Quantitative Assessment of Climate Change Impacts On Forest Ecosystem


 Characterizing and predicting the response of terrestrial ecosystems to global change is one of the key challenges of contemporary ecology and ecological conservation. The impact of climate change on forest ecosystem has been widely studied, but it rarely uses method of multi-index fusion for quantitative evaluation. In this study, forest ecosystem in Heilongjiang Province was investigated. Based on remote sensing, meteorological observation, ground survey, geographic information, MAXENT model, CASA model, carbon sequestration potential model of Zhou Guangsheng-Zhang Xinshi, pixel dichotomy model and Savitzky Golay Filter model were used. On this basis, we analysed the change characteristics of forest distribution, net primary productivity and vegetation coverage. We also established a model for evaluating the impact of forest ecosystem change on climate, and made a quantitative assessment of the impact on climate. Our results indicate the following: (1) From 2001 to 2019, the forest area in Heilongjiang Province ranged from 2.34 × 105 to 2.46 × 105 km2, the forest NPP ranged from 40.48 to 555.32 gC/m2/a, and the vegetation coverage ranged from 42.42% to 67.64%, both of which showed a significant upward trend; (2) The values of forest ecological role were significantly positively correlated with the climatic potential; (3) The results of climate impact assessment of forest ecosystem change showed the contribution rate of climate change to forest ecosystem change was negatively correlated with forest coverage, which varied from 4.79% to 18.07% in different regions (cities) of the province. This study contribute to improving evaluating influence of climate change on forest ecosystem.


Quantitative Assessment of Climate Change Impacts on Forest Ecosystem
8 Characterizing and predicting the response of terrestrial ecosystems to global 9 change is one of the key challenges of contemporary ecology and ecological 10 conservation. The impact of climate change on forest ecosystem has been widely 11 studied, but it rarely uses method of multi-index fusion for quantitative 12 evaluation. In this study, forest ecosystem in Heilongjiang Province was 13 investigated. Based on remote sensing, meteorological observation, ground ecosystem is one of the main terrestrial ecosystems, and it is also the most complex terrestrial 48 ecosystem. It has high biological productivity and biomass (Melillo et al, 1993), as well as rich 49 biodiversity (Tilman et al, 1996;Nadrowski et al, 2010;Yu et al, 2008). Ecosystem distribution is 50 an important signal of forest ecosystem status. There are large areas of forest vulnerability in 51 Northeast and Southwest China. The subtropical evergreen deciduous broad-leaved mixed forest, 52 cold temperate mountain coniferous forest and temperate deciduous broad-leaved mixed forest 53 become more vulnerable under climate Change (Wan et al, 2018). The effect of temperature on 54 distributing plant species in forest-steppe ecotone of northern (Liu et al, 2015) and boreal forest 55 (Wu et al, 2017) in northern China was greater than that of precipitation. Air temperature 56 increasing obviously effected the ecotone of alpine coniferous forests. The areas of suitable 57 distribution regions for alpine tundra, subalpine forest, cold-temperate coniferous forest, and 58 temperate mixed forest decreased continuously; however, the areas for warm-temperate deciduous 59 broad-leaved forest and temperate grassland increased (Liu et al, 2017 Figure 1 Average yearly temperature, precipitation (a) and topography (b) in the study area 123 124 3. Data collection and methodology 125 126 3.1 NDVI data set reconstruction 127 128 In this study, we used the Global MOD13Q1 data as the basic data source of land use classification.

129
The data come from EOS data center of NASA LPDAAC (The Land Processes Distributed Active 130 Center). Which is level-3 product in the Sinusoidal projection with 250-meter resolution. The 131 coverage of each area is 10°×10° lat/long, and the data areas used in Heilongjiang Province are 132 h25v03, h25v04, h26v03 and h26v04. The data contains Normalized Difference Vegetation Index 133 (NDVI) and Enhanced Vegetation Index (EVI) data sets. In this study, MODIS Reprojection 134 Tool was used for stitching and projection conversion (conversion to WGS-1984-UTM projection).

135
The temporal spectral data of vegetation index can reflect the dynamic of vegetation growth, 136 and high quality NDVI time series data is of great significance for regional and global ecological 137 and  The meteorological data derived from the daily data of 80 meteorological stations in Heilongjiang 163 Province from 1951 to 2019, including daily average temperature, daily precipitation, daily 164 average relative humidity and other basic meteorological. Climate change research must be based 165 on reliable data. However, due to the influence of station migration and others, most of the 166 measured data are not uniform in sequence, which will cover up the reality and produce false 167 climate change. It is necessary to check the uniformity and quality control of the data before 168 conducting climate change analysis to maximize the reliability of the data (WEI F, 1999; REN et 169 al, 1998).

170
We used SPSS 22 to conduct normal test for the daily temperature and relative humidity data 171 in each climate zone. The significance level was P ≥0.05 which indicating most of the daily 172 temperature and relative humidity data in each climate zone were approximately normal 173 distribution, and no standardized processing was required in the analysis. The change of daily 174 precipitation does not have gradual and continuous characteristics. It's not the independent 175 variable to judge whether the precipitation data obey the normal distribution, but the monthly 176 precipitation is the independent variable. The test results show that most of the temperature and 177 precipitation data used in this study are subject to normal distribution. Therefore, no 178 standardization is done when further analysis is not required. is below 5℃ and the average temperature is below 10℃ in the hottest month. Deciduous woody 186 plants in frigid zone cannot survive when the maximum temperature is higher than 5℃ in the 187 coldest month and the average temperature is higher than 21℃ in the hottest month. We 188 According The forest distribution range in the study area was extracted based on the classification results of 202 "IGBP Global Vegetation Classification Scheme" of MODIS MCD12Q1 data. Forests consist of 203 the following nine categories: (1) Dominated by evergreen conifer trees (canopy >2m). Tree 204 cover >60%, (2) Dominated by evergreen broadleaf and palmate trees (canopy >2m). Tree 205 cover >60%, (3) Dominated by deciduous needleleaf (larch) trees (canopy >2m). Tree cover >60%, 206 (4) Dominated by deciduous broadleaf trees (canopy >2m). Tree cover >60%, (5) Dominated by 207 neither deciduous nor evergreen (40-60% of each) tree type (canopy >2m). Tree cover >60%, (6) 208 Dominated by woody perennials (1-2m height) >60% cover, (7) Dominated by woody perennials 209 (1-2m height) 10-60% cover, (8) Tree cover 30-60% (canopy >2m), (9) Tree cover 10-30% 210 (canopy >2m).

211
Landsat series data of the same period were used to test the classification accuracy of the 212 combined data. We analyzed the classification accuracy of land use types (forest and other land 213 use types) from 2001 to 2019 by using error matrix and Kappa analysis method based on LANDSA 214 data of the year. The overall classification accuracy was more than 85%, and the Kappa coefficient 215 was more than 0.82. have little influences. The AUC is the highest which the characteristic parameter is "threshold" 234 and the control frequency is 1. Jackknife test showed that the main temperature factors affecting 235 forest distribution were the average temperature of the warmest season, the annual average 236 temperature and the lowest temperature of the coldest month. And the main water factors were 237 annual precipitation, seasonal variation of precipitation and the driest season precipitation. The vitality of an ecosystem reflects its carrying capacity and anti-disturbance ability. We (1) Monthly vegetation coverage simulation model based on integrated eco-meteorological: Where, n is actual sunshine duration; N is maximum sunshine duration; Ra is exoatmospheric 294 solar radiation; a and b are fitting coefficient。 295 PAR was calculated according to the ratio of PAR to RS of 0.48: 296 (4) Temperature stress coefficient 298 Where, Tε is temperature stress coefficient; Ta is average monthly temperature (Unit: oC); Tmin、 300 T max and T opt are the minimum, maximum and optimum temperatures for photosynthesis, 301 respectively (Melillo et al，1993).

302
(5) Water Stress Coefficient 303 The ratio of monthly actual evapotranspiration (E) to potential evapotranspiration (PET) was used 304 to estimate the water stress coefficient (W), and the calculation formula is as follows: 305 Where, Δ is saturation vapor pressure gradient; γ is humidity calculation constant; R n is monthly 308 net radiation，and the calculation formula is as follows: 309 Where, α is surface albedo; R nl is net long-wave radiation 311 312 We combine the Analytic Hierachy process (AHP) (Saaty, 1977;1980) and Shannon Entropy 332

Evaluation model of meteorological impact of forest ecosystem change
Index (Shannon, 1948) to determine the model weight. Which not only avoids the subjectivity of 333 the subjective weighting method, but also avoids the randomness of the objective weighting 334 method. Perform consistency test on the weight results calculated by the above two methods, and 335 calculate the combination weight after passing the consistency test. See Table 1  Information Center of the Chinese Academy of Sciences, with the spatial resolution of 90m. 365 The provincial administrative division data required by the study comes from the 1:250000 366 basic geographic information issued by the China Meteorological Administration. It's 367 topologically checked to remove the gaps between provincial boundaries and county boundaries. 368 The provincial administrative division data and the location data of meteorological observation The suitable distribution area of the forest was between 3. 72 × 10 5 and 4. 14 × 105 km 2 , with 380 an average of 3. 93 × 10 5 km 2 , which had no significant change trend and no significant correlation 381 with the actual distribution area of the forest. The actual distribution area of forest in the study 382 area may be affected by both meteorological and non-meteorological conditions. 383  Figure 4-b). Figure 4-c shows that most areas of the province, except the southwest, 396 are suitable for forest distribution. And that all the forests in the province are distributed in the 397 suitable distribution area, with 82.82% of the area suitable for forest distribution in the long term 398 (11 -19 years) (Figure 4-d). And the actual distribution area of forests is highly consistent with 399 the potential distribution area. The NPP (ranges from 440.48 to 555.32gC/m 2 /a with an average of 507.05gC/m 2 /a) is generally 411 slightly lower than the NPP climate potential (ranges from 460.38 to 577.62gC/m 2 /a with an 412 average of 515.67gC/m 2 /a,). They all showed a significant increase trend of 4.71 gC/m 2 /a and 413 3.92gC/m 2 /a respectively ( Figure 5) (Figure 6-a). The 420 regional distribution characteristics of NPP climate potential (Figure 6-b) was consistent with the 421 forest NPP. The climatic potential of NPP ranged from 392.64 to 628.23gC/m 2 /a and mainly 422 concentrated in the vicinity of 462.13gC/m 2 /a, followed 560gC/m 2 /a.

423
The forest NPP in most regions is lower than the NPP climatic potential (Figure 6-e), but 424 some regions of the eastern mountains exceeds the climatic potential which is calculated based on 425 meteorological observation data. One of the reason may be that the interpolation accuracy of 426 meteorological data will be affected under the condition of complex terrain, which will affect the 427 estimation accuracy of NPP. The NPP (Figure 6 -C) and NPP climatic potential (Figure 6 -d) of 428 most forests showed increasing trends, with the largest increasing trends at 4.21 gC/m2/a and 2.98 429 gC/m2/a, respectively. the increasing trend of forest actual NPP was greater than the increasing 430 trend of NPP climate potential in most areas (Figure 6 -f)  of forest NPP and the real forest NPP, calculated as the climate potential of forest NPP (Fig. 6-b) minus 439 the real forest NPP (Fig. 6-a) c d e f The vegetation coverage (ranged from 42.42% to 67.64%) was generally lower than the climate 446 potential of vegetation coverage (ranged from 58.34% to 72.71%). And there was a significant upward 447 trend of 0.51%/a and 0. 49%/a respectively. was less than that of realistic coverage in most of the province, but the opposite has occurred in 461 parts of the Lesser Khingan Range (Figure 8-   Because the NPP climate potential, vegetation coverage climate potential and potential 481 distribution area of forest are all calculated based on meteorological factors, which can represent 482 the climate potential of each ecological function driven by meteorological factors, this study 483 analyzed the relationship between the actual value of each ecological function and the potential 484 climate potential. 485 Fig. 9 shows that the actual values of forest ecological function elements are significantly 486 positively correlated with the climate potential. The NPP climate potential was calculated based 487 on annual precipitation and average daily temperature from 0℃ to 30℃. Therefore, the actual The linear fitting relation of NPP climat potential and realistic NPP was y=0.49x+221.81, 505 Coefficient of determination (Figure 9- The contribution rate of meteorological conditions to forest ecosystem changes is 11.08% in the 517 province, and the contribution rate is between 4.79% (Mudanjiang) and 18.07% (Daqing). 518 Considering the forest coverage and location of each city, it can be seen that the forest coverage 519 in Songnen Plain (0.01% -43.22%) and Sanjiang Plain (17.24% -42.37%) is relatively low, and 520 the contribution rate of meteorological conditions to forest ecosystem change is relatively high, 521 with an average of 14.29% and 11.25%, respectively. However, the urban forest coverage in the 522 Greater Kingan Mountains (96. 36%), Lesser Kingan Mountains (44. 38% -91. 92%) and Eastern 523 Mountains (71. 09%) was higher, and the contribution rate of meteorological conditions was 524 relatively low, averaging 8. 96%, 8.27% and 4. 79%, respectively. 525 526 Figure 10 Contribution rate of meteorological conditions (gray Bar) and forest coverage (red Bar) of 527 forest ecosystems in 13 cities of Heilongjiang Province 528 529 Forest coverage is calculated as the percentage of forest area in the total area of a region. 530 There was a significant negative correlation between the contribution rate of meteorological 531 conditions (y) and forest coverage (X), and the linear regression equation was y = -5.94x + 107.67. 532 Coefficient of determination was 0.4748, Prob=0.009, and sample points participating in linear 533 fitting were 13. 534 535 5. Discussions 536 537

Influence of climatic factors on NPP 538 539
One of the concerns of our research is the impact of climate factors on NPP. The results 540 showed that the forest NPP in Heilongjiang Province increased significantly from 2001 to 2019. 541 The coupling effect of annual precipitation and daily mean temperature (0-30 ℃) was the main 542 meteorological factors of forest NPP change in the province, and the contribution rate was 16. 543 31%. This result was much lower than the conclusion reached by Wang  Due to the impact of climate change, boreal coniferous forest in Heilongjiang Province will face 573 the severe challenge of being replaced by other biological communities, and the distribution area 574 may be reduced, but this does not mean that the potential distribution area of boreal forest 575 ecosystem will be reduced (Liu et al, 2017). In recent years, rising temperatures have led to a 576 sustained and island-like degradation of permafrost in northeastern China. This change is pushing 577 northern ecosystems into an unbalanced state, which may affect the relative role of climate factors 578 and fire in determining vegetation distribution. The results of this study showed that although the 579 actual distribution area of forest in Heilongjiang Province was distributed in the potential 580 distribution area (4.1.2 Figure 4), there was a significant correlation between the proportion of 581 actual distribution area and the proportion of potential distribution area (4.4.1 Figure 9-c). 582 However, the actual distribution area of forests in the province is not consistent with the potential 583 distribution area (Figure 3 of Section 4.1.1). This shows that the actual distribution area of forest 584 is not only affected by meteorological conditions, but also strongly disturbed by fire, outbreak of 585 pests and diseases, human production activities and so on, but the results of such disturbances are 586 not enough to completely change the distribution trend of forest ecosystem in a larger area. 587 588

Influence of climatic factors on vegetation coverage 589 590
The results show that the average vegetation coverage of the province has a significant 591 increasing trend from 2001 to 2019, and the contribution rate of climate change to the change of 592 vegetation coverage is 32.61%. On the one hand, climate warming can promote vegetation growth 593 in cold areas or high altitude areas in the north; On the other hand, the increase of vegetation cover 594 will also have an adverse effect on land surface temperature. The increase of vegetation coverage 595 reduces the background (no snow) and snow-covered surface albedo, resulting in a significant 596 increase in surface absorption of solar radiation, and amplifies the feedback between snow cover, 597 surface albedo and absorbed solar radiation (Zhang et al, 2007). Snow-vegetation interaction 598 warms northern land in spring, resulting in a rapid increase of vegetation coverage in spring and 599 prolongs the length of growing season (Peng et al, 2011). From the point of view of heat, it 600 provides favorable conditions for the improvement of vegetation coverage. In addition, although 601 the increase of vegetation coverage reduces the intensity of soil evaporation, it increases the 602 vegetation transpiration, even if there is no significant impact on the total precipitation, but it may 603 change the pattern of precipitation, so that the precipitation in some areas decreases, resulting in 604 the reduction of vegetation coverage. Which may also be one of the reasons for the decrease of a 605 small amount of forest vegetation coverage in the southern part of the Lesser Kinggan Mountains 606 and the northern slope of the Eastern Mountains (Figure 8-c). 607 608 Influence of climatic factors on forest ecosystem 609 610 Climate is one of the main factors affecting the distribution pattern and functional 611 characteristics of terrestrial vegetation types, which affects the composition of biological 612 communities in ecosystems by affecting physiological processes such as photosynthesis, 613 respiration and phenology of vegetation, thus changing vegetation distribution, NPP and 614 vegetation coverage. Due to the inconsistency of vegetation coverage and NPP changes, there is 615 great uncertainty in assessing the climate driving effect by using a single index (Ding et al, 2020). 616 In this study, the impacts of climate change on forest ecosystems were assessed with different 617 weights by integrating forest distribution, NPP and vegetation coverage indexes, which reducing 618 the uncertainty generated by single index assessment to a certain extent. The results showed that 619 the contribution rate of meteorological conditions to forest ecosystem change ranged from 4. 79% 620 to 18. 07%, and there was a significant negative correlation between climate contribution rate and 621 forest coverage. From the perspective of landscape ecology, the larger the patch area of the forest 622 type, the more conducive it will be to the abundance and quantity of species, the extension and 623 interconnection of the food chain, and the reproduction of the secondary species, so as to gain 624 greater anti-interference and restoration ability. Therefore, for the same external disturbance, areas 625 with higher forest cover rate have a lower impact on their ecological functions than those with 626 lower forest cover rate. 627 628 6. Conclusion 629 630 From 2001 to 2019, the forest area in Heilongjiang Province showed a trend of increasing first 631 and then decreasing, and both NPP and vegetation coverage showed a significant upward trend. 632 The contribution rate of meteorological conditions to forest ecosystem change varies from 4.79% 633 to 18.07% in different cities. There was a negative correlation between the impact of 634 meteorological conditions on forest ecosystem and forest coverage, that is, the higher the forest