Targeting area and comparing the effect of different land use/land cover (LULC) scenarios on greenhouse gases (GHGs) emission reduction (Case study: Hyrcanian forests in Iran)

Background: Because the greenhouse gases (GHGs) emissions are known to be strongly influenced by land use/land cover (LULC) change, deforestation and degradation (REDD) mechanism has attracted much attention as a strategy for understanding how different LULC scenarios effect on the GHGs emissions. Transition to other LULC types is of the major challenges of Iran's Hyrcanian forests Golestan province. To consider how LULC change scenarios affect GHGs, REDD project was executed in a period of 30 years (2018- 2048) at intervals of 5 years. In this regard, study area was divided into the project area and leakage belt based on the Multi Criteria Evaluation (MCE) derived forest suitability map. In the baseline scenario, it was assumed that the trend of past LULC changes will continue. Results: By implementation of the project scenario, some degradation activities were controlled. Project scenario was executed with different project success rates (PSR) of 90, 80, 70, 60 and 50% to examine its efficiency rate in reducing GHGs emissions. According to the results, 38206.8 hectares of forests within the project area will be destroyed by 2047 under the baseline. The destroyed area will be reach 39784.4 hectares in the leakage belt. The highest rate of forest destruction in the project area will occur in the last 5 years (1352 hectares per year), so the highest CO 2 and non-CO 2 emissions equal to 662655.3 tons/year and 278.94 tCO 2 e/year will happen in the last 5 years (2042-2047). Based on the results, reducing the PSR affected the efficiency of the project scenario. The highest and lowest rates of emissions reduction were observed respectively with PSR of 90 and 50%. Conclusions: That's very important for developing countries especially Iran that are facing many challenging forest conservation decisions. This study innovated in methodology by integrating the MCE into the REDD steps. The MCE as a spatial targeting method could be applied to increase the efficiency of the REDD project, as we illustrated for the case of Hyrcanian forests.


45
Forest ecosystems provide a wide range of multiple ecosystem services (ESs) that are 46 important for sustaining life on earth and maintaining the integrity of the ecosystems 47 (Bauhus et al., 2010;Gamfeldt et al., 2013;Miura et al., 2015;Mri, 2017;Tolessa et al., 48 2017). One of the most important forest ESs is climate regulation (Costanza et al., 2017;49 Chu et al., 2019). On one hand, carbon accumulates through growth of trees into forest 50 growing stock. On the other hand, Land use/land cover (LULC) change activities impact 51 carbon stocks (Vauhkonen and Packalen, 2018). By continuing global forest decline 52 (Köthke et al., 2013), FAO (2010) predicts that the current annual global forest loss is 53 about 13 million hectares. The Intergovernmental Panel on Climate Change's fourth 54 assessment report (IPCC, 2014) estimated that agriculture, forestry and other LULC 55 changes specially deforestation contribute 24% of global anthropogenic greenhouse 56 gases (GHGs). As between the eras from 1750 to 2011, about 180 PgC was released to 57 the atmosphere due to LULC change, mainly deforestation (IPCC 2014). In the absence 58 of potential land mitigation and adaptation policies, climate change can affects many 59 parts of the environment and multiple ESs (Etemadi et al., 2012). 60 Therefore, reducing land use related GHGs emissions represents a significant climate 61 change mitigation strategy (Collen et al., 2016). One approach to doing so emerged in 62 2005, by the Reducing Emissions from Deforestation (RED) program (Pistorius, 2012). 63 The United Nations Framework Convention on Climate Change (UNFCCC) introduced 64 RED as a simple monetary mechanism for reducing forest related carbon emissions in 65 developing countries (UNFCCC, 2005). Over the past few years, the scope of RED 66 mechanism was notably extended and now includes forest degradation (the second D in 67 REDD). Also plus activities including sustainable forest management, conservation of 68 forest carbon stocks, enhancement of forest carbon stocks and safeguard forest non 69 carbon values led to the REDD+ project (Vije, 2015). REDD is a global environmental 70 governance mechanism with the objective to slow and eventually halt deforestation and 71 forest degradation from LULC change in developing countries by providing an economic 72 incentive to keep carbon stored in vegetation and soils (Angelsen and Wertz-73 Kanounnikoff, 2008;Skutsch and Van Laake, 2008;Angelsen and Brockhaus, 2009;74 Parker et al., 2009;Arévalo et al., 2020). In recent years, REDD projects have attracted 75 much attention around the world as a policy to regulate climate change on a national 76 and regional scale. So far, many developing countries include Indonesia, Philippines,  Massarella et al., 2018;Guadalupe et al., 80 2018; Sheng et al., 2016). REDD mechanism requires information on LULC change and 81 carbon emission trends from the past to the present and into the future (Harris et al.,82 2012; Eastman, 2015;Capitani et al., 2019;Arévalo et al., 2020). Because the emissions 83 of GHGs are known to be strongly influenced by LULC change (Cooper et al., 2020; 84 Hundera et al., 2020), scenario analysis with LULC models can play a major role in 85 providing information to decision makers. By purpose of spatial targeting of REDD studies, LULC change models such as Land Change Modeler (LCM), Geomod, CA-87 MARCOV and CLUE-S have been used for the LULC changes prediction, specially forest 88 loss trend (Feng et al., 2020;Tang et al., 2020;Parsamehr et al., 2019;Mena et al., 2017;89 Bununu et al., 2016;Kim, 2010;Hewson et al., 2019;Redowan, 2019). Some models 90 such as LCM that permits the simulation of future scenario is integrated with a REDD 91 steps to determine and model anthropogenic GHGs emission reductions (Bununu et al., 92 2016). In addition to the importance of LULC change models, to increase the REDD 93 context efficiency, site selection is also one of the success determinants. In other words, 94 REDD as a least cost policy to achieve climate regulation, obviously depends on how and 95 where it is implemented (Blom et al., 2010;Lin et al., 2014;Atela et al., 2014). Amiri and Zargham, 2015) are belong to the Euro-Siberian biome (Zohary, 1973;101 Browicz, 1989). The high precipitation and mild climate of the Hyrcanian region 102 facilitates broad-leaved dense forests (Noroozi, 2020 Project (BioCF), was implemented in the nine steps under two baseline and project 131 scenarios (Fund, 2008). In the baseline scenario it was assumed that the trend of past 132 LULC changes especially forest loss will continue, and in the project scenario some of 133 the degradation and deforestation activities were controlled and stopped (Eastman,134 2014). The REDD steps are explained in detail in the next sections. In the first step of the REDD project, the temporal boundaries, carbon pools and the 138 spatial boundaries of the reference region, project area and leakage belt are defined 139 (Fund, 2008). The duration of the REDD methodology activity must be at least 20 years 140 (Fund, 2008). The spatial boundary of the reference region is the spatial delimitation of 141 the analytic domain from which information about rates, drivers and patterns of LULC 142 change will be obtained, projected into the future and monitored (Fund, 2008). The 143 project area is the area of land on which the project proponent will undertake the 144 project activities. In contrast, the leakage belt is the land adjacent to the project area in 145 which baseline activities are likely to be displaced from inside the project area (Fund,146 2008). Five carbon pools include above-ground, below-ground, dead wood, litter and 147 soil organic carbon are potentially eligible in REDD methodology (Fund, 2008). Because  The MCE methodology operates based a series of raster layers of environmental 158 parameters that are assumed to have significant influence on land suitability for a 159 specific LULC category (Mahiny and Clarke, 2012). There are several steps to conduct 160 the MCE analysis and it begins by specifying a collection of ecological and 161 socioeconomic factors that are deemed to influence a given LULC category. In this 162 regard, raster layers of such factors retain digital values that are quantified through 163 different measurement levels (i.e. nominal, ordinal, interval and ratio). Therefore, by 164 implementing the fuzzy set theory (Zadeh, 1965), factor layers could be fuzzified 165 (standardized) and become ready for map integration. The fuzzified scores within each 166 raster space allow map overlay procedure and acknowledge uncertainty in data layers 167 and also provide a means for compromising between different opinions on the 168 importance of each layer. Before map integration, factor layers are weighted to indicate 169 their relative importance. In this regard, the analytical hierarchy process (AHP) is 170 applied (Saaty, 1980) to establish a logical basis for conducting several pairwise 171 comparisons between factors and matrix computations (Table 2).

173
There is also a second category of digital map layers used in the MCE analysis that 174 depict absolutely unsuitable lands for the LULC category under study, called 'constraint' 175 layers, and they retain only 0 and 1 values to indicate impossibility and possibility, 176 respectively, of developing the targeted LULC category (Table 3).  Table 2  The goal of this step is to collect and analyze spatial data in order to identify current 199 LULC conditions and to analyze LULC change during the historical reference period 200 within the reference region, leakage belt and project area (Fund, 2008).  The goal of this step is understanding who is deforesting the forest (the agent) and 212 what drives LULC decisions (drivers and underlying causes) (Fund, 2005). The driver 213 variables in this study include (1) elevation, (2) slope, (3) aspect, (4) distance from 214 roads, (5) distance from forest edge, (6) distance from residential areas, (7) distance from agriculture, (8) distance from rangelands and (9) distance from water resources.

216
The elevation variable was prepared from a topographic map and Aster DEM data. Slope 217 and aspect variables were also extracted from DEM. Distance operator in GIS was used 218 for the generation of distance variables. V-Cramer statistic in the LCM was used to 219 investigate the predictive power of variables. As a rule, V-Cramer values above 0.15 and 220 0.40 are respectively appropriate and good (Eastman, 2015). The objective of this step as the core of the baseline scenario in REDD methodology is 223 to locate in space and time the baseline deforestation expected to occur within the 224 reference region, project area and leakage belt (Fund, 2005). Modeling the land use are carried out over a five years period (Fund, 2005 stock changes (C-Baseline) and non-CO2 emissions (Fund, 2005). For calculating the C-  year t is calculated as follows: Where ∆ is Total baseline carbon stock change at year t (tCO2e).

263
Conversion of forest to non-forest classes by fire is a source of non-CO2 gases 264 emissions (CH4 and N2O) (Fund, 2005). When sufficient data on such forest fires are 265 available from the historical reference period, emissions of non-CO2 gases from biomass 266 burning can be estimated (Fund, 2005). In order to obtain fire information in the forests Due to the importance of PSR and LR factors in GHGs emissions, several REDD 295 scenarios were determined in this step. In this regard, Efficiency Rate (ER) was defined 296 (Eq. 8). In any scenario, the values of PSR and LR factors were changed and the impact 297 of changes on reducing GHGs emissions were assessed ( based on previous studies as well as information contained in IPCC reports (Fund,302 2008). In order to sensitivity analysis of the model, the change of the forest carbon 303 stocks was examined in different states. In each state, forest carbon stocks were steadily 304 reduced or increased (Table 5). In the baseline state, the estimated amount of carbon 305 stocks in section 2.2.1 was used. In the other four states, 25 and 50% were added or 306 reduced to the amount of carbon stocks, respectively (Fig. 3). In this study, we demonstrated how to use a MCE-derived forest suitability map to 319 identify the most suitable areas for project area and leakage belt. Based on the value 320 range of the forest utility map, the amount of 335087.9 ha was allocated to the project 321 area and 203163.7 ha to the leakage belt ( Fig. 4.b). Without REDD implementation, 322 31443.8 ha of forests within project area would be destroyed until 2047 year. In order 323 to prevent of this problem, in the first step of REDD project, greater scores of forest 324 suitability map were allocated to project area and lower scores were assigned to 325 leakage belt. With this approach, more suitable forest areas were conserved as project 326 areas and the leakage belt was the land adjacent to the project area in which destructive 327 activities were likely to be displaced from inside the project area (Fund, 2008).  Comparing the two sub periods (1984-2008 and 2008-2018), the forest changes follow 363 a similar decreasing trend (Table 6). Although the deforestation rate is somewhat 364 slower in the second period (-2.1%/-2.8%). This problem showed that during the 365 second sub period (2008-2018), deforestation has not only occurred at the expense of 366 agriculture, but the other LULC classes have also grown. In contrast to the first period, 367 rangeland increased with positive growth rate (+0.9%/-0.19%).

368
Steps 3 to 5 were performed with LCM module in TerrSet software. In the baseline 369 scenario was assumed the continuation of deforestation change rates over the past 370 years  in the reference region, project area and leakage belt. Variables 371 include slope, elevation, distance from rangelands, distance from croplands, distance 372 from human made areas, distance from forest edges, distance from roads and distance 373 from rivers were significantly correlated with deforestation during 1984-2002 (Cramer 374 values>0.11) and used as explanatory variables for projection of the rate and location 375 of future deforestation (Fig.6). 376 <Fig. 6>

377
The variables slope, elevation and distance from rangeland were the most significant 378 according to the Cramer's V test (Table 7). The lowest result was obtained for the "aspect". This means that this variable had low significance in deforestation prediction.

380
These results confirm the findings of Shooshtari and Gholamalifard (2015) in the Neka 381 watershed, Shooshtari et al (2020) in the Ghara-su basin and Zabihi et al (2020)   The quantity of changes extracted from the Markov chain (Table 8) (Eastman, 2015). As studies on 412 LULC change analysis in north of Iran showed that Hyrcanian forests were the main 413 contributor to increase agriculture (Minaei and Kainz, 2016;Asadolahi et al., 2018;414 Nasiri et al., 2019;Aghsaei et al., 2020;Shooshtari et al, 2020). Studies in other parts of 415 the world also confirm that deforestation has taken place at the expense of agricultural 416 activities (Fernandes et al., 2020). In Golestan province like other areas in northern 417 Iran, most of deforestation was occurred surrounding forest areas (Shooshtari and  Fig. 9 shows CO2 emission rate within the project area under the baseline scenario.

430
Given that the most of forest destruction in the project area occurred in the last 5 years, 431 so the highest emission equal to 757119.1 tones occurred during 2042-2047. Table 9 432 indicates the reduction in CO2 emissions due to carbon sequestration by changing 433 forests to other LULC types. According to the results of Table 9, the net CO2 emission 434 rate within the project area was obtained from equation (4). Fig. 10 shows CO2 emission As shown in Fig. 13, C-REDD was estimated from equation (7) under different project 499 scenarios. According to the results, the highest and lowest rate of net CO2 emissions 500 reduction occurred respectively in the first and fifth scenarios with PSR, LR, and EC of 501 90, 10, 80 % and 50, 50, 0%, respectively. A similar trend was observed for net non-CO2 502 emissions reduction (Fig. 14).

503
The goal of developing less successful REDD project scenarios was to assess the 504 impact of the unsustainable land use policies. As inappropriate decision making 505 approaches reduced the success rate of the REDD project up to the baseline scenario.

534
Also this study innovated in methodology by integrating the MCE into the REDD steps.

535
The MCE as a spatial targeting method could be applied to increase the efficiency of the 536 REDD project, as we illustrated for the case of Hyrcanian forests. Our approach in using