Two decades of active fire data in Huascaran Biosphere Reserve (Peru): patterns and future challenges to prevent wildfires.


 Background

The Andes are recognized as center of biocultural and biological diversity that provide multiple ecosystem services threatened by different anthropogenic forces. Here, fires are widely used in agriculture and other traditional activities with the potential to become larger-scale, unmanageable “wildfires” and contribute to the degradation of both ecosystems and human livelihoods. In this study, we assess the temporal and spatial pattern of MODIS and VIIRS Active Fire Data (AFD) and their relationships to 9 environmental conditions and human activities across the three zones of the Huascaran Biosphere Reserve (HBR), Ancash, Peru between 2001 and 2020.
Results

Xeric climatic conditions seem to drive the temporal AFD pattern across and within years in the HBR. Across years, an oscillating increment of active fire alerts was evident, with years 2016 and especially 2020 displaying the highest AFD records of the two decades. Both correspond to years with the lowest mean annual precipitation. Seasonally, AFD also peaked at the end of the dry season (August and September), but AFD continue until late November of 2016 and 2020. The amount of people on the other hand, measured in terms of population and number of rural villages, is the main driver of where AFA occur, increasing along the limits of the Huascaran National Park (HNP) and the Buffer Zone (BZ) and in weakly controlled areas within the park.
Conclusions

We conclude that a combination of both, xeric conditions, and human pressures, seems to interact and produce high numbers of AFD along key Andean ecosystems like Puna grasslands and shrublands that are not ecologically adapted to this type of pressures. Land transformation and a limited capacity of control over key areas within the HNP seem to play an important role in where and when wildfires occur. Finally, although 2020 was an arid year, the covid-19 pandemic would have exacerbated the number of AFDs that occurred in the area, demonstrating that the effects of the pandemic go beyond health problems.

for water, land, food and other resources from local people bring socio-environmental con icts, that, if not managed and well understood, often lead to degradation of both ecosystems and human livelihoods Fire is widely used in the Andes for agriculture and other traditional activities (Borsdorf and Stadel 2015a). In agriculture it is used to clear land in a process of burning and rotating land plots to expand or improve land productivity also known as slash-and-burn (Kleinman et al. 1995). It is also used for eliminating secondary vegetation and littler, renewing pastures for livestock, and sometimes for largescale deforestation for new farming areas (Beck et al. 2008). Traditional uses include extensive burnings in June 24th as part of the festivities of Inti Raymi and the Campesino's day while other traditions aimed to "call" the rain during the end of the dry season (August -September). Yet despite their controlled uses, res frequently escape and become larger-scale, unmanageable "wild res" that cause Puna pastures, shrublands and native forest loss and degradation with strong impacts on the Andean biodiversity southeast Australia (Mariani et al. 2016). Hence, these environmental and human factors do not act independently and rather constantly interact, making hard to forecast wild res events at different scales.
As a result, uncontrolled re in areas where they are not part of the natural regime is a real threat for multiple ecosystems and biodiversity (Borrelli et al. 2015; MolinaKnapp 2019). In the Andes, natural res are uncommon as lightings, which usually ignited natural res, mostly occur during the wet seasons when most plants, that can serve as fuel, are wet. In response, different alternatives have been proposed to reduce their impact especially in areas of high biodiversity (Young 1997). One alternative was the development of several ecotourism or nature-base tourism initiatives with the intention of providing a more environmentally friendly economic options to local people, providing a "guide" of good practices to reduce their impact on nature. Yet, results are ambiguous and not conclusive as re used by local people continued independently of these economic activities given that agriculture is a fundamental part of rural livelihoods (Young 1997). A more restrictive alternative is the establishment of Natural Protected Areas (NPA) where wild res are legally forbidden and penalized but in practice, some of these efforts are di cult to achieve. Among them, National Parks have the highest level of restriction and therefore, expected to minimize the occurrence in wild res within their boundaries. However, although this restrictive approach has shown important results in terms of biodiversity conservation, it has been criticized by not incorporating local humans needs and, in some cases, by directly affecting them by, for examples, forced displacements (Kabra 2018). The Man and the Biosphere (MAB) Program from UNESCO, complement the NPS' objective by integrating people's role in the conservation of areas of high biodiversity, aiming to enhance the relationship between people and their environments by using different interdisciplinary approaches (UNESCO 2021). This conservation efforts within Biosphere Reserves are normally structured over three areas: (i) the "Core Area (CA)" that always is a National Park and protects biodiversity, (ii) a "Buffer Zone (BZ)" that surround the core area and local people used for compatible activities and (iii) the "Transition Zone (TZ)" that surrounds the buffer zone and where most local cities are located, fostering socio-cultural and ecologically sustainable economic activities (UNESCO 2021). Today, the World Network of Biosphere Reserves spans over 129 countries with around 257 million of local people living in there.
In this study, we evaluate the relationships of res, measured in terms of the number of Active Fire Data (AFD), and different environmental conditions and human activities in the three zones of the Huascaran Biosphere Reserve (HBR), Department of Ancash, Peru between 2001 and 2020. Speci cally, we aim to (i) describe the temporal and spatial pattern of AFD, (ii) assess their relationships with climatic and human factors, (iii) identifying the most affected ecosystems and (iv) discussing the effectiveness of three conservation initiatives: ecotourism, the establishment of Huascaran National Park (HNP), and the Huascaran Biosphere Reserve (HBR). We hypothesized that while environmental conditions would drive the temporal pattern of AFD across years and seasons, human activities would be the main driver of where AFD occur. We also predict a greater effectivity of Huascaran National Park (HNP) as a tool to conserve one of the main sites of High Andean biodiversity.

Study area
We carried out the study in the Huascaran Biosphere Reserve (HBR) (Latitude: -9.439056°, Longitude: -77.347048°), located in the Department of Ancash, Peru ( Fig. 1a-d). The 340,000 ha of the core zone (CZ) is protected by Huascaran National Park (HNP) since 1975 (Fig. 1b), covering most of Cordillera Blanca which is considered the highest tropical mountain range in the world (2,400-6,768 m asl), and was declared a World Heritage Site by UNESCO in 1985 (Shoobridge 2005; SERNANP 2017). Cordillera Blanca is anked by two valleys located in the buffer and transition zones ( Fig. 1c-d), the Callejón de Huaylas (east) and Los Conchucos (west), where ~ 728,000 inhabitants depend on the ecosystem services that glaciers, forests, grasslands, peatlands, and other high Andean ecosystems provide (Shoobridge 2005).
Main cities, like Huaraz (118,836 hab. in 2017), are in the transition zone (Fig. 1d). This area also includes part of Cordillera Negra, an ice-free mountain range running parallel to Cordillera Blanca and the coast for about 180 km and is dominated by a mixture of intermontane shrubby forests and Puna grasslands. HBR is one of the six Biosphere Reserve that currently exist in Peru and has the main objective of conserving the high Andean biodiversity, landscape beauty and key ecosystem services, such as water, to ~ 2,600 populated areas (UNESCO 2021).
The climate across HBR varies with elevation and is locally modi ed by topography. The mean annual temperature decreases with elevation, but is relatively stable within years at each elevation, albeit with "helada" events (drastic drops in temperature for short periods of the day, usually before sunrise) between May and August ( Fig. 1e) (Kaser and Georges 1997; SERNANP 2017). In contrast, mean annual precipitation (640-1400 mm) increases with elevation and shows strong seasonality, decreasing sharply between June and August ( Fig. 1f) (Kaser and Georges 1997). As a result, primary productivity in different plant communities, such as wetlands, scrublands, and puna grasslands, show strong seasonal variation, although forested areas, being evergreen, remain without apparent changes in their structure or primary productivity (Morales et al. 2018). res. Both MODIS and VIIRS detect res of near real-time (NRT) at 1 km and 375 m resolution, respectively and in both cases the recorded coordinate represents the center of either ~ 1 km or 375 m pixel that contain one or more res. Therefore, the coordinates do not necessary represent that exact location of the re but rather, that at least one active re event is present in the extent (LANCE 2021).
Multiple points are expected to represent large res events. Compared with MODIS, VIIRS allows to detect smaller aming res and provide a less underestimated depiction of re presence. For each re alert, latitude and longitude coordinates, acquisition date, time, and elevation (m asl) were recorded. Because the same re event could have been detected by both satellites, we excluded MODIS records that overlapped VIIRS in date and location. Each event was then classi ed according to its location within the core (n = 734), buffer (n = 1,195) or transition (n = 4,275) zone. Processed data was standardized by area (number of AFD /10,000 hectares) to account for differences in extension of each zone, excluding 170,816 ha within the CZ that correspond to in ammable areas like glaciers, lakes or rocks.

Environmental and anthropogenic covariates
We use 9 environmental covariates measured within 1-km diameter from each AFD and from a set sample of 5,000 random points where none re alerts were reported. This environmental covariables included geographic, land productivity and climatic information. We obtained elevation (m) and land surface slope (degrees) from the 30-m resolution Digital Elevation Model (DEM) provided by the Shuttle Radar Topography Mission (Farr et al. 2007). Net Primary Productivity (NPP), understood as the rate at which energy is stored as biomass by autotrophs through, for example, photosynthesis in plants (Waide et al. 1999), and in our case re ects the amount of biomass on ecosystems, was derivate from the average interannual Normalized Difference Vegetation Index -NDVI expressed as: 1 where Band 5 and Band 4 correspond to the near-infrared and visible red wavelengths Landsat 8 images of 30-m resolution between 2013 and 2020. We used Google Earth Engine (GEE) to analyze and generate this data, ltering those images with more than 25% of cloud cover (Gorelick et al. 2017). We also use the average daily solar irradiation (Wh/m2/day) of the dry season (June -August) between 1986-2010,

Results
Active Fire Alert pattern on time and space A total of 6182 AFD were recorded across the three zones of the HBR between 2001 and 2020 ( Table 1) Seasonally, AFD principally occurred during the months of August and September which correspond to the end of the dry season (Fig. 2b). Yet, during 2016 and 2020, res started before (June -July) and continue until late November (Fig. 3). The yearly cumulative patterns also show that year 2020 exhibited 3.6, 4.6 and 3.6 times more active re alerts than most previous years for the core, buffer, and transition zones respectively (Fig. 3). Spatially, AFD were related with the distribution of people (Fig. 4). Within the CZ, res were less frequent and concentrated to speci c areas of the norther (Quebrada Cedros), center (Quebradas Honda and Potaca) and south (Quebrada Yanayacu and Pumashimi) of Cordillera Blanca. The highest concentration occurred along the CZ-BZ boundary where many people live. Indeed, this pattern was highly correlated to the number of people and rural villages present across the three zones of the reserve (Pearson's r coe cient: 0.99, p < 0.05; Figure S1). According to the last two national surveys of INEI, 592 and 219 people lived inside the national park (CZ) by 2007 and 2017 respectively while around 40 and 33 thousand people lived in the BZ at these times (Table 1).

Fire and environmental and human relationships
Active re alerts peaked around 2,800-3,000 m asl to then decreased with elevation (Fig. 5a). This peak roughly coincides with the buffer zone of the HBR whereas a second and a third smaller peaks Along the gradient of primary productivity (NDVI), re alerts were maximum at NDVI values between 0.1-0.2 which correspond to areas mainly dominated by Puna grasslands. In regions with NDVI values between 0.2-0.5, that correspond to bottom valleys with scrubby and forested areas, the number of AFD were slightly lower but relatively constant (Fig. 5b). This trend was corroborated after quanti ed the number of AFD among landcover types, were Puna grassland, shrublands and agricultural elds accounted for 45.92%, 30.27% and 18.2% of them (Table 2). Peatlands 0 (0) 0 (0) 0 (0) 1 (0.14) 1 (0.14) Forest Plantations 0 (0) 0 (0) 0 (0) 1 (0.14) 1 (0.14)  As temperature decrease with elevation, res were also related with warmer areas where the Mean Annual Temperature (MAT) range between 12-17 C (Fig. 4c). Its relationship with precipitation (MAP) on the other hand, showed two peaks: one with dryer areas where MAP ranges between 250-600 mm and are in the BZ and TZ in the norther part of Callejon de Huaylas and second, related with wetter areas that ranges between 750-850 mm and are located along the glacial valleys and in the central and southern part of Callejon de los Conchucos ( Supplementary Information Fig. S2).
The relationship of AFD against solar radiation and terrain slope suggests that res are more frequent in steep areas (> 30 degrees) independently of the amount of solar radiation. Indeed, the range of solar radiation (~ 4000-5000 Wh/m2/day) where the number of AFD decreases correspond to atter areas of the glacier valleys (Fig. 5e-f) whereas the northern and southern slopes usually receive higher (> 5000 Wh/m2/day) or lower amounts (< 4000 Wh/m2/day) respectively. Finally, AFD remained highly constant along most part of the HFP index (Fig. 5g) whereas monthly re alerts decreased with monthly precipitation (Fig. 5h). On the other hand, the number of AFD was negatively related to the cumulative yearly precipitation (Pearson's r coe cient: -0.55; p = 0.013).

Discussion
We evaluate the extend to which active res alert in the Huascaran Biosphere Reserve were related to environmental and human population features between 2001 and 2020. The number of re alerts constantly increased during last years, especially in the buffer zone were most people uses re in agriculture; and reached its maximum in 2020 during the rst wave of the global pandemic. A combination of both, xeric conditions, and human pressures, seem to exacerbate re alerts occurrence along key Andean ecosystems like Puna grasslands and shrublands that are not adapted to this type of pressures. Moreover, land transformation and a limited capacity of control over key extensive areas within the National Park seem to play an important role in where and when res occur. Hence, the effectivity of different conservation initiatives will depend on understanding how climate shapes long term and seasonal xeric condition -including the distribution of biomass that serve as fuel for re -and how is this related to different human activities principally associated to agricultural practices.

Climate and human determinants of res in Cordillera Blanca
While climate can provide suitable condition to wild re to prosper, human activities usually trigger them, While climate drives the temporal pattern of res, the spatial distribution of res seems to be driven by people. In our study site, AFD was highly related to the number of people living on each zone of the HBR (Table 2; Figure S1). We believe that the use of re in agriculture and in minor proportion, accidental events caused by tourist, are the main underlying factors behind this relationship.
At the end of the dry season, people located in rural areas start to feel the effects of water scarcity and therefore, the traditionally use of re for land preparation for a new agricultural year start (Moran et al. 2006; Manta et al. 2018). This practice is widely used in the Andes and other agricultural regions and usually involves the burn of stubble -the set of remains of stems and leaves that remain on the ground after cutting a crop called "rastrojos" in Spanish-. In many opportunities, however, burning rastrojos can get out of control if it is not well managed, causing severe damages to native forest and other ecosystems. Plotting our results on a map, clearly shows that most AFD within the HNP occur close to the BZ and around the entrance of densely populated glacial valleys (Figs. 1 and 4). According to the HNP o ce, most wild res recorded within the HNP started with this type of practice in the BZ, where agricultural activities are a fundamental part of the livelihoods and culture of local people. But what is in the buffer zone that can make "rastrojos" so di cult to control? The answer could be Eucaliptus and Pines forest.
It is known that landcover type can de ne re intensity and ammability of an area (Paton et al. 2014).
As in many parts of the Andes, Eucalyptus globulus and Pinus radiata represent key exotic species for rural livelihood whereas also reduces human pressure on native forest for logging (Luzar 2007). For several decades, both were extensively planted in the surroundings of the Cordillera Blanca, becoming part of the landscapes of the buffer and transition zones (Byers 2000). However, both have important characteristics that make them good candidates to prosper in red conditions (Mutch 1970).
Eucalyptus forests are considered " re-dependent" trees, due to a high natural ammability that allow them to live in balance with res (Hodgson 1968). One characteristic that increase its ammability is the presence of oils in the leaves, which have low igniting point and required less oxygen that other plants to burn (Webb 1968). Another property is the low rate of decomposition of the litter, caused by the volatility of the oils in the leaves, which produces an accumulation of litter on the ground and leaves it exposed to the periodicity of res (Agee et al. 1973;Mutch 1970). For Pines on the other hand, Franzese and Raffaele (2017), indicate that in addition to have speci c survival adaptations to re, such as thick bark and selfpruning capacity (Higgins and Richardson 1998), there are two very important factors that would explain the synergistic relationship between pines and re: (1) serotiny, which is the prolonged storage of seeds in the canopy and is closely related to the responses of trees to recurrent res (Keeley, 2012;Lamont et al. 1991), and (2) the ability of seedling to invade and prosper in burned and unburned habitats. Therefore, re seem to be a good facilitator of Pinus radiata success in habitats affected by res, as recruitment would be carried out quicker in burned areas.
Under technical supervision, different practices have been carried out to manage these types of forest, including continuous thinning and prescribed re in humid conditions to control or eliminate the amount of fuel (litter) accumulated in the understory (Agee et al. 1973). Additionally, the use of controlled re is also vital, so destructive res do not occur (Mutch 1970). However, most local people that practice agriculture in HBR do not necessarily follow these practices but conduct the burn of restrojos quite often. We believe that the monitoring and manage of res in the buffer and transition zones are fundamental to protect the native biodiversity of HNP. However, this may require set of well planed activities that include environmental education, logistic and personal preparation, and facilities for the ecological restoration of affected areas.
How are wild res managed in the HBR?
Wild res in the HBR should be managed in a hierarchical way by different institutions and key actors responsible for speci c tasks, including prevention, the response, and restauration (Vallejo and Alloza 2014). However, in practice, the burden falls mainly on the HNP, the park rangers and local people who somehow respond to these events in a rather voluntary way full of conviction. The National Service of Protected Natural Areas (SERNANP) has an organic strategy for the control of wild res framed in three components: (1) prospective that seeks to understand the causes of these res, (2) corrective that seeks to prevent them and (3) reactive that it is implemented in response to a wild re event (SERNANP 2018).
On the other hand, the management of re risk control is framed in the prevention, reduction, preparation, response, estimation, rehabilitation, and restoration of the affected spaces and are implemented by both SERNANP and the National Forest and Wildlife Service. (SERFOR). Currently, the PNH administration is the main responsible for strengthening the response capacities to wild res that occur in the CZ of the RBH. These include ongoing training for park rangers and the inclusion of local people through the formation of Participatory Vigilance Committees (Comites de Vigilancia Participativa). In the buffer and transition zone, these activities most be part of the goals of various institutions that include SERFOR, Defensa Civil, and municipal and regional governments (see. Gobierno Regional de Ancash 2020, Plan de prevención y reducción del riesgo de desastres ante incendios forestales de la región Ancash 2020-2023). However, these actions do not seem to be su cient to counteract the misuse of re that often unleashes wild res that are harmful to nature and man. According to Manta (2017), there is still a lack of training for key actors, as well as little equipment, and a high turnover of personnel that together prevent an adequate prevention, response, and management of wild res in the area. We believe that it is necessary to empower good practices in the use of re, mainly associated with livestock and agricultural practices, as well as to be able to differentiate between destructive and useful re through controlled and prescribed burns.
The legislation behind the use of wild re appears to be an aggravating factor for its management in the HBR. Within the HNP, res are prohibited for several reasons including that they are not part of the high Andean natural systems and can be highly destructive of its biodiversity. However, the legal framework that applies outside the HNP is unclear given basic laws that on one hand allow the use of re and on the other hand, laws that contravene and prohibit them (e.g. Law 25268, art.5; Law 29763, DS 018-2015-MINAGRI, DS 021-2015, art. 127,128,137.3 subsection a, b, c; Penal Code of Peru, Art 310). In addition, legal gaps and the lack of updated supranational and national regulations, and the legal bureaucracy mean that environmental processes and charges throughout the entire RBH are led up to 100%, leaving many criminal behaviors that cause terrible wild res unpunished (Manríquez and Rascón, 2020). Given this, the establishment of better criteria and goals of environmental protection in the technical part with clear de nitions for different types of ecosystems is especially important.

Effectivity of the Huascaran National Park and Biosphere Reserve
Our results clearly shown that, after correcting by the amount of ammable area, the CZ has signi cantly less occurrence of AFD that the BZ and the TZ, indicating that the protection role of the HNP is being largely effective. Certainly, there are certain areas constantly affected over the years that we hypothesis is given to logistic di culties to control remote areas where centers of control are absent. La Quebrada Cedros or the sector between the towns of Chacas and Huari, for example, are poorly patrolled areas with di cult access and it is where most of the AFD are concentrated. This contrasts strongly with areas such as Quebrada Llanganuco or Ulta where surveillance is constant, and the number of AFD is almost nonexistent (see Fig. 1 and Fig. 4). The areas that make up the ZA and ZT do not have signi cant differences in the number of AFD, suggesting that the environmental policies for re prevention and control implemented in these two areas are equally ine cient.
Fires during the 2020 The management of wild res was especially di cult during 2020, a year where the COVID-19 pandemic not only caused many human losses but forced strict lockdowns that stopped many economic activities, including tourism. In 2020, many management activities of the HNP had to be suspended or modi ed, including the work with local communities and tourism, which represents the main economic income for many local people. It is certainly possible that this economic paralysis contributed to more people to shift to work the land, having made use of re to clean unused areas for agriculture and livestock.
Nevertheless, our results also show that 2020 had drier conditions that could have contributed to this high re occurrence. Therefore, the pattern of re in 2020 could be the result of a combined effect of dry weather conditions and an increased number of people redirecting their efforts to agricultural activities that involved the use of re. Similar conclusion was made in Colombia, and other part of the globe (Amador-Jimenez et al. 2020).

Conclusions
Our study showed that active res alerts increased during the last two decades in response to both, xeric conditions during droughts years and the increase of human population in the buffer and transition zones of the HNP. After correcting by area, the core zone has signi cantly less number of re alerts than the buffer and transition zone, suggesting a greater effectivity in the control of res by the Huascaran National Park. However, remote, and not well patrolled areas are frequently affected by wild res that usually start in the buffer zone. Overall, Puna grasslands and shrublands were the most affected ecosystems followed by agricultural zones in the buffer and transition zones. The development of longterm re management strategy in cooperation with local communities is key to mitigate the negative effects of wild res in biodiverse mountain systems like the Huascaran Biosphere Reserve.

Declarations
Ethics approval and consent to participate Not applicable. Figure 1 Huascaran Biosphere Reserve (HBR) (a) protects most of Cordillera Blanca and local people across three zones: the core (b), buffer (c) and transition zone (d). In (a), we also shown main roads, villages, controlled areas (C. Areas), control sites within the HNP. Monthly interannual variation on temperature (e) and precipitation (f) characterize the climatic seasonality along the year. All pictures by CSSR.