Land use/land cover changes and regional climate over the Loess Plateau during 2001–2009. Part II: interrelationship from observations
- First Online:
- Cite this article as:
- Fan, X., Ma, Z., Yang, Q. et al. Climatic Change (2015) 129: 441. doi:10.1007/s10584-014-1068-5
- 1k Views
Afforestation efforts in China resulted in significant changes in vegetation coverage over the Loess Plateau during 2001–2009. While regional climate conditions dominate the distribution of major vegetation types, human activities, primarily afforestation/reforestation and the resultant land use/land cover (LULC) changes (LULCC) and their impacts, are the focus of this study. A new attribution method was developed and applied to observed data for investigating the interrelationships between climate variation and LULCC. Regional climate (temperature and precipitation) changes are attributed to climate variation and LULCC; LULCC is attributed to climate variation and human activities. Climate attribution analysis indicated a larger contribution ratio (based on comparison of standard deviations of each contributing factor-induced climate changes and that of total change) from climate variation than from LULCC (0.95 from climate variation vs. 0.35 from LULCC) for variations in temperature. Impacts on precipitation indicated more spatial variations than those on temperature. The spatial variation of LULCC impacts on precipitation implied that human activities might have larger impacts on precipitation in the region’s arid north than in its humid south. Using both leaf area index (LAI) and areal coverage of each of the major land types, LULCC attribution analysis suggested that LULCC observed in the 2000s resulted primarily from human activities rather than climate variations (0.99 contribution ratio from human activities vs. 0.26 from climate variation).
Interactions between land use/land cover (LULC) change (LULCC) and regional climate variations have been found to be complicated, yet at the same time very important for environmental adaptation and management (Pielke 2005; Bonan 2008; Liu et al. 2008). Investigations are urgently needed, particularly, to study the effects of government policies and human activities that have been undertaken in an attempt to mitigate some of the impacts resulting from the degradation of the natural environment (e.g., Tao et al. 2004; Hvistendahl 2012). An example of these is the well-known Chinese government-funded Grain for Green Project (GFGP), which encourages the return of terrace-farming croplands to forests/grasslands in order to restore their natural ecosystems (Tao et al. 2004). The GFGP project is widely adopted in the Loess Plateau region and throughout China. The regional climate of the Loess Plateau is mostly continental arid to semi-arid. Moreover, it has experienced a drying and warming trend since the 1970s (Ma and Fu 2006). These climatic conditions make the Loess Plateau one of the most sensitive ecological regions in China, which is fragile and susceptible to climate changes (Liang et al. 2008). This is the primary reason the region was selected as the initial location for the GFGP, and is also the region on which we focus our investigation.
It is well known that vegetation distribution is primarily determined by climatic and environmental factors such as temperature and precipitation, as well as elevation, accumulated sunshine, and winds (Zhang et al. 2006). Climate change has been found to affect ecosystems in many respects, including LULCC and changes in vegetation (Goward and Prince 1995; Theurillat and Guisan 2001). On the other hand, results also show the importance of LULC in land–surface–atmosphere interactions (Dirmeyer et al. 2010). It has been demonstrated that the state of LULC could impact mesoscale convective development and circulation, meso- and regional-scale heat and moisture fluxes, precipitation, and temperature (e.g., Clark and Arritt 1995; McPherson 2007; Mahmood et al. 2010). As a result, it is expected that LULCC would modify the existing nature of land–surface–atmosphere interactions and, consequently, weather and climate. Many local-, regional-, and global-scale studies have demonstrated this assertion (e.g., Chase et al. 2000; Chen et al. 2012).
Afforestation may cause different impacts pertinent to various spatial and temporal scales, as well as having both localized and remote impacts. Some of the local impacts include both warming and cooling, as well as drying and wetting, all resulting from modifications to physical, chemical, and biological properties such as albedo, surface roughness, soil water capacity, carbon flux, and biomass productivity. Remote impacts include, for example, increased precipitation in downwind areas (Jiao et al. 2005). Large-scale afforestation may also increase the severity of water shortages in arid regions (Cao et al. 2007).
Therefore, climate and LULC interact and cause concurrent mutual impacts on each other while climate variations and LULCC also result from external forcings and human activities. However, these intermingling factors and interacting processes results in only one set of reality. It is always desirable but challenging to isolate and recognize the role of each individual factor or process so that applicable strategies may be devised toward a sustainable climate and environment. One widely used approach is numerical modeling using general circulation models (GCMs) in single-forcing sensitivity experiments. For example, Bonan (1997), Fu (2003), Pielke (2005), Pitman et al. (2009), and Avila et al. (2012) studied impacts of LULC as single forcing on climate simulations using GCMs. Kumar et al. (2013) devised a method to identify LULC impacts on climate from multiple GCM simulations. These studies have either prescribed LULCCs for sensitivity experiments or utilized difference of present and preindustrial LULC conditions to represent the LULCC, most commonly, cropland fractions (e.g., Noblet-Ducoudre et al. 2012). However, the attribution of the changes remains a challenge due to the limited resolution and capability of GCMs to take into consideration of the dynamically and concurrently changing LULC and climate, especially for the recent decade afforestation efforts and climate change in the study area of China.
Given that afforestation and its associated changes in LULC and regional climate have already taken place, this study is to: 1) analyze observed climate and LULC data to reveal their interrelationships; 2) devise a method to attribute: a) changes in LULC to either climate change and/or human activities (agricultural, afforestation/deforestation, etc.) and b) changes in regional climate to either natural climate variations and/or human activity-induced LULCC; and 3) apply the method to observational data directly to extract information about these changes.
2 Data and methodology
In this study, we focus on the study area shown in Fig. 1a, which covers the same geographical area as in FAN14. Specifically, it encloses the latitude-longitude region of (32–41°N, 102–114°E) and encompasses roughly 112.2 × 104 km2 over the Loess Plateau. Following FAN14, Moderate Resolution Imaging Spectroradiometer (MODIS) LULC and leaf area index (LAI) data are used in this study, which were available from 2001 to 2009 at 1-km horizontal and 8-day temporal resolutions (Strahler et al. 1999). The dimensionless LAI data indicate the amount of available leaf material and, thus, can be used as an indicator of the vigor of vegetative growth.
In addition, Normalized Difference Vegetation Index (NDVI) time series, derived from the Advanced Very High Resolution Radiometer (AVHRR) on board the National Oceanic and Atmospheric Administration (NOAA) polar-orbiting satellites, were also used (Goward et al. 1991). NDVI is an indicator of the vegetative greenness of an area and, therefore, changes in NDVI can reveal changes in vegetation. NDVI data were available from 1981 to 2006 at 8-km horizontal and 15-day temporal resolutions (Goward et al. 1991).
Station-observed precipitation and surface air temperature data, obtained from the China Meteorological Administration, were also used. For each station, its LULC type, LAI, and areal coverage of a particular LULC type is determined from the 1-km pixels within a distance of 25 km from the station. Specifically, LULC type for a station is assigned the dominant MODIS LULC type among the pixels; LAI uses the average LAI of the pixels; and areal coverage of any LULC type is represented by the number of pixels of that type.
2.2 Interrelationship analysis
FAN14 reported that the large increase in vegetation coverage recently observed on the Loess Plateau indicated the effects of both climatic and human influences. Similarly, regional climate variation may be influenced by both natural climatic fluctuations as well as human activities. Within the scope of this study, we focus only on human-induced LULCC as the human activity under consideration.
In order to analyze their interrelationship, and ideally to isolate the impacts of natural climatic variations and human-induced changes, we take a simple approach to extract information directly from observations. Here, we are dealing with two challenges in properly attributing the impacts, each with two contributing factors. 1) Climate variation attribution: climate variations are represented by temperature (T) and precipitation (P). The two contributing factors are natural climatic variations in T and/or P, along with human-induced LULCC. 2) LULCC attribution: LULCCs are represented by changes in LAI or areal coverage. The two contributing factors are natural climate variations and human-induced changes. We tackle these two attributions using two different scenarios. The first focuses on climate variation attribution within different zones of LULC; the other focuses on LULCC attribution under different climate conditions.
Although the vegetative cover changed over the 9-year period 2001–09, an obvious pattern exists where major LULC types consistently dominated and were sustained in certain areas (Fig. 1a). According to the spatial distribution of these major LULC types, the study area can be divided into four primary zones as shown in Fig. 1a, where B is used for barren or sparsely vegetated areas, G for grasslands, C for croplands, and F for forests. The areas in between these four zones form the three transition zones T1, T2, and T3 along the borders of B and G, G and C, and C and F, respectively. Transition zone T1 is primarily covered by open shrublands, T2 by closed shrublands, and T3 by woody savannah. According to FAN14, most of the changes in LULC occurred in these transition zones where the two major LULC coverages on either side of a transition zone advanced or retreated for a given year.
Also shown in Fig. 1a are the weather stations from which temperature and precipitation data were used in this study. Stations where LULC types remained unchanged throughout 2001–09 are marked by black plus signs, whereas those stations where LULC types did change, mostly along the transition zones, are marked by red dots.
2.2.1 Climate variation attribution
The attribution method of Kumar et al. (2013) avoided large number of sensitivity modeling experiments by comparing neighboring changing and non-changing LULC regions’ responses to a same all-forcing experiment. However, only a representative cropland-change was used and the dynamical change of LULC and its interaction with climate was neglected due to the nature of prescribing the LULC in the GCMs (though the interactions between a given LULC and the atmosphere are considered in most climate models). The method proposed below takes into consideration of the reality that LULC change occurs concurrently with climate change, which intermingled with human interference.
All of the above calculations, derived for temperature (T), are readily applicable to precipitation (P) as a climate variable.
2.2.2 LULCC attribution
Climate-induced LULCC (i.e., LC′) should correlate with climate variables. From the analysis in FAN14 and below in Section 3.1, LULCC correlates with precipitation, while precipitation has a negative correlation with temperature. Even human-manipulated croplands are strongly related to precipitation, as agricultural activities in the Loess Plateau region rely significantly on precipitation availability. Therefore, LC′ may be estimated by first establishing a regression relationship between L and the two climate variables P and T, and then calculating LC′ from variations in P and T (i.e., PC′ and TC′). This procedure applies to all four primary LULC zones. In calculating the regression relationship for LAI, all stations in zones B, G, C, and F were used; however, for the areal coverage of an individual LULC type, only those stations within the same zone were used. Finally, LH′ can be estimated from Eq. (5). As was done in Eq. (4), the ratios of contribution from climate variation (rC) and human activities (rH) can be obtained from the standard deviations of LC′, LH′, and L.
3.1 Climate variations within areas of major LULC zones
This subsection first discusses climate variations in the areas where LULC types were sustained and remaining stable throughout the study period. It is shown in Fig. 1 that precipitation gradually increases from Barren → T1 → Grassland → T2 → Cropland → T3 → Forest, which essentially represents a path curving from the northwest down to the south (see Fig. 1a). A distinct difference in precipitation of about 100–200 mm exists between each of the four primary LULC zones. The one-standard deviations indicated by the dashed lines also indicate that the variability experienced a gradual and consistent change across these zones. The average temperatures show that the grassland is the coolest zone, cooler even than the barren area, which also reflects the effect that the grass and associated wetter-than-barren climate had on its temperature. The variation in air temperature lies within a range of 1–2 °C, except for the cropland zone, where a variation of more than 5 °C is recorded. It is primarily due to human interference that cropland encompasses such a wide region (Fig. 1a and d), leading to a larger overall variation.
In the barren zone (Fig. 2a), temperature and precipitation show a clear inverse relationship, with a correlation coefficient r = −0.75 that passes the 95 % t-test confidence level. A very weak correlation exists between the total area of barren land and the precipitation, even though the actual total area contains contributions from human activities. These correlations imply that the total barren land area decreases (increases) when it is cool and moist (warm and dry) and favorable (unfavorable) for vegetative growth. In the grassland zone (Fig. 2b), the inverse relationship between temperature and precipitation is also clearly seen (r = −0.75, 99 %). Although a drying trend occurred after 2003, the decreasing trend in the grassland area was much smaller, implying impacts from human interference. The relationship between the total area of grassland and the precipitation is thus positive but weak (further discussions are presented below in Section 3.2). Croplands and forests are located in relatively warm and wet regions, and the inverse relationship between temperature and precipitation is weaker and insignificant in these zones. Cropland is a human-activity-dominated land type; however, in the Loess Plateau area, the rainfed farming has a strong dependence on local climate, with more precipitation necessary for the expansion of croplands. Therefore, a stronger positive relationship between total area of cropland and precipitation is seen (r = 0.45, 90 %, Fig. 2c). The forest zone displays similar variations as the croplands, but the correlation is not as significant (Fig. 2d).
The variations in precipitation (Fig. 3d–f) exhibit distinct patterns in the three transition zones, implying that the impact from the two contributing factors is more variable. The contribution from climate variation is smaller than for temperature, with rC = 0.71, 0.77, and 0.88, implying greater spatial variation in the northern arid zones than in the humid south, since climate contributions are obtained from two neighboring LULC zones. The contribution from LULCC resulted in rL = 0.91, 0.96, and 0.31 in the three transition zones. Owing to the complex nature of precipitation, many factors may contribute to the change in precipitation apart from the two considered herein. Nonetheless, the difference in rL suggests that LULCC may have stronger impacts in the northern two arid and semi-arid transition zones than in the southern humid one. One possible reason is that any LULCC that occurs in less vegetated arid and semi-arid zones may modify evapotranspiration and associated latent heat partition more greatly than the densely vegetated forests and nearby woody savannahs in the southern humid Qingling Mountains.
3.2 LULCC under similar climate conditions
As pointed out by FAN14, vegetation distribution over the Loess Plateau closely follows the distribution of mean annual precipitation. Spatial correlations were calculated for the percentage of annual precipitation anomaly for each year from 2001 to 2009 and all the years from 1980 to 2010 in order to identify years that had similar patterns in precipitation variation. The time period was chosen to match the availability of LAI and NDVI data. The LAI data have been found to accurately reflect changes in vegetation and match well with satellite imagery (FAN14).
Based on the results of a spatial correlation analysis, four pairs of years were identified as having high spatial correlation coefficients and similar distribution patterns for the anomaly of annual precipitation. These year-pairs were 1995/2008 (A), 1994/2002 (B), 2000/2005 (C), and 1996/2007 (D), which had spatial correlations of r = 0.68, 0.67, 0.64, and 0.55, respectively. Shown also in Fig. 4a are magnitudes of precipitation and temperature anomalies, with the four year-pairs marked. The magnitudes of the precipitation anomaly for the four year-pairs are comparable. The areal-averaged annual precipitation anomaly was negative for pairs A, B, and C, and positive for pair D; in the latter, 2007 was slightly drier than 1996. The magnitudes of the temperature anomaly within each pair were also alike, except for pair D, where 1996 was close to climatology but 2007 was about 1˚C warmer. Thus, the two years in pair D exhibited different climatic conditions than the other three when temperature is taken into consideration. Both years were wetter and warmer than normal, but 2007 was relatively drier and warmer than 1996.
An analysis based on the areal coverage demonstrates similar relationships as shown for LAI for the Cropland and Forest zones (Fig. 6g–h). However, the area of barren land exhibited a positive relationship with precipitation (Fig. 6e), while the area of grassland showed a negative relationship (Fig. 6f). Nonetheless, the relative contributions of climate variation and human impact are consistent with the conclusions reached for LAI above.
4 Summary and discussions
Climate variations in areas of similar land use were slightly milder than in the transition zones in between. On the other hand, an analysis of historically similar climate years provides evidence for the existence of a clear trend in increasing vegetation cover and growth in the 2000s relative to the previous decade. These demonstrate that climate variations exhibited in the primary LULC zones arise from mainly natural fluctuations. Changes in LULC under similar climate conditions primarily resulted from human activities.
Climate variations were attributed to a combination of its own natural interannual fluctuations and LULCC contributions. Contribution ratios indicated larger contributions from climate fluctuations (an average rC of 0.95) than from LULCC (an average rL of 0.35) for variations in temperature. Impacts on precipitation indicated larger spatial variations than on temperature. The spatial variation of LULCC impacts on precipitation implied that human activities might have had greater impacts on precipitation in the arid north than in the humid south, possibly due to the larger impacts on evapotranspiration and associated latent heat partitions over less vegetated arid regions than densely vegetated humid regions (e.g., forest and nearby woody savannahs).
LULCCs were attributed to a combination of climate variations and human activities. Both LAI and areal coverage of each of the primary LULC types were analyzed. The average relative contribution ratios from human activities (afforestation) were rH = 0.99, while the average contribution ratios from climate variations were just rC = 0.26. These suggest that the LULCCs of major land type zones occurred in the 2000s resulted primarily from human activities rather than natural climate variations, which is consistent with the findings reported in FAN14.
The present observation-based study suggests coupled and dynamic relationships between climate variations and LULCCs. One limitation of this study is that the limited observational record left some uncertainties in the results, particularly in the attribution analysis for the precipitation and the areal coverage of different land-cover types. We suggest that modeling study would be the next logical step to further understand the pathways and mechanisms of impacts of afforestation on climate.
This research was supported by the National Basic Research Program of China (2012CB956201), the National Natural Science Foundation of China (41275085), the Knowledge Innovation Program of the Chinese Academy of Sciences (KZCX2-EW-202), and the Special Fund for Meteorological Scientific Research in Public Interest (GYHY201106028). Jeremy Krieger at the University of Alaska Fairbanks helped editing and proofreading the manuscript.
Open Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.