Fire occurrence and fire mitigation strategies in a grassland reforestation area in the Philippines

Forest Policy and Economics 64 (2016) 35–45

Contents lists available at ScienceDirect

Forest Policy and Economics journal homepage: www.elsevier.com/locate/forpol

Fire occurrence and fire mitigation strategies in a grassland reforestation area in the Philippines Rico C. Ancog ⁎, Leonardo M. Florece, Ozzy Boy Nicopior School of Environmental Science and Management, University of the Philippines Los Baños, Laguna, Philippines

a r t i c l e

i n f o

Article history: Received 15 March 2015 Received in revised form 8 January 2016 Accepted 8 January 2016 Available online xxxx Keywords: Grassland fires Remote sensing Philippines Fire mitigation strategies Reforestation Uplands

a b s t r a c t While fire has been identified as a major factor negatively affecting success of reforestation projects in the Philippines, no study so far has been conducted on the types and extent of fire mitigation strategies being implemented by upland farmers. This study presents an analysis of the occurrence of grassland fire, the factors influencing it, and the fire mitigation strategies employed by small farm-holders in four adjacent upland communities in Carranglan, Nueva Ecija, Philippines. With a total of 779 fire incidents recorded from 2002–2014 covering an estimated overall total of 19,500 ha, grassland areas in Carranglan warrants urgent rehabilitation. The number of fire occurrence in Carranglan were found to have significant positive relationship with temperature, compound topographic index (CTI) and aspect (P b 0.05). In response, upland farmers implement several fire mitigation strategies that are either vegetational and structural in nature. The most common practices include regular grass cutting and the establishment of fire lines. Using simple linear regression, the number of fire mitigation strategies implemented by upland farmers is very highly associated to the years of residence in the area, number of children in the household, age of respondents, educational level, livelihood, and farm lot ownership (P b 0.01). The limited number of fire-mitigating strategies implemented by the upland farmers may have contributed to the frequent occurrence of grassland fires in Carranglan. Strategies that would significantly increase farmers' awareness of and capability in implementing various fire-mitigating strategies must be incorporated in a fire hazards management program to ensure success of reforestation projects, which is a key approach in rehabilitating the upland ecosystems. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Reforestation is a major activity among developing countries as a direct response to the rapid decline of tropical forests (FAO, Food and Agriculture Organization of the United Nations, 2013; Le et al., 2014). At the minimum, it involves planting of trees on cleared land to establish tree cover (de Jong, 2010) that in the long-run is expected to provide vital ecosystem services (Sayer et al., 2004). Depending on the identified cause of forest degradation, reforestation projects are being designed and implemented using a wide array of approaches, types of technologies and resources, level of stakeholder participation, and implementing actors. Over the past decades, success of reforestation projects vary across different countries depending on a combination of technical, socioeconomic and environmental factors (Chokkalingam et al. 2005; Le et al., 2014). Grassland ecosystems formed as a result of deforestation and degradation of forestlands particularly require immediate reforestation activities (FAO, Food and Agriculture Organization of the United Nations, ⁎ Corresponding author at: School of Environmental Science and Management, University of the Philippines Los Baños, College, Los Banos, Laguna, 4031, Philippines. E-mail address: [email protected] (R.C. Ancog).

http://dx.doi.org/10.1016/j.forpol.2016.01.002 1389-9341/© 2016 Elsevier B.V. All rights reserved.

2013). Zanne and Chapman (2001) observe that forest regrowth is often slow in cleared and abandoned tropical forests. In the Philippines, grasslands are classified under two distinct types: (1) Imperata cylindrica-dominated areas, which generally represent degraded, acidic, low organic matter, and dry soil areas susceptible to erosion; and (2) Saccharum spontaneum-dominated areas that occur in areas where there is adequate moisture (Concepcion and Samar, 1995; Magcale-Macandog and Galinada, 1998). For years, reforestation programs in the Philippines have focused on Imperata-dominated landscapes that are prone to environmental hazards such as wildfires, landslides, among others. The spread of Imperata is associated with a decline in soil fertility contributing to reducing crop productivity (Van Noordwijk et al., 1996). The rehabilitation of Imperata grasslands will require a much better understanding of their area, distribution, and characteristics (Garrity et al., 1996). Over the past four decades, however, reforestation in the Philippines has gained limited success (Chokkalingam et al., 2005; Rebugio et al., 2007; Israel and Israel and Lintag, 2013). In terms of the institutional aspect, many reforestation projects face similar challenges. These include low acceptability among the local people, non-involvement of key stakeholders in reforestation planning, untimely release of funds, improper mode of reforestation implementation, regular occurrence of

36

R.C. Ancog et al. / Forest Policy and Economics 64 (2016) 35–45

fire, poor site quality, and poor site-species matching (Rebugio et al., 2007; Florece, 2010). Analysis on where, when, and why fires occur could enable more effective and forward-looking damage reduction strategies to significantly mitigate fire risks especially in reforestation areas (Prestemon et al., 2001). Fire has been part of the ecological processes of grasslands (Collins, 1992; Fuhlendorf and Smiens, 1999). The probability of fire occurence is greatest in areas with high biomass accumulation (Fuhlendorf and Engle, 2004). The role of fire in grasslands and savannahs is well recognized and can either be beneficial or detrimental depending on its use. Grasslands have been found to maintain biodiversity if burnt every 1–3 years in Australian savannahs (Setterfield, 2002); reduce fuel load in California and other areas in the western United States (Faulkner et al., 1989); manage invasive species (DiTomaso et al., 1999); and eliminate old foliage and encourage regrowth of grasses for the benefit of browsing livestock. In rural Philippines, fire appears to be socially condoned because it is a cultural norm in preparing swidden farms. Using fire in land preparation has been documented to be the leading cause of fire escaping to adjoining reforestation projects (Florece, 2010). As fire could negatively affect the biophysical components of the ecosystem, reforestation managers and decision makers opt to adopt a fire exclusion policy. Goldammer and Penafiel (1990) expressed that forest managers and fire control officers are still running behind the wildfire problem as they are inadequately prepared. Balancing ecosystem management, human rural development, and fire risk remains a major concern to natural resource agencies (Dombeck et al., 2004; Sturtevant et al., 2009). Thus, a better approach toward integrated fire management that would allow reaping the benefit from fire but excluding the negative impacts of uncontrolled fire, is still wanting. While fire has been identified as a major factor negatively affecting success of reforestation projects in the Philippines (Chokkalingam et al., 2005; Rebugio et al., 2007; Israel and Lintag, 2013), no study so far has been conducted on the types and extent of fire mitigation strategies being implemented by upland farmers who are also engaged in reforestation projects. To accelerate success of reforestation projects, an analysis of the factors that affect fire occurrence and the use of fire mitigation strategies among upland farmers is imperative. Schemes and programs that could heighten implementation of fire mitigation strategies could then be crafted accordingly. This study provides an analysis of grassland fire occurrence, factors influencing it, and fire mitigation strategies of upland farmers in a large grassland area in Carranglan, Nueva Ecija, Philippines—an area that for years has been subject to several reforestation programs. Specifically, this paper aims to identify the types of fire mitigation strategies among the upland farmers and determine the factors affecting the number of fire mitigation strategies employed by upland farmers.

2. Materials and methods 2.1. Study site The Philippines provides a classic example of deforestation common to many developing countries. The Food and Agriculture Organization of the United Nations (FAO) (2013) explains that the cycle of tropical deforestation typically begins with excessive logging that results in logged-over forests being converted for agricultural uses. Unproductive farmlands are subsequently abandoned after being subjected to intensive anthropogenic activities, soil degradation, recurring disturbances (especially fires), and isolation from intact forests, among other reasons. Being one of the 17 mega-diverse countries (Mittermeir, 1997), the loss of vital forest ecosystems in the Philippines resulted in the decline of a wide array of ecosystem services (Rebugio et al., 2007; Lasco, 2008). In response, a number of reforestation efforts particularly on open grasslands and denuded lands were implemented across the country (Harrison et al., 2004; Sayer et al., 2004; Chokkalingam et al., 2005).

Among the recipients of massive reforestation projects is the municipality of Carranglan in Nueva Ecija (15°59′42″ N, 121°1′40″ E), which is the focus of this study. It is an agriculture-based town with a total land area of 70,531 hectares (ha) (Fig. 1). It is the northernmost municipality of the province of Nueva Ecija in the island of Luzon. The landscape is characterized by rugged mountainous terrain that ranges from nearly level, undulating, to hilly slopes as it approaches the Caraballo and Cordillera mountain ranges in the north and the Sierra Madre mountain range in the east. The major land uses in Carranglan are grassland, alienable and disposable land, forest plantation, and forestland vegetated predominantly by secondary growth forests (Peras et al., 2009). There are two soil types in Carranglan: Annam sandy clay loam and Umingan sand (PhilGIS, Philippine GIS Data Clearinghouse, 2013). Recommended crops for the Annam soil type include upland rice, lowland rice, coconut, corn, plantation crops, root crops, vegetables, and perennial trees whereas the Umingan soil series is suitable for corn, vegetables, lowland rice, coconut, sugarcane, and perennial trees (Carating et al., 2014). The municipality has Type I climate with a pronounced wet season from May to November and dry during the rest of the year. The observed normal annual precipitation rate is 1854.9 millimeters (mm) (PAGASADOST, Philippine Atmospheric, Geophysical and Astronomical Services Administration-Department of Science and Technology, 2014). Land in Carranglan is under the administrative jurisdiction of several government agencies as it hosts the watershed that supplies water to the hydroelectric power plant in Pantabangan, Nueva Ecija. The presence of several government agencies represents varying and oftentimes competing interests with regard to resource use and conservation efforts. So far, large tracts of grasslands currently abound in Carranglan despite the attempts and investments to rehabilitate the watershed. While grassland fires may have positive ecological benefits (e.g., improving the pasture areas of some farmers through the regrowth of grass species like Themeda triandra and Imperata cyclindrica for the production of thatch), most grassland fires affect regenerating tree species that in six to eight years would have improved the carbon stock and the biophysical and microclimate of the site for the entry of other plant species. Generally, policies on the use of fire in regulating biomass in upland areas in the Philippines exist but are not widely practiced. 2.2. Overview of reforestation projects in grassland areas in Carranglan, Nueva Ecija For a long time, a number of reforestation projects have been implemented as a key strategy to rehabilitate the grassland areas of Carranglan, Nueva Ecija (Table 1). The overall intention of these reforestation projects is to facilitate succession of these grassland areas to shrubland or a forest stand. These projects were mostly initiated by the Department of Environment and Natural Resources (DENR) of the Philippines (e.g., Program on Forest Ecosystem Management [PROFEM] and the Philippine–Japan Reforestation Project) in the 1970s. Other government agencies such as the National Irrigation Authority and the National Power Corporation that have stakes in Carranglan also engage in reforestation programs. A recent reforestation project that also covers Carranglan, is the National Greening Program (NGP), a multigovernment agency effort aimed at planting about 1.5 billion trees covering 1.5 million ha in the Philippines for a period of six years (2011–2016). The reforestation projects in Carranglan were particularly targeted to rehabilitate its grassland areas. In terms of area coverage, many of these projects covered all of the villages in the municipality of Carranglan. The total area covered ranged from 30–54,090 ha. In terms of components, these projects included reforestation, afforestation, plantation establishment, replanting, forest protection, agroforestry establishment, and community and institutional strengthening. The DENR in collaboration with a number of foreign and local partners initiated and implemented these projects. In the 1970s to 1980s, most

R.C. Ancog et al. / Forest Policy and Economics 64 (2016) 35–45

37

Fig. 1. Map of the study areas in Carranglan, Nueva Ecija, Philippines.

of the reforestation projects usually employed top-down approaches with the national government playing the major role. However, this changed starting in the 1990s when community-based forest management programs were implemented. Analysis of the current land cover condition of Carranglan indicates minimal success of the previous reforestation projects (Fig. 1). Grassland remains to be the major land cover among the reforestation areas in Carranglan with I. cylindrica (cogon) as the dominant species (Fig. 2 and Fig. 3). To change this condition, a number of reforestation projects were implemented in these grassland areas. However, an important cause of the limited success of the reforestation projects in Carranglan is due to fires in the forests and grasslands that have become a recurring phenomenon (Florece, 2010). The grassland reforestation areas of Carranglan has distinct climatic condition that make it particularly conducive to fire. Water, fuel wood and wild animals in Carranglan are now limited as climatic condition is extremely harsh when the dry season starts in November and ends in May with precipitation averaging 1700 mm and with an average temperature of about 28 °C (Florece,

2010). Large areas of the established plantations by previous reforestation projects were also continuously subjected to poaching either for charcoal or lumber. Thus, aside from the geophysical characteristics being comparable with many of the upland ecosystems in the Philippines, the high incidence of grassland fires in the reforestation projects was the basis in the selection of Carranglan as the study site. 2.3. Analysis of fire occurrence using MODIS-MCD45A1 This study covered the grassland reforestation projects in Carranglan, Nueva Ecija, Philippines. Spatially-explicit records on fire incidences were extracted from MCD45A1, the code name for the burned area product of Moderate Resolution Imaging Spectroradiometer (MODIS). This product has a resolution of 500 meters (m) and provides daily fire records expressed in a Julian-year date on a per pixel basis (LP DAAC, Land Processes Distributed Active Archive Center, 2014). MODIS is the sensor on board the Terra and Aqua satellites, which were launched in December 1999 and May 2002, respectively, as part of

Table 1 List of grassland reforestation programs implemented in Carranglan, Nueva Ecija, Philippines. Year

Reforestation program/project

Major implementor and partners

Estimated area (ha)

1938–1991 1976–1992

Carranglan–San Jose Reforestation Project Republic of the Philippines–Japan Reforestation Project Regular Reforestation Program Integrated Social Forestry Program Contract Reforestation Program NPC yearly reforestation project Community Based Forest Management (CBFM) Program

Department of Environment and Natural Resourcesa (DENR) DENR and Japan International Cooperation Agency (JICA) DENR DENR DENR and National Irrigation Authority (NIA) National Power Corporation (NPC) DENR

54,090 2596 823 856 900 30–40/year Covers all grassland areas of Carranglan

1989–1992 1995–1997

Source: Above list was based on available records from the Forest Management Bureau of the Department of Environment and Natural Resources (DENR) of the Philippines. a Prior to 1987, the DENR was named Bureau of Forest Development (BFD).

38

R.C. Ancog et al. / Forest Policy and Economics 64 (2016) 35–45

Fig. 2. Burned areas in Carranglan, Nueva Ecija, Philippines.

NASA's Earth Observing System. MODIS provides exceptional capability over other sensors in terms of fire monitoring and was designed to include characteristics specifically for fire detection (Justice et al., 2006). Boschetti et al. (2009) elaborates the algorithm for producing the MCD45A1 product. However, this type of data only provide information on the location of fire and not on its actual size and shape. It should be noted also that while potential underestimation due to the inability of MODIS to detect small fires occurring at less 500 square meters (m2) (Tran et al., 2008; Anaya and Chuvieco, 2012), this is compensated by overestimation of some fires (Kugbe et al., 2012) because each fire data point is approximately 25 ha (Boschetti et al., 2009). With the collected data on fire occurrence and land cover, data preparation and processing followed. Each MCD45A1 raster data has pixels

coded with values corresponding to the approximate day of burning as well as codes indicating unburned areas, water, or lack of data (Boschetti et al., 2009). Only those pixels with codes indicating the day of burning were considered in the subsequent analysis. These data points were treated as fire events. The study processed all the MCD45A1 data from 2002–2014. The frequency of fires per data point in monthly and yearly bases were derived. The total number of fire incidences was counted regardless of date which means that those data points with more than one record were all recounted. In counting the number of days affected by fire, only the dates with fire records were considered. This means that recurrent fires and simultaneous events in different spatial locations were disregarded.

Fig. 3. Active fire in Carranglan, Nueva Ecija, Philippines.

R.C. Ancog et al. / Forest Policy and Economics 64 (2016) 35–45

The relative distribution of fire events over different land cover classes was also determined by overlaying the fire data points over land cover data. The land cover data of Carranglan was extracted from the nationwide land cover data obtained from the National Mapping and Resource Information Authority (NAMRIA) (2010) of the Philippines. 2.4. Analysis of factors influencing fire occurence Factors that influence the occurrence of fire in Carranglan was analyzed using regression analyses. The number of times (or recurrence) a single data point records a fire was taken as the dependent variable (Table 2). A value of “0” translates to “absence” which also means that there was no recorded fire incident for that datapoint for the whole study period (2002–2014). On the other hand, values 1 to 7 indicate the number of times fire occurred for the whole study period as based on MODIS data. A total of 13 independent variables were included in this study as the explanatory variables grouped under three components namely: physical/topographic, climate, and human activity (Table 2). The digital elevation model (DEM) data used was from the Shuttle Radar Topography Mission (SRTM) with a resolution of 90 m. The SRTM 90 m was downloaded from www.cgiar-csi.org/data/srtm-90m-digitalelevation-database-v4-1. Four other data sets were derived from SRTM 90 m: elevation, slope (%), aspect, and CTI. To match the resolution of the MODIS burnt fire data, SRTM was resampled to 500 m. As described in Prasad et al. (2007) and Vadrevu et al. (2009), CTI is a function of upstream contributing area and slope. This index is also known as wetness index. The distribution of water over an area is influenced by lateral flow and thus controlled by elevation differences.

39

Wetness index is given as Wetness Index ¼

ln ðAÞ tanðBÞ

where ln is the natural logarithm, B is the slope angle, and A is the upslope area per unit width of contour. In general, the index is a measure of the tendency of water to accumulate at any point on a slope. Areas with low CTI values represent places with small catchments and steep slopes. Areas with high CTI values represent places with large catchments and gentle slopes. With the accumulation of water, it is hypothesized that higher CTI values would result to less occurrence of fire. Climate data used in this study were downloaded from WorldClimGlobal Climate Data repository (www.worldclim.org/bioclim). Temperature (in °Celsius) and precipitation (in mm) data were averages of 50year observations. Human activity indicators were limited to proximity to roads, rivers, and crops (annual and perennial). Data on roads and rivers were taken from the Geofabrik website (www.geofabrik.de) while annual/perennial crops were extracted from the NAMRIA 2010 nationwide land cover data. Euclidean distance method was employed to determine the relative distances. Houses and infrastructure areas were taken from Google Earth and then proximity analysis was done using Euclidean distance (Table 2). As for the vegetation-related indicators, land cover was extracted from National Mapping and Resource Information Authority (NAMRIA) (2010) while both normalized difference vegetation index (NDVI) and normalized difference moisture index (NDMI) were derived

Table 2 Variables analyzed in the regression analysis for the determinants of fire occurrence (2002–2014) in Carranglan, Nueva Ecija. Independent variables

Data type

Units

Value range

Remarks

Data source

Physical/topographic indicators Elevation Numeric masl

40 to 2040

SRTM 90 m resampled to 500 m

Slope

Numeric Percent

0 to 65.9

Extracted from resampled SRTM DEM

Aspect

String

Nominal Nominal

Extracted from resampled SRTM DEM

Compound topographic index

Numeric Index

3.86 to 19.40

Derived from resampled SRTM DEM

www.cgiar-csi. org/data/srtm-90m-digital-elevation-database-v4-1 www.cgiar-csi. org/data/srtm-90m-digital-elevation-database-v4-1 www.cgiar-csi. org/data/srtm-90m-digital-elevation-database-v4-1 www.cgiar-csi. org/data/srtm-90m-digital-elevation-database-v4-1

Climate indicators Mean annual rainfall

Numeric mm

50-yr average, 1950–2000

World Climate Data, www.worldclim.org/bioclim

Mean annual temperature

Numeric °C

1892 to 2807 18.6 to 26.3

50-yr average, 1950–2000

World Climate Data, www.worldclim.org/bioclim

Euclidean distance

www.geofabrik.de

Euclidean distance

www.geofabrik.de

Euclidean distance

NAMRIA 2010 land cover

Euclidean distance

Google Earth

derived from Landsat 5 image

NAMRIA 2010 land cover Resampled to 500 m, www.glovis.usgs.gov

derived from Landsat 5 image

Resampled to 500 m, www.glovis.usgs.gov

Human activity indicators Distance to roads Numeric Meters Distance to rivers

Numeric Meters

Distance to annual/perennial crops Distance to settlements

Numeric Meters Numeric Meters

0 to 8810.86 0 to 3055.88 0 to 7874.41 0 to 9632.06

Vegetation-related indicators Land cover String Nominal Nominal Normalized difference Numeric Index −0.25 to vegetation index (NDVI) 0.719 Normalized difference Numeric Index −0.4375 to moisture index (NDMI) 0.3333 Dependent variable Recurrence

Numeric Count

0 to 7

The number of times a single data point records a fire. “0” means there was no recorded fire incident on that datapoint for the whole study period. “1” to “7” indicate the number of times a fire occurred for the whole study period.

40

R.C. Ancog et al. / Forest Policy and Economics 64 (2016) 35–45

from 2010 Landsat images. All were resampled to a resolution of 500 m (Table 2).

3. Results and discussion 3.1. Socio-demographic condition of upland communities

2.5. Analysis of fire mitigation strategies This study also analyzed the fire mitigation strategies implemented by the upland farmers who are engaged in reforestation projects. In the context of this study, fire mitigation strategies broadly included all activities being intentionally implemented to prevent, reduce, control, and regulate the occurrence and spread of fire. In the absence of an earlier local study, the types of fire mitigation strategies were determined during initial field work and following Schulte and Miller (2010). These fire mitigation strategies are classified as vegetation-related, structural in nature, on-site fire response actions, and prescribed burning. Vegetation-related fire mitigation strategies include adjustments in the plants such as pruning, thinning, salvage cutting, etc., to reduce the plant biomass that may serve as additional fuel if fires occur. On the other hand, structural fire mitigation strategies include engineering and physical adjustments in the farm such as firelines and firebreaks using concrete and physical materials. On-site fire response actions are those implemented by farmers right at the onset of grassland fires, which include fire suppression and water sprinkling, among others. Lastly, prescribed burning is the intentional introduction of fire in favorable weather and fuel conditions (Schulte and Miller, 2010). Analysis on the views of stakeholders in fire-prone areas was found to be critical in instituting effective strategies to mitigate fire (Valente et al., 2015). As such, a series of household interviews were implemented from December 2012 to March 2013 covering four adjacent barangays (villages) in the municipality of Carranglan, Nueva Ecija, Philippines, namely: Burgos, Joson, General Luna, and San Agustin. These study areas were selected as they have recorded higher fire occurrences from 2002–2012 based on the NASA's Fire Information for Resource Management System (FIRMS).1 This was cross-validated using secondary data and key informants from the municipal government of Carranglan. The four villages were part of the many reforestation projects initiated by government agencies and other concerned groups. In fact, all respondents of the study were cooperators of a number of reforestation projects particularly the latest community-based forest management (CBFM) program. A pre-tested questionnaire was administered in face-to-face interviews with 130 randomly selected household heads (level of confidence = 95%; variability = 0.5; precision = ± 10%) in the four villages covered in this study. The data and information collected were: socioeconomic profile, farming activities, and grassland fires and management. Simple linear regression was employed to analyze the determinants of the number of fire mitigation strategies used by the upland farmers. A number of preliminary analyses were done to make sure that no assumptions on normality, linearity, multicollinearity, and homoscedasticity were violated. The number of fire mitigation strategies (m) being implemented by the upland farmers were taken as the dependent variable while farmers' socioeconomic and farming characteristics were used as explanatory variables. In relation to the local policy encouraging farmers to implement more fire mitigation strategies, it was assumed in this study that a higher number of fire mitigation strategies being implemented by farmers would be good in controlling the occurrence and spread of fires. All data processing for this study was done using SAS (Ver. 9.0).

1

URL: http://earthdata.nasa.gov/data/near-real-time-data/firms.

The upland communities in the Philippines generally experience pervasive poverty particularly in areas that are environmentally challenged due to paucity of resources (Rola, 2011). This can be clearly seen in the upland communities covered in this study where most of them eked out a living from upland areas that are perennially affected by environmental hazards, particularly fire. Overall, most of the upland farmers were born in Carranglan (56.9%) and have resided in their respective barangays for an average of 33.95 years (Table 3). The mean land area of the respondents' home lots is 1.22 ha, which is comparable to other areas in the Philippines. While most of the respondents (60%) work as farmers, some of them are employed as hired laborers (10%) and engage in non-farm small businesses (15.4%) or as government employees (6.9%), among others (Table 3). The mean monthly income of the respondents is PHP 8563.32 (USD 1 = PHP 41.00), which is below the 2012 per capita poverty threshold of PHP 10,121.00 set by the Philippine Statistics Authority. Among respondents who are employed as farmers, most (72.3%) are tenants. The farmlots owned by farmers have a mean area of 2.42 ha, which are mostly cultivated as agroforestry (63.8%). Farmers have been engaged in farming from 10.53 years (minimum in Joson) to as long as 43.52 years (maximum in San Agustin). 3.2. Fire occurrence using remotely-sensed data The MCD45A1 data show that of the total 17 villages comprising the municipality of Carranglan, nine (Bunga, Burgos, General Luna, Joson, Piut, Puncan, R.A. Padilla, Salazar, and San Agustin) have been affected by fires (Fig. 4). It can be clearly observed that some data points have more than one record of fire occurrence suggesting that fires are recurrent and more frequent in some areas in Carranglan. Fire regimes occur when fire events are recurrent in a given area over a certain period of time, which could be due to climatic and socio-economic factors (Pezzatti et al., 2013). In this study, fire occurrences were found to be particularly clustering in the villages of Burgos, Joson, San Agustin, and General Luna. These villages are largely dominated by I. cylindrica, which could explain its vulnerability to grassland fire (Magcale-Macandog and Galinada, 1998; Garrity et al., 1996). The village of Joson recorded the most number of fires (n = 186), followed by San Agustin (n = 167), Bunga (n = 146), and Burgos (n = 141). Moreover, Joson also had the highest number of recorded fires during the El Niño years of 2005 (n = 37) and 2010 (n = 38). The estimated total burned area in Carranglan, Nueva Ecija for 2002–2015 is 19,500 ha based on MCD45A1 data. This covers about

Table 3 Socio-demographic characteristics of respondents from sampled villages in Carranglan, Nueva Ecija, Philippines (2013). Socio-demographic variables

Total (n = 130) Mean (SD)

No. of years living in the area Home lot area (ha) No. of children Total household size Education (no. of yrs) % farming as main livelihood Monthly income (PHP) Farm land area (ha) Total years farming % farm land under tenancy Farming methods (m2) Monocropping Agroforestry Kept as forest

Freq (%)

33.95 (15.9) 1.22 (1.9) 3.13 (2.2) 4.69 (1.9) 8.34 (3.0) 79 (60.8) 8563.32 (12,954.77) 2.42 (3.0) 25.11 (20.8) 94 (72.3) 28 (21.5) 83 (63.8) 19 (14.7)

R.C. Ancog et al. / Forest Policy and Economics 64 (2016) 35–45

41

Fig. 4. Burned areas in Carranglan, Nueva Ecija, Philippines from 2002 to 2014.

27.64% of the total land area of Carranglan (70,531 ha). Within the 13year duration covered in this study, grassland fire averages at 1600 ha per year (Fig. 4). Across different land use types, most (88%) of the fire events within the 13-year time period occur in grassland, which were the same areas earlier subjected to reforestation projects. The high occurrence of fires in grassland areas in Carranglan is consistent with previous studies showing that grassland areas are the most vulnerable to wildfires (Prestemon et al., 2001; Sturtevant and Cleland, 2007; Sturtevant et al., 2009). Currently, most upland areas in Carranglan are grasslands signifying their nonsuccession to a shrubland or a forest stand (Fig. 1). The MCD45A1 data show a total of 779 fire incidents occurring in 449 days from 2002 to 2014. This translates to an annual average of 37 days with recorded fires (Fig. 5). This is equivalent to an average of 65 fire incidents in the last 13 years. The most “fire active years” were 2005 (94 days with fire) and 2010 (73 days with fire) while the year with the most number of fire incidents is 2005 (n = 148). Fire incidents occur beginning January and extending until May in a given year. Interestingly, there have been no recorded fire events from June to December. The trend clearly shows that fuel accumulation happens from June to December with the first fires occurring in January, leading to more incidences of fires. With the decrease in available fuel, fire occurrence becomes less frequent from February until May. Additionally, the temporal pattern of forest fires as observed from space by MODIS is consistent with that of the local people's observations during the household interviews. Of all the fire-affected months, February consistently had the most number of fire incidents. All these information from MCD45A1 confirms the common observation that fires in Carranglan repeatedly take place in grasslands. This implies that with the aid of fire, grasslands can become a selfmaintaining ecosystem.

3.3. Factors associated with fire occurence Examining the factors that influence fire occurrence helps to find ways of proper management. The number of fire occurrences in Carranglan were found to be positively related to temperature, CTI, and aspect (P b .05). Consistent with the literature (Brenkert-Smith et al., 2012; Fuhlendorf and Engle, 2004), results of this study show that higher temperatures in an area could lead to higher frequency of fire occurrences. Higher temperatures would lead to dry conditions where fires are

Fig. 5. Fire incidents in Carranglan, Nueva Ecija, Philippines covering 2002–2014.

42

R.C. Ancog et al. / Forest Policy and Economics 64 (2016) 35–45

highly probable. In Carranglan, good accumulation of fuel, such as the biomass from I. cylindrica, results in increased incidence of fires. This condition is particularly observable during the El Niño years. Aspect also plays an important role in the recurrence of fire in Carranglan. Higher frequency of fire is recorded during the northeast monsoon (amihan) period in October to March. During this period, Carranglan experiences dry wind blowing in a westward direction since it is in the interior and leeward portion of the eastern part of Luzon Island that faces the Pacific Ocean. Slope was not a significant predictor of fire occurrence in Carranglan, thus, grassland areas in low elevations with flat slopes and heavily dominated by grasses showed high frequency of fire occurrence. Though it has been earlier hypothesized that higher CTI values may lead to less fire occurrence, an inverse relationship was found to be the case in Carranglan. The CTI, which measures the tendency of water to accumulate at any point on a slope, was found in this study to be a positive predictor of fire occurrence. It should be noted that Carranglan has large catchment areas with gentle slopes. The CTI may be high in some areas but what was actually observed were small intermittent creeks present only during the rainy season. The high dominance of I. cylindrica that act as fuel may have offsetted the expected reduction effects of CTI resulting in more fires in Carranglan. In addition, farmers in Carranglan attested that farming activities are more widespread in areas that are relatively wetter. Many farmers related that intentional burning of the grasslands is high in areas with optimal wet conditions because grasses can easily regrow and produce fresh biomass that their livestock prefer (Table 4). 3.4. Frequency and types of implemented fire mitigation strategies Table 5 shows the distribution of respondents per number of implemented fire mitigation strategies. To exhaustively document typologies of fire mitigation strategies, respondents were asked to liberally identify strategies that they implement in preparation and/or in response to the prevention and management of grassland fires in their respective farmholdings and in their community. Overall, this study confirms that upland farmers in Carranglan do implement a number of fire mitigation strategies. About 45.4% and 33.1% of the respondents enumerated either one or two fire mitigation strategies, respectively, while 15.4% claimed to have practiced none at all. Only about 6.2% of the respondents in Carranglan claimed to practice three or more fire mitigation strategies. Comparing the computed values of the mean number of implemented fire mitigation strategies across the villages (1.3 strategies), the highest mean value was noted in General Luna with 1.42, which is Table 4 Factors influencing fire occurrence (2012–2014) in grassland reforestation areas in Carranglan, Nueva Ecija, Variables Elevation Slope Mean annual rainfall Mean annual temperature Distance to annual/perennial crops Distance to rivers Distance to roads Moisture Index Normalized Difference Vegetation Index Compound Topographic Index Distance to settlements Land cover Aspect Pseudo R-square = 0.8709. ⁎ Significant at 5%.

Codes

Estimates

Std. error

y-intercept Elevation Slope Rainfall Temperature dist_crops

3.704953 −0.1831189 −0.1436234 −0.2152435 0.19023436 −0.1857502

0.933977 0.00013 0.001104 0.000143 0.027061 ⁎ 0.00001

dist_river dist_roads NDMI NDVI

−0.0682905 −0.109544 −0.2611451 −0.2618804

0.000018 0.000007 0.108546 0.108363

CTI 0.01877071 dist_settlements −0.1533396 Land cover −0.0855971 Aspect 0.03531437

0.004855 ⁎ 0.0001 0.0001 ⁎ 0.0001

slightly higher than the mean values of other barangays covered in this study. One-way ANOVA showed a statistically significant difference in terms of the number of fire mitigation strategies being implemented across the villages [F(22,107) = 0.016, P b .05]. While farmers in these villages maintain considerable sharing of knowledge, respondents related that fire mitigation strategies is not among the information they share unlike fertilizer inputs, plant varieties, and potential sources of financial loans. So far, the number and type of fire mitigation strategies appear to be largely dependent on personal experience alone. In terms of types, upland farmers in Carranglan implement fire mitigation strategies that are either vegetational (n = 66, 51%) or structural (n = 52, 40%) in nature (Table 5). The common vegetation-related fire mitigation strategies include regular pruning and cutting down of older plants and establishment of fire breaks and fire lines using fire-resistant plants such as Gmelina arborea (yemane) and Bauhinia malabarica (alibangbang). These fire mitigation strategies were implemented by the farmers themselves as part of their land preparation activities in every cropping cycle. The affinity of the farmers to vegetation-related fire mitigation strategies may be due to their familiarity with these practices, which do not require additional farming inputs. For structural-related fire mitigation strategies, the most commonly noted responses include the establishment of structural fire lines using local and makeshift materials and land excavation or digging of 1–1.5 ft bordering the farmlot, among others (Table 5). Moreover, several respondents have likewise mentioned that they are already using chemical sprays (i.e., herbicides) to control growth of plants and in effect fuel production, which is a relatively uncommon practice as this entails additional cost in land preparation and maintenance. Overall, only three farmers (2%) claimed to have implemented prescribed burning, which shows how rare the practice is among upland farmers in the Philippines. Thus, respondents confirmed that they would need training on fire mitigation strategies, particularly on prescribed burning, to increase their familiarity, awareness, and confidence in implementing such practices. 3.5. Determinants of the number of implemented fire mitigation strategies A total of six variables (α = .01) were found to predict the number of fire mitigation strategies implemented by the upland farmers in Carranglan (Table 6). These include number of years of residence in the area (NoYrs), total number of children in the household (NoChild), age of respondents (Age), educational level (Educ), farming as the main livelihood (Mainjob), and total farm land area (LandAsset). The regression analysis also confirms that the factors that would enhance the practice of fire mitigation strategies among upland communities are very much related to the characteristics of the farming households. The total number of children in the household (NoChild), age of the farmer-respondents (Age), and educational level (Educ) of the household head positively affect the number of fire mitigation strategies being implemented by the upland farmers. As commonly practiced in many upland areas in the Philippines, the large number of children in a household could serve as additional labor in implementing fire mitigation strategies. Being less mechanized and with minimal financial resources, tending a farm in upland areas in the Philippines is labor-intensive. Implementing fire mitigation strategies is taken as an additional effort by the farmers. Thus, children are assigned the additional task of implementing fire mitigation strategies. The age of the household heads and their educational attainment also showed significant positive relationship with the number of fire mitigation strategies implemented by the upland farmers. This shows that farmers who are older and more educated are more open and likely able to implement fire mitigation strategies in the farm. With upland farming in the Philippines engaged in by those who are usually less educated, this finding requires a more long-term need of enhancing the education status of the farmers to enable them to be more open to implementing additional technologies (i.e., fire mitigation strategies),

R.C. Ancog et al. / Forest Policy and Economics 64 (2016) 35–45

43

Table 5 Frequency and types of fire mitigation strategies implemented by upland farmers in Carranglan, Nueva Ecija, Philippines (2013). Frequency of fire mitigation strategies implemented

Villages (Barangays) Burgos (n = 49)

Joson (n = 30)

General Luna (n = 28

San Agustin (n = 23)

Total (n = 130)

N

%

N

%

N

%

N

%

N

%

0 (none) 1 2 N3 Total Mean S.D.

4 15 10 1 30 1.26 0.73

13.3 50.0 33.3 3.3 100.0

10 19 16 4 49 1.28 0.88

20.4 38.8 32.7 8.2 100.0

4 10 12 2 28 1.42 0.83

14.3 35.7 42.9 7.1 100.0

2 15 5 1 23 1.21 0.67

8.7 65.2 21.7 4.3 100.0

20 59 43 8 130 1.30 0.80

15.4 45.4 33.1 6.2 100.0

Types of fire mitigation strategies implementeda,b 1. Vegetation-related (cutting and transfer of organic matter such as grasses to other areas) 2. Structural and engineering work (water impoundment and/or collection) 3. Field action (chemical spraying) 4. Prescribed burning of organic matter on-site 5. None

14 18 6 – 4

28 36 12 – 8

21 27 11 – 10

70 89 36 – 33

23 8 8 1 4

83 28 28 4 14

11 9 6 2 2

48 39 23 9 9

69 62 31 3 20

53 48 23 2 15

a

Respondents could provide multiple answers. Percentage computed for the types of fire mitigation strategies implemented would mean the number of respondents implementing a specific fire mitigation strategies (m) relative to the total sample size multiplied by 100. b

which they may not be familiar with but could be instilled in them as timely and necessary. In Carranglan, farmers practicing fire mitigation strategies who are older and relatively more educated also enjoy certain levels of influence among their upland counterparts. Hence, government programs could tap them as cooperators to set up demonstration farms that would show that fire mitigation may be instrumental in supporting farm activity. On the other hand, with negative regression weights, variables such as the number of years of residence in the area (NoYrs) and total farm land area (LandAsset) indicate that farmers who have stayed longer and have larger farm land implemented fewer fire mitigation strategies. This is directly related to a lack of appreciation of and familiarity with fire mitigation among the upland farmers. Most of the farmers who have resided longer in Carranglan do not see the need to implement fire mitigation strategies because many of them were not aware of such practices. It should be noted that these farmers who have lived longer in their locations have experienced land tenure insecurity until the

implementation of forest stewardship programs (such as CBFM) in the 1990s. However, for farmers who have resided longer in Carranglan, land security is perceived as complete autonomy in terms of how their farms will be cultivated. Thus, many farmers related that government initiatives to incorporate fire management strategies into their farm operations were seen as intrusions into their independence. This case is further compounded by the fact that fire management has not become a norm among upland communities yet. Several key informants further related that while an ordinance requiring farmers to practice fire mitigation strategies has been passed, no one has been penalized for noncompliance. It was also found that farmers with relatively larger total farm land area (LandAsset), reported implementing fewer fire mitigation strategies. One potential explanation for this is the common notion among upland farmers that fire control is not a priority since their landholding is relatively larger than others and potential loss due to fire is less. Moreover, the negative beta coefficients for the variable Mainjob show that

Table 6 Results of the simple linear regression analysis on the determinants of the number of fire mitigation strategies implemented. Parameter estimates Variables

Codes

Estimate

Std. error

Wald

VIF

Number of years living in the locality Total home lot area Type of ownership of home lot Religion Total number of children in the household Total household size Gender Age of the respondent Civil status of the respondent Educational attainment of the respondent Place of birth Agriculture as the main job of the household head Total combined monthly income of the household Total farm land area Total number of years farming Types of land ownership of the farm land Types of farming methods Practicing crop rotation

NoYrs LotArea HowQcrd Religion NoChild TotalHH Gender Age CivilStatus Educ Native Mainjob Income Landasset YrsasFarmr LandOwnership WaysFarming Croprotation

−0.05096 −0.19438 0.24827 −0.04465 0.35034 0.02051 0.23545 0.06810 0.05917 0.40579 −0.03118 −0.24895 −0.00001 −0.00009 0.02610 0.04529 0.48198 0.29536

0.02306 0.15681 0.20901 0.11299 0.15898 0.19757 0.60207 0.03014 0.11676 0.11260 0.72871 0.11784 0.00002 0.00004 0.01997 0.28842 0.58794 0.33978

4.88511 1.53653 1.41094 0.15615 4.85625 0.01078 0.15293 5.10712 0.25684 12.98657 0.00183 4.46338 0.30749 4.41709 1.70901 0.02466 0.67204 0.75566

2.455 1.756 1.704 1.827 1.825 2.105 1.615 2.487 1.251 1.586 1.919 1.65 1.623 1.803 3.796 2.215 2.335 2.366

Pseudo R-square = 0.520300641. –2 Log Likelihood = 132.7997333. VIF = variance inflation factor, which as shown above have values less than 10 signifying that there is low chance of multicollinearity among variables included in this study. ⁎ Significant at 10%. ⁎⁎ Significant at 5%.

⁎⁎

⁎⁎ ⁎ ⁎⁎ ⁎⁎ ⁎ ⁎⁎ ⁎ ⁎⁎

44

R.C. Ancog et al. / Forest Policy and Economics 64 (2016) 35–45

respondents with farming as their main occupation implemented fewer fire mitigation strategies further signifying the likely tendency of the upland farmers in Carranglan to stay true to their farming practices that do not include fire mitigation. Upland farmers related that fire mitigation strategies would mean additional work requiring additional farm inputs, and thus increasing operational costs. As confirmed in the analysis of the socioeconomic characteristics of the upland farmers in Carranglan, lack of funds was cited during interviews as a major constraint in implementing fire mitigation strategies.

3.5.1. Implications for upland reforestation programs The rehabilitation of grassland areas in the Philippines warrants urgent attention because of their high vulnerability to fire hazards. As confirmed by this study, grassland fires indeed remain a perennial threat in Carranglan hampering the success of reforestation programs and significantly affecting marginal upland communities. Analysis on the trends and characteristics of fire occurrence in the upland areas in Carranglan suggests that reforestation efforts must prioritize the villages where forest fires are clustering. Moreover, the operational capability to implement fire mitigation strategies, both at the farm and municipal levels, must be significantly improved. Thus, planting activities in the uplands should be done once substantial fire control mechanisms are in place. A number of fire mitigation strategies are already being implemented by farmers but these need to be enhanced given the urgency of combating grassland fire hazards. Given that high fire recurrence is observed in the grassland rehabilitation areas in Carranglan, the number of fire mitigation strategies implemented by upland farmers need to be increased. These reforestation areas, which were severely affected by fire, remain to be under grassland vegetation. With this experience, any upland reforestation program must not only focus on the number of trees planted and its area coverage but must also be complemented with a fire hazards management program that encourages the use of locally-specific fire mitigation strategies to be actively implemented by the upland farmers. It should be noted, however, that inadequate financial resources among upland farmers remain a major constraint from investing in strategies to mitigate and control fire. Again, government support that would enhance farmers' awareness and knowledge of grassland fire, as well as their skills and capability in implementing fire mitigation strategies, is necessary. Results of this study further indicated that the number of fire mitigation strategies implemented by upland farmers is influenced by their socioeconomic attributes such as the number of children, household size, educational attainment, farm land area, number of years farming, and their apparent affiliation to their village as expressed in their place of birth and the number of years of residency in the village. Decisionmakers, natural resource managers, and users should be made aware of how to take these significant socioeconomic variables into consideration in crafting policy. To increase the number of fire mitigation strategies to be implemented by the upland farmers, a massive information and education program on fire management is urgently required. Policy should mobilize those farmers who have high likelihood of adopting pre-defined fire mitigation strategies as characterized by their socioeconomic attributes. This study highlights that while farmers' knowledge of a few fire mitigation strategies is necessary, it is not enough. What is equally important is for farmers to have the incentive to implement fire mitigation strategies. In the current state, socioeconomic conditions of the upland farmers in Carranglan make them financially incapable of investing in fire mitigation strategies. At the individual level, immediate gains from farming is clearly the priority over long-term conservation targets. Likewise, critical to the effective management of fire is the understanding of the factors that influence its occurrence. In the case of Carranglan, the occurrence of fire is influenced by temperature, CTI, and aspect. Thus, a fire mitigation program must ensure that such factors are effectively dealt with. With these information, authorities could identify the

areas in Carranglan that would be particularly prone to fire, which could then be prioritized for fire management action. The stakeholders' views on fire would matter in how mitigation programs could be formulated and effectively implemented (Valente et al., 2015). Many farmers are in agreement that their knowledge on fire mitigation strategies needs to be enhanced. Policy makers should capitalize on this finding and raise the need to catalyze information dissemination and knowledge sharing about grassland fire mitigation. It may facilitate collective decision making by being open to discussions on opportunities and constraints related to grassland fire mitigation in upland reforestation areas. Similarly, farmers should be empowered through a consultative and collaborative process that takes into account the empirical results of this study to guide more in-depth analysis of the social, institutional, and cultural context of why farmers practice such minimal fire mitigation strategies. Specifically, current and future upland rehabilitation programs, particularly reforestation projects, must be designed and implemented with clear incorporation of fire mitigation strategies. Alongside the integration of fire mitigation into upland rehabilitation initiatives, there is a need to increase awareness and capacitate upland farmers in implementing fire mitigation strategies across the various aspects of their farm operations. Such a program could take off by first augmenting the current paucity of fire mitigation strategies, both in terms of frequency and types, that are being implemented by the upland farmers. As shown in the study, the number of fire mitigation strategies implemented by the upland farmers is significantly influenced by a confluence of factors such as the number of years of residence in the area, total number of children in the household, age of household head, educational level, if farming serves as the main livelihood, and total farm land area. Therefore, these variables must be taken into consideration in the formulation of policy initiatives and institutional arrangements to increase fire mitigation strategies implemented. In the long-term, such efforts could be expected to significantly contribute to accelerated success of reforestation programs in the upland areas and heightened rehabilitation of the upland ecosystems. 4. Conclusion This study provides an analysis of grassland fire occurrence, the factors influencing it, and the fire mitigation strategies of upland farmers in a large grassland area that for years have been subjected to grassland reforestation programs. The fire occurrence and its extent clearly highlights the seriousness of fire as a major factor negatively affecting the rehabilitation of an upland grassland area in the Philippines. This study, likewise, confirmed that upland farmers in the Philippines only know and practice a few fire mitigation strategies, which are dwarfed by the high number of fire occurrences. The high frequency of fire occurrence over the years remains to be a persistent problem jeopardizing large reforestation efforts, which in turn, have largely contributed to the minimal success of grassland area rehabilitation in the Philippines. With the observation that fires in Carranglan repeatedly occur in grasslands, it can be deduced that with the aid of fire, grasslands can become a self-maintaining ecosystem, which thwarts the overall goal of transforming grasslands into climax forest ecosystems. Acknowledgments This study was supported by the Commission on Higher Education (CHED) of the Philippines through 2013-SESAM-COD project. Assistance of Elizabeth Gatchalian and Clarissa Ruzol are gratefully acknowledged. References Anaya, J.A., Chuvieco, E., 2012. Accuracy assessment of burned area products in the orinoco basin. Photogramm. Eng. Remote. Sens. 78 (1), 53–60.

R.C. Ancog et al. / Forest Policy and Economics 64 (2016) 35–45 Boschetti, L., David R., and A. A. Hoffmann. 2009. MODIS Collection 5 Burned Area ProductMCD45User's Guide (Ver 2). Accessed (January 2013). http://198.118.194.10/QA_ WWW/forPage/MODIS_Burned_Area_Collection51_User_Guide_3.0.1.pdf. Brenkert-Smith, H., Champ, P., Flores, N., 2012. Trying not to get burned: understanding homeowners' wildfire risk-mitigation behaviors. Environ. Manage. 50, 1139–1151. Carating, R.B., Galanta, R.G., Bacatio, C.D., 2014. The Soils of the Philippines. Springer Netherlands, Amsterdam. Chokkalingam, U., Kurniawan, I., Ruchiat, Y., 2005. Fire, livelihoods, and environmental change in the Middle Mahakam peatlands, East Kalimantan. Ecology and Society 10, 1–26. Collins, S.L., 1992. Fire frequency and community heterogeneity in Tallgrass Prairie vegetation”. Ecology 73 (6), 2001–2006. Concepcion, R.N., Samar, E.D., 1995. Grasslands: development attributes, limitations and potentials. Paper Presented During the First National Grassland Congress, ERDB, College, Los Baños, Laguna, September 26–28. de Jong, W., 2010. Forest rehabilitation and its implication for forest transition theory. Biotropica 42 (1), 3–9. DiTomaso, J.M., Kyser, G.B., Hastings, M.S., 1999. Prescribed burning for control of yellow starthistle (Centaurea solstitialis) and enhanced native plant diversity. Weed Sci. 47, 233–242. Dombeck, M.P., Williams, J.E., Wood, C.A., 2004. Wildfire policy and public lands: integrating scientific understanding with social concerns across landscapes. Conserv. Biol. 18 (4), 883–889. Faulkner, J.L., Clesch, E.E.C., Sanders, W.L., 1989. Use of prescribed burning for managing natural and historic resources in Chickamauga and Chattanooga National Military Park, USA. Environ. Manage. 13 (5), 603–612. FAO (Food and Agriculture Organization of the United Nations). 2013. “FAOSTAT Database.” Accessed (January 2014). http://faostat.fao.org/site/291/default.aspx. Florece, L.M., 2010. Rural people and DENR personnel's views and perceptions on reforestation in the Philippines. Journal of Nature Studies 9 (1), 59–68. Fuhlendorf, S.D., Engle, D.M., 2004. Application of the fire-grazing interaction to restore a shifting mosaic on tallgrass prairie. J. Appl. Ecol. 41 (4), 604–614. Fuhlendorf, S.D., Smiens, F.E., 1999. Scaling effects of grazing in semi-arid savanna. J. Veg. Sci. 10 (5), 731–738. Garrity, D., M. Soekardi, M. Van Noordwijk, R. De la Cruz, P.S. Pathak, H.P.M Gunasena, N. Van So, G. Huijun, and N.M. Majid. 1996. “The imperata grassland of tropical Asia: area, distribution and typology.” Agrofor. Syst. 36 (13): 3–29. Goldammer, J.G., Penafiel, S.R., 1990. Fire in the pine-grassland biomes of tropical and subtropical Asia. In: Goldammer, J.G. (Ed.), Fire in the Tropical Biota: Ecosystem Processes and Global Challenges. Springer-Verlag, Berlin. Harrison, S.R., Emtage, N.F., Nasayao, B.E., 2004. Past and present forestry support programs in the Philipines, and lessons for the future. Small Scale For. Econ. Manag. Policy 3 (3), 303–317. Israel, D.C., Lintag, J.H., 2013. Assessment of the Efficiency and Effectiveness of the Reforestation Program of the Department of Environment and Natural Resources Final Report. Philippine Institute for Development Studies, Manila. Justice, C., Giglio, L., Boschetti, L., Roy, D., Csiszar, I., Morisette, J., Kaufman, Y., 2006. MODIS Fire Product: Algorithm Theoretical Basis Documentation (Ver. 2.3). NASA/EOS ATBD. Kugbe, J.X., Fosu, M., Desta, T.L., Denich, M., Vlek, P.L.G., 2012. Annual vegetation burns across the northern savanna region of Ghana: period of occurrence, area burns, nutrient losses and emissions. Nutr. Cycl. Agroecosyst. 93 (3), 265–284. Lasco, R.D., 2008. The reforestation value chain for the Philippines. In: Snelder, D.J., Lasco, R.D. (Eds.), Smallholder Tree Growing for Rural Development and Environmental Services: Lessons from Asia. Springer, The Netherlands, pp. 193–206. Le, H.D., Smith, C., Herbohn, J., 2014. What drives the success of reforestation projects in tropical developing countries? The case of the Philippines. Glob. Environ. Chang. 24, 334–348. LP DAAC (Land Processes Distributed Active Archive Center). 2014. “Land Processes Distributed Active Archive Center.” Accessed (January 2015). https://lpdaac.usgs.gov/. Magcale-Macandog, D. and W. Galinada, W. 1998. “Fire occurrence in grassland-based smallholdings.” In Improving Smallholder Farming Systems in Imperata Areas of

45

Southeast Asia: Alternatives to Shifting Cultivation edited by K. Menz, D. MagcaleMacandog, and I.W. Rusastra. Canberra, Australia: Australian Centre for International Agricultural Research. Mittermeir, R.A., 1997. Megadiversity: Earth's Biologically Wealthiest Nations. CEMEX/ Agrupación Sierra Madre, Mexico City. National Mapping and Resource Information Authority (NAMRIA), 2010A. Land Cover for the Philippines (Manila, Philippines). PAGASA-DOST (Philippine Atmospheric, Geophysical and Astronomical Services Administration-Department of Science and Technology). 2014. “Climatological Normals.” Accessed September 3, 2014. http://kidlat.pagasa.dost.gov.ph/index.php/ climate/climatological-normals. Peras, R.J., Pulhin, J.M., Lasco, R.D., Cruz, R.V., Pulhin, F.B., 2009. Climate variability and extremes in the Pantabangan–Carranglan watershed, Philippines: assessment of impacts and adaptation practices. Journal of Environmental Science and Management 11 (2), 14–31. Pezzatti, G.B., Zumbrunnen, T., Burgi, M., Ambrosetti, P., Conedera, M., 2013. Fire regime shifts as a consequence of fire policy and socio-economic development: an analysis bsed on the change point approach. Forest Policy and Economics 29, 7–18. PhilGIS (Philippine GIS Data Clearinghouse). 2013. “Country Vector and Raster Datasets.” Accessed September 20, 2014. http://philgis.org/freegisdata.htm. Prasad, V.K., Badarinath, K.V.S., Eaturu, A., 2007. Biophysical and anthropogenic controls of forest fires in the Deccan Plateau, India. J. of Envi. Mgt. 861, 1–13. Prestemon, J.P., Pye, J.M., Butry, D.T., Holmes, T.P., Mercer, D.E., 2001. Understanding broadscale wildfire risks in a human-dominated landscape. For. Sci. 48 (4), 685–693. Rebugio, L.L., Pulhin, M., Carandang, J., Peralta, A.P., Camacho, E.O., Bantayan, L.D.N.C., 2007. Forest restoration and rehabilitation in the Philippines. In: Lee, D.K. (Ed.), Keep Asia Green: Volume 1 Southeast Asia. International Union of Forest Research Organizations (IUFRO), Vienna, Austria. Rola, A., 2011. An upland community in transition: institutional innovations for sustainable development in rural Philippines. Los Banos, Laguna. SEARCA and ISEAS, Philippines. Sayer, J., Chokkalingam, U., Poulsen, J., 2004. The restoration of forest biodiversity and ecological values. For. Ecol. Manage. 201 (1), 3–11. Schulte, S., Miller, K.A., 2010. Wildfire risk and climate change: the influence on homeowner mitigation behavior in the wildland–urban interface. Soc. Nat. Resour. 23 (5), 417–435. Setterfield, S.A., 2002. Seedling establishment in an Australian tropical savanna: effects of seed supply, soil disturbance and fire. J. Appl. Ecol. 39 (6), 949–959. Sturtevant, B., Cleland, D., 2007. Human and biophysical factors influencing modern fire disturbance in Northern Wisconsin. Int'l. J. of Wildland Fire 16, 398–413. Sturtevant, B.R., Miranda, B.R., Yiang, J., He, H.S., Gustafson, E.J., Scheller, R.M., 2009. Studying fire mitigation strategies in multi-ownership landscapes: balancing the management of fire-dependent ecosystems and fire risks. Ecosystems 12 (3), 445–461. Tran, A.T., Dinh, N.D., Nguyen, D.T.H., Phonekeo, V., 2008. Forest fire risk mapping by using satellite imagery and GIS for Quang Ninh Province, Vietnam. Paper Presented at the Asian Conference on Remote Sensing (ACRS) http://www.a-a-r-s.org/acrs/ proceeding/ACRS2008/Papers/PS%202.8.pdf. Vadrevu, K.P., Eaturu, A., Badarinath, K.V.S., 2009. Fire risk evaluation using multicriteria analysis—a case study. Environ Monit Assess 166, 223–239. Valente, S., Coelho, C., Ribiero, C., Liniger, H., Schwilch, G., Figueiredo, E., Bachmann, F., 2015. How much management is enough? Stakeholder views on forest management in fire-prone areas in central Portugal. Forest Policy and Economics 53, 1–11. Van Noordwijk, M., Hairiah, K., Partoharjono, S., Labios, R.V., Garrity, D.P., 1996. Foodcrop-based production systems as sustainable alternatives for Imperata grasslands? Agrofor. Syst. 36 (1–3), 55–82. Zanne, A.E., Chapman, C.A., 2001. Expediting reforestation in tropical grasslands: distance and isolation from seed sources in plantations. Ecol. Appl. 11 (6), 1610–1621.