of changes identified as described by annual burned area. A correlation between the potential impact of anthropogenic factors on fire regimes leading to greater burned area was investigated.

Study area
Our region of interest, the Republic of Sakha (Yakutia), is the biggest region of Russia located on the north-east of the country (figure 1).
The Sakha Republic has a complicated topography and typical 'continental climate' with very low temperatures between −50°C and −60°C in the winter, in turn summer temperatures can be as high as 29°C-38°C (Sofronova et al 2014). Also, the climate is very dry. The amount of annual precipitation, especially in the central districts, is equal to the semi-arid territories with frequent droughts during summer (Korzhuev 1965). Precipitation in most of the territory is as low as 150-250 mm which along with frequent drought create favorable fuel conditions for fires (Tomshin and Solovyev 2018).
In recent times, the Sakha Republic, our study area, has had the largest area of fire disturbance in the Russian Far East. Chen et al (Chen et al 2014), using analyses of Landsat data, showed that Eastern Siberia Forests occupy 82% of the territory of the republic, or 254.7 million hectares (Minprirody 2018). The territory of the Republic consists of four vegetation zones: boreal taiga forests, tundra, forest-tundra and arctic deserts. The arctic vegetation occupies 26%, and the boreal-74%, where our research is focused on. Open coniferous stands dominate in Yakutian forests. Larch species (Larix cajanderi, L. gmelinii, L. sibirica, L. czekanowskii) comprise 90.5%, pine (Pinus sylvestris) 7.3% and Siberian spruce (Picea obovata) 0.6% of the tree cover (Protopopova and Gabysheva 2016).
In Russia, forest inventories identify all forests by five classes of fire hazard based on landscape/ecosystem indicators from Class 1 (highest fire danger), to Class 5, where fires occur only under extremely unfavorable conditions (Shvidenko and Goldammer 2001). The average natural fire hazard risk class of Yakutian forests is reported as 2.9, which is about the median fire hazard risk in the Russian classification of the natural fire hazard risk. However, forests with the highest fire risk (1-2 class) account for 33.7% of forests (Russian Federation 2018).

Data
To assess historical fire occurrence in the Sakha Republic we used fire records from 1996 to 2018, obtained from the Sukachev Institute of Forest wildfire database (general details of this database can be found in Ponomarev et al 2016); also those who are interested can access it through sending requests to the Database developer, the second author of the present article (Shvetsov, E.G., eugeneshvetsov11@yandex.ru). Fire causes and seasonality data were obtained from annual reports of the Yakutian branch of the Aerial Forest Protection Service of Russia which are based on satellite (ISDM-Rosleskhoz, available online on https://nffc. aviales.ru/main_pages/index.shtml, website requires authorization), aerial and ground observations of fires (collected by Kirillina, K. in August 2017). There are 19 forestry districts, which are the main territorial unit of management in the field of use, protection, and reproduction of forested areas in the country. For analysis we chose five forestry districts having high fire activity including Amginsky, Gorny, Khangalassky (in central Sakha Republic), Verkhnevilyuisky and Vilyuisky (in western Sakha Republic).
As supporting research data, we used: • GIS data for preparation of the regional fire maps from the Russian Open GIS Portal 'GIS-Lab' (http://gis-lab.info), in Russian language.
• Information on economic use of forests and forested land collected from regional forest policies and legislation, Government reports on environmental protection, and plans and strategies on socioeconomic development of the Sakha Republic.

Methodology
Fire data obtained from the Sukachev Institute of Forest wildfire database (satellite burned area estimates and number of fires) and the Yakutian Branch of Aerial Forest Protection Service of Russia (fire cause and seasonality) were aggregated to regional (Sakha Republic) and local (for selected forestry districts) levels and were compiled into the fire variables for analysis, including: • Number of fires-total number of forest fires, ignited by both lightning and anthropogenic factors.
• Burned area-total area affected by forest fires, including areas covered by forests and forested land.
• Number of anthropogenic fires-all fires, ignited by both human and industry.
• Fire season-the period of the year when fires occurred, in days.
• Peak fire month-month with the highest number of registered fires/largest extent of burned area.
First, to analyze the historical fire activity we derived historical trends of annual number of fires and burned area for Sakha Republic. Standard errors were calculated for all applicable averages.
To make a more robust analysis of burned area extent we used satellite burned area estimates obtained by Shvetsov E.G. for the entire Republic and five selected forestries with increased burned area for 1996-2018 from the Sukachev Institute of Forest wildfire database, which contains active fire detections from NOAA AVHRR (1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007) and TERRA/AQUA MODIS (2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018). This wildfire database was generated using a multistep process including: (1) contextual active-fire detection; (2) creation of fire polygons from adjacent fire pixels and (3) correction of resulting polygons. The detection procedure for AVHRR data was generally similar to the HANDS algorithm (Fraser et al 2000). The processing chain for MODIS data was based on the approach of Giglio (Giglio et al 2003) considering several adjustments in background characterization and detection probability estimation (Shvetsov 2012). The algorithm used in the Institute of Forest differs mainly by aggregating fire detections into fire polygons and subsequently using a correction procedure. This correction procedure was based on comparison between the burned areas measured using low resolution data (AVHRR/ MODIS) and moderate resolution data (Landsat) for the test sample of fires (~5% of all annually detected fires) and obtaining the linear regression equations for several fire size classes. Then these equations were used to correct fire polygons for the rest of the fires detected using low resolution data (Ponomarev and Shvetsov 2015).
Second, changes in the temporal evolution of fire regime features were investigated using the Rodionov regime shift detection method (Rodionov 2004). Rodionov's method helps to find inflection points indicating regime shift, i.e. years when shifts from one phase to another occurred. To detect increases in fire activity, the total annual burned area data were analyzed for abrupt positive shifts in the mean, where shift is a statistically significant deviation from the mean value of the current regime, in this case that year becomes a potential change point (more in Rodionov and Overland 2005). We used Rodionov's regime shift detection software (Version 6.2), in which the following parameters could be adjusted: probability level, cut-off length, outliers weight parameter (Huber parameter) and subsample size (see relevant parameters in Meyn et al 2010). All parameters were dependent on the original fire data.
To detect long-term trends in selected fire features during the study period , we used the Mann-Kendall trend test. The slopes of trends were estimated by Sen's slope estimator. The trend analysis was completed for burned area only as our analysis was centered on burned area. The trend analysis was carried out using XLSTAT Software. We compared conducted trend analysis using the MAKESENS program for annual time series data developed by the Finnish Meteorological Institute (Salmi et al 2002).
Both Rodionov's regime shift detection method and the Mann-Kendall trend test were chosen due to their robustness to autocorrelation.
Finally, we assessed fire seasonality and causes based on descriptive analysis and literature review. Information on the length of fire season and peak fire months (month with the highest number of registered fires) were obtained from the Yakutian Branch of Aerial Forest Protection Service of Russia. Information on peak fire period depending on when the largest burned area extent was obtained from Sukachev Institute of Forest wildfire database for selected fire seasons (years).

Results and discussion
4.1. Recent fire occurrences and burned area extent, 1996-2018 This 22 year period was chosen because during this period of time the Sakha Republic underwent rapid industrialization, which we are proposing impacted the fire activity. Further, as mentioned above we have satellite data from AVHRR and MODIS for the period starting from 1996.
There were 13620 fires and 39 120 600 ha burned area from 1996, the year from which we had satellite data, to 2016 in the Sakha Republic; with average annual values being 592±85.92 fires and 1700 895.7±359 792.57 ha burned ( figure 2).
These findings represent a departure from the results of previous historical fire studies for Sakha Republic (Nikolaev et al 2012): there is a simultaneous growth in the number of fires, which is used for fire seasonality analysis below, and in the extent of annual burned area.

Interannual variability and trends in fire regime
Burned area on both regional (Sakha Republic) and local (five selected forestry districts) levels became the subject for this analysis.
The results of the trend analysis of burned area are presented in table 1.
Significant increases in the burned area were detected for the Sakha Republic overall, and three out of five researched forestry districts, which were located in the central (Gorny) and western parts of the Republic. These changes might be explained by the existence of a strong warming trend and increasing anthropogenic impact. From year 2000, the Sakha Republic became much more rapidly industrialized. The central part has been impacted by agro-industry and population growth. The western part has been impacted by rapid industrial development.
Burned area on a regional scale showed significant abrupt increase in 2012 (figure 3). One forestry district (Verkhnevilyuisky, Western part of the Republic) out of five showed abrupt significant increase of burned area observed starting from 2013 ( figure 4). In the following we aim to show a correlation between anthropogenic factors involving new, rapid developments in mining, refining, and agroindustry.

Changes in fire seasonality and changes in burned area
Duration of fire season in the Sakha Republic is a period between the first and last registered fires in the protected forest areas. Historically the start of the fire season was usually the first week of May and the fire season lasted until mid-September. The length of fire season varies among forestry districts; and might vary depending on the weather conditions of a particular year.
The seasonal distribution of fire activity also varies. Fires at the beginning and the end of the fire season usually have shorter duration, with smaller burned area due to more moderate fuels and fire weather conditions for large and long-lasting fires. The peak of fire occurrence falls in July-early August (Solovyev et al 2009).
Analysis of historical fire records showed that the average duration of the fire season shifts to a longer fire season, from 115±3.81 d from 2005 up to 128±2.78 d starting from 2011 (summarized in table 2). Moreover, in the 2010s the extension of peak fire period was detected. Previously, the month with the peak (maximum) number of fires had been July, but in the latest decade (since 2010) the peak fire period was recorded as being throughout a three-month period from May to July. In the 2000s, peak fires were recorded only in July (besides 2009); however, since    Table 3 presents the analysis of peak fire period based on the largest extent of burned area.
From table 3 we can see a marked increase in burned area almost steadily starting from 2011. Also, the fire season from 2009/2010 starts in April and lasts until October. Notably there were also significant amounts of burned area in the new onset month of April and the ending month of October, which impacts not only the length of fire season, but also increases total burned area extent. July, however, was still mostly the month with the largest burned area. Also, there exists a second (post-2016) rapid and intensive increase of cumulative burned area during the most recent fire seasons (2017 and 2018).

Fire causes
According to the regional fire statistics, during the study period 1996-2016 (the latest year in which the binary breakdown into lightning-anthropogenic was used), the majority of forest fires in the Sakha Republic had an anthropogenic character-47.5%. The second main cause of fires was lightning-43.3% ( figure 5).
As in the other boreal regions of the world, the lightning-ignited fires in the Sakha Republic historically have been responsible for more burned area than anthropogenic fires. However, this situation started to change in recent decades. In our case, the determination of causality is complicated by the fact that the regional forestry legislation does not require the on-site observation or suppression of fires if they are located in remote areas and do not make a threat to human settlements (Government Degree N177 from 25.05.2016 'On the approval of forest fire control zones in the territory of the forest fund of the Sakha Republic Yakutia'). Thus, the causes of 'remote fires' are not thoroughly assessed; and by default, lightning is designated as their main possible cause, especially if they occur in the areas with high lightning activity (Solovyev et al 2009).
Lightning fires are typically most likely to occur during June-August following the seasonal dynamics of lightning activity (Solovyev et al 2010). Spring (May fires) and Fall (September fires) fires were typically called anthropogenic. Spring fires were thought mainly occur due to traditional agricultural burning practice needed for renewal of agricultural fields. This procedure was part of Yakutian cattle breeding and  (2010,2014,2015) Note. Data source: The annual reports of the Aerial Forest Protection Service of Russia in the Sakha Republic based on satellite, aerial and ground fire observation data, available from the corresponding author by email, in English. grazing culture. Fall fires were usually blamed on local livelihood and recreational activities such as hunting, picnicking, berry and mushroom picking (Protopopova and Gabysheva 2015). Within the historical prevalence of lightning fires on the territory of Sakha Republic, since the early 2000s the ratio of anthropogenic fires has begun to increase, as shown in figure 5. Further, from the mid-2000s anthropogenic fires started to prevail. To test this assertion, we performed a test for a statistically significant difference in cause attribution for 2000-2016 study period, which was found to be statistically significant (P=0.03). At this point we might suggest that new sources of anthropogenic fires may be attributed not only to farms, hunting and recreational areas, but also to increasingly populated and industrialized areas (Central and Western forestry districts). The same goes with new industrial sites and development areas. This was shown in figure 6, which gives the spatial correlation between the location of industrial and agricultural areas and fire activity. Figure 6 shows that the distribution of burned area extent is higher in developed industrial and agricultural areas. The growth of burned area extent also can be observed near new industrial sites. The high burned area extent in agro-industrial areas such as most of the forestry districts in Central Yakutia (Amginsky, Gorny and Khangalassky) can be explained by the vulnerability of the dry climate of this region to the agro-industrial practice of large-scale burning, especially before new planting, which occurs at the beginning of fire season in May. In regards to the percentage of anthropogenic fires, the increase was detected in five forestry districts and mostly attributed to the zones of new industrial and agro-industrial development, which might be related to involvement of more forested areas into industrial, agro-industrial and infrastructural development in the Sakha Republic.

Conclusion
We characterize historical fire regimes of the Sakha Republic, Russia-an area with limited fire data with Map key: Cross-hatched areas-agricultural districts, bold striped areas-industrial district, striped areas-new industrial site, and red circles-burned area, in (× 1000 ha). Visualization of total burned area from raw data compiled by Shvetsov, E. G., available from Shvetsov, E.G. in English, by email. statistical analysis of spatiotemporal changes of historical fire trends by supplementing regional fire statistics with satellite burned area estimates.
There is some evidence of a long-term increase of fire activity during the period 1996 to 2018 in the Sakha Republic, illustrated by figure 2 (to 2018) and 4. The peak fire occurrence period is now extending across a three-month period (May to July).
5. Were registered positive shifts in the temporal evolution of burned area after 2000, on both regional and local (forestry district) levels.
6. In this compilation of statistics (regional fire data and satellite burned area data) the ratio of anthropogenic fires seems to be increased, especially in districts with developed industry/agriculture and new industrial sites/development zones.
We believe that it should be possible to use the proposed approach for assessment of historical fire activity in other regions with limited data on fire, but where knowledge about risks of fire is vitally needed.

Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request, specifically, datasets on burned area and fire seasonality available from EG Shvetsov at the Sukachev Institute of Forest (eugeneshvetsov11@yandex.ru).

Appendix
Air temperature changes for study area (five selected forestry districts).