Effectiveness of China's National Forest Protection Program and nature reserves

Abstract

There is profound interest in knowing the degree to which China's institutions are capable of protecting its natural forests and biodiversity in the face of economic and political change. China's 2 most important forest-protection policies are its National Forest Protection Program (NFPP) and its national-level nature reserves (NNRs). The NFPP was implemented in 2000 in response to deforestation-caused flooding. We undertook the first national, quantitative assessment of the NFPP and NNRs to examine whether the NFPP achieved its deforestation-reduction target and whether the NNRs deter deforestation altogether. We used MODIS data to estimate forest cover and loss across mainland China (2000–2010). We also assembled the first-ever polygon dataset for China's forested NNRs (n = 237, 74,030 km2 in 2000) and used both conventional and covariate-matching approaches to compare deforestation rates inside and outside NNRs (2000–2010). In 2000, 1.765 million km2 or 18.7% of mainland China was forested (12.3% with canopy cover of ≥70%)) or woodland (6.4% with canopy cover <70% and tree plus shrub cover ≥40%). By 2010, 480,203 km2 of forest and woodland had been lost, an annual deforestation rate of 2.7%. Forest-only loss was 127,473 km2 (1.05% annually). In the NFPP provinces, the forest-only loss rate was 0.62%, which was 3.3 times lower than in the non-NFPP provinces. Moreover, the Landsat data suggest that these loss rates are overestimates due to large MODIS pixel size. Thus, China appears to have achieved, and even exceeded, its target of reducing deforestation to 1.1% annually in the NFPP provinces. About two-thirds of China's NNRs were effective in protecting forest cover (prevented loss 4073 km2 unmatched approach; 3148 km2 matched approach), and within-NNR deforestation rates were higher in provinces with higher overall deforestation. Our results indicate that China's existing institutions can protect domestic forest cover. La Efectividad del Programa Nacional de Protección de Bosques y las Reservas Naturales en China Existe un interés profundo por conocer el grado al que las instituciones en China son capaces de proteger sus bosques naturales y biodiversidad de frente al cambio económico y político. Las dos políticas de protección de bosques más importantes de China son el Programa Nacional de Protección de Bosques (PNPB) y las reservas naturales a nivel nacional (RNN). El PNPB se implementó en el 2000 en respuesta a las inundaciones causadas por la deforestación. Emprendimos la primera evaluación cuantitativa a nivel nacional del PNPB y las RNN para examinar si el primero logra su objetivo de reducir la deforestación y si el segundo la impide por completo. Utilizamos datos de MODIS para estimar la cobertura y prdida de bosque en toda la China continental (2000–2010). También ensamblamos el primer conjunto de datos de polígonos de las RNN boscosas de China (n = 237, 74,030 km2 en 2000) y usamos tanto estrategias convencionales como de correspondencia de covarianza para comparar las tasas de deforestación dentro y fuera de las RNN (2000–2010). En el 2000, 1.765 millones de km2, o el 18.7% de la China continental, estaban arbolados (12.3% con una cobertura de dosel ≥70%) o cubiertos por bosque (6.4% con una cubierta de dosel <70% y con cubierta de árboles más arbustos ≥40%). Para el 2010, ya se habían perdido 480,203 km2 de bosque, una tasa anual de deforestación del 2.7%. La pérdida del bosque puro fue de 127,473 km2 (1.05% anualmente). En las provincias del PNPB, la tasa de pérdida del bosque puro fue de 0.62%, lo que fue 3.3 veces más bajo que en las provincias fuera del programa. Además, los datos de Landsat sugieren que estas tasas de pérdida son sobreestimados debido al gran tamaño de pixel de MODIS. Así, China parece haber alcanzado, e incluso excedido, su objetivo de reducir la deforestación a 1.1% anualmente en las provincias del PNPB. Aproximadamente dos-tercios de las RNN de China fueron efectivas en la protección de la cobertura del bosque (pérdida prevenida 4073 km2 con la estrategia sin correspondencia; 3148 km2 con la correspondencia), y dentro de las RNN, las tasas de deforestación fueron mayores en las provincias con una deforestación mayor en general. Nuestros resultados indican que las instituciones existentes en China pueden proteger la cobertura doméstica de bosque. China covers one of the greatest ranges of climates and ecological diversity in the world, in total containing ∼10% of all species living on Earth, including over 34,000 vascular plant species and over 7500 vertebrate species, many of which are endemic to China (World Bank 2001a; MEP PRC 2013). Maintaining this high-species diversity is China's large variety of ecosystems, including 212 natural types of forest (World Bank 2001a; MEP PRC 2013). However, China's biodiversity, ecosystem services, and natural landscapes are being degraded at a rapid rate due to swift economic growth and institutional constraints (Economy 2007; Kahn & Yardley 2007; Wang et al. 2007b; Liu et al. 2008). With the decline of its natural forests, China has witnessed deterioration in biodiversity. At least 200 plant species are extinct, 15–20% of China's higher plant species are endangered, 233 vertebrate species are on the edge of extinction, and more than 61% of wildlife species are affected by habitat loss (World Bank 2001a; Liu & Diamond 2005; MEP PRC 2010). In contrast, China has made important political commitments to conservation as a signatory of major international biodiversity treaties and by implementing many domestic conservation laws and regulations and targeting conservation programs for charismatic species such as panda, tiger, primates, and cranes. Arguably the most important class of conservation measures in China is its forest-protection policies. From 1970 to 1996, when weak forest-protection laws and enforcement prevailed, the annual deforestation rate was approximately 2.67–3.36% (Li & Yang 2000), and this loss contributed to severe soil erosion and flooding (Zhang et al. 2000). In particular, deforestation and cultivation on steep slopes in the upper Yangtze River led to the devastating 1998 flood (Zong & Chen 2000; Zhang et al. 2001). In response, in 2000 the Natural Forest Protection Program (NFPP) was implemented across 17 provinces in 3 regions: the upper reaches of the Yangtze River, the upper and middle reaches of the Yellow River, and certain “key state-owned forestry regions,” as designated by the government (Figs. 1 & 2). Final expenditures on the NFPP reached RMB118.6 billion (∼US$19.2 billion) from 2000 through 2010 (SFA PRC 2010a). One of the NFPP's main objectives was to reduce commercial timber extraction by 62.1% through a ban on commercial logging on 0.445 million km2 (SFA PRC 2000). Given that the annual logging rate during 1994–1996 across mainland China was 3.0% (Li & Yang 2000), to achieve a 62.1% reduction in logging implies that the annual deforestation rate within the NFPP zone needed to be reduced to 1.1% (3.0% × [1−62.1%]) during 2000–2010. To judge the NFPP's effectiveness, we can compare China's achieved deforestation rate with this target. China also protects natural forest within its nature reserve (NR) network. Starting with its first NR in 1956, China had established 2669 NRs by 2012. These reserves cover 14.9% of China's land area (1.43 million km2). A total of 363 are national-level nature reserves (NNRs) that account for 62.9% of the area of all NRs (MEP PRC 2013) and are meant to cover the most important and biodiverse ecosystems. Consequently, NNRs receive the highest level of protection and state funding. For example, in 1999 NNRs received $113/km2, while other NRs received $53/km2 (Liu et al. 2003). However, the effectiveness of China's NFPP and NNR programs in protecting natural forest is disputed (Liu et al. 2001; Wang et al. 2013). Even the precise borders of the NNRs have been mostly unknown to the public, and estimates of China's forest cover from 2000 to 2010 vary by a factor of 1.7 (1,209,000–2,054,056 km2), in part due to varying definitions of forest (World Bank 2001b; Hansen et al. 2010; SFA PRC 2005, 2010b). Moreover, because economic development typically takes precedence over biological conservation at local governance levels (Liu & Diamond 2005), it is widely reported that natural forest continues to be cleared or converted to low-biodiversity tree plantations (rubber, Eucalyptus, and fruit trees) (Li 2004; Li et al. 2007; Wang et al. 2007b; Xu 2011; Zhai et al. 2012). Given China's large biodiversity endowment and high level of forest cover and the profound domestic and international interest in knowing the degree to which modern China's governing institutions are capable of protecting biodiversity, we mapped annual forest cover over mainland China from 2000 through 2010 with a uniform remote sensing dataset (MOD13Q1) and the randomForest algorithm (Breiman 2001; Liaw & Wiener 2002; Bartholome & Belward 2005; Clark et al. 2010); quantified deforestation with a 3-year moving window from 2000 to 2010; used conventional unmatched and covariate-matching analyses (Andam et al. 2008; Joppa et al. 2008) to estimate how much closed forest in the NNRs avoided deforestation during the logging-ban era; and compared our results with the recently released, Landsat-based global dataset of deforestation (Hansen et al. 2013). We sought to determine whether 2 important nature-protection institutions in China, the NFPP and the NNR system, have been effective. Estimates of forest-cover loss provide rare quantitative measures of effectiveness, due to the NFPP's public, quantitative deforestation reduction targets and because NNR status prohibits deforestation outright. The MOD13Q1 dataset layers we used to map forest cover included a 16-day composite of red, near infrared, and mid-infrared reflectance, Normalized Difference Vegetation Index, and Enhanced Vegetation Index at 231.7 m resolution (Solano et al. 2010). Low-quality pixels due to clouds and shadows were filled by the mean values of the previous and following scenes. In addition, for all of mainland China, we downloaded processed Landsat data (forest extent 2000 and forest loss 2000–2010) from the Global Forest Change site (http://earthenginepartners.appspot.com/science-2013-global-forest, accessed 2 Nov 2014). Although 30-m resolution Landsat data were also available, the computing cost of analyzing 11 years of Landsat data would have been prohibitive, and, more importantly, the MOD13Q1 dataset provided 23 scenes of uniform data each year, while Landsat could not. Thus, the MOD13Q1 data allowed us to confirm that an apparently deforested pixel remained so for 2 years after the event, thereby increasing confidence. The Shuttle Radar Topography Mission (SRTM) dataset was downloaded from USGS (http://srtm.usgs.gov/index.php). Elevation, slope, and a topographic position index (TPI) were calculated from the 90-m resolution SRTM. The TPI of a focal pixel was defined as the difference between its elevation and the mean elevation of all pixels in an 11×11 grid centered on the focal pixel. The TPI identifies pixels that are either higher (peaks) or lower (gorges) than the surroundings and thus captures local inaccessibility. Using Arc2Earth software (http://www.arc2earth.com/) with ArcGIS (Ver. 10.0, ESRI), we synchronized our MODIS data with high-resolution images in Google Earth (http://www.google.com/earth) and then digitized and visually assigned 13,869 polygons (minimal edge >500 m) to 3 cover classes based on Google Earth images (summer and autumn 2012): forest (tree canopy cover ≥70%), woodland (tree canopy cover <70% and tree plus shrub cover ≥40%), and non-wooded land (tree cover <40%, including open shrub, grassland, farmland, urban, open water, etc.). Tree canopy cover was visually estimated by crown perimeter. With 23 scenes of MOD13Q1 product per year, we used the random forest algorithm to classify annual land cover over mainland China from 2000 through 2010 (Breiman 2001; Clark et al. 2010; Dorais & Cardille For each year, was From 2000 to 2010, we forest loss and because our was to the effectiveness of the NFPP and NNRs in reducing and the of forest and because most in China is due to low-biodiversity plantations 2008; & Xu We defined deforestation as of forest to non-wooded land and deforestation as of forest and woodland to non-wooded deforestation is a because it has a higher of for For the we to of forest to woodland or woodland to We used 3-year moving to forest loss (forest to the first of a window was forest and the 2 years were and was a pixel as deforested in that is for forest cover is over to MODIS pixel as we for one NNR in between years to have been deforested and then To we 200 deforestation pixels at random in and China that could be as deforested or high-resolution in Google was in all The of and polygons of all NNRs on of the of Protection were the NNR and reported to the MEP by each NNR the on of These polygons are the data on NNR vary due to NNR We the effectiveness of the NNRs at the reserve level and at the level because NNRs are at both At each we used both unmatched and matched (Andam et al. 2008; Joppa et al. 2008) to estimate from pixels that were forest in 2000 and to determine these pixels were deforested forest in In the unmatched a of forested pixels in 2000 inside the NNR and outside the NNR from the NNR and compared how many pixels inside and outside remained forested in The matched also compared a of pixels inside and outside the of inside and outside pixels were matched by such as slope, and to forest The matched was more effective at the of reserve status per se because it for the of reserves to be in more remote and thus part of protection from inaccessibility. At the of were 363 NNRs in mainland these at least km2 of forest cover and were over provinces. these we in 2 provinces and because closed forest could be outside the reserve to At the NNR we pixels of forest in 2000 inside an NNR and pixels outside the NNRs within the We calculated as is the total of forest pixels inside the NNR and as is the total of forest pixels outside the NNR and all the other NNRs within the A was used to in which deforestation in the reserve over to outside the reserve 2008; et al. 2008) or deforestation is lower to reserve due to governance or et al. Wang et al. 2013). By pixels to within of NNRs in the we that pixels were from At the we also pixels of forest in 2000 inside all NNRs and pixels of forest in 2000 outside the NNRs within the was the of 2000 and of forest pixels in 2000 inside all NNRs within the is the of and the total forest pixels in 2000 outside all NNRs within the The effectiveness of NNRs within a was calculated as = deforestation rate outside deforestation rate deforestation rate outside = of pixels converted to outside and deforestation rate inside = of pixels converted to inside For the matched our were elevation slope, and to forest For each within-NNR we pixels (Andam et al. 2008; et al. 2013) within of m of m of and of The pixel with the was the (Andam et al. 2008). For the we times and used the For the NNR we times and used the The effectiveness of NNRs within a was calculated as = deforestation rate of matched outside all deforestation rate of matched within all We also NNR effectiveness with Hansen et Landsat To we of the Hansen et al. pixels 1 Hansen pixel. for each Hansen tree cover was calculated as the mean of the pixel We the for We then unmatched and analyses for NNRs at the level as we with the MODIS data also In 2000, 1.765 million km2 or 18.7% of mainland China was in forest or woodland The 17 provinces in the NFPP of China's forest cover 2). and of forest and with the of and the total was 2). deforestation in the in and These provinces for of (Figs. 1 & 2). The provinces with the highest and are designated state-owned forest within the had the highest of forest cover in the and for of from 2000 to 2010 2). For the mainland provinces with forest area than km2 in 2000, forest cover deforestation rate = = = From 2000 to 2010, was 480,203 km2, or of the forest and woodland in million km2 or of mainland China forest or woodland in The provinces for deforestation were in the and including the 3 non-NFPP provinces, and and 2 NFPP provinces, and which contributed of (Figs. 1 & 2). China's 363 natural forest both inside and outside borders to be These which 74,030 km2 of forest in the 2000, 4073 km2 of forest loss or 3148 km2 of loss The matched and unmatched approaches The unmatched NNRs as effective a of forest with and the matched NNRs as effective. both NNRs were as and NNRs were as effective by at least one We also the effectiveness of NNRs at the governance including all provinces with forest cover Thus, and which of km2 forest and forest cover was in were from the NNRs in 23 of provinces with forest as by both approaches of of The higher the deforestation rate in the the more effective the NNRs to be in that = = = = Hansen et Landsat data of provinces effectiveness with with both approaches 23 of of with previous & Zhang 2010; Hansen et al. most of the forest and woodland cover was in China's and by the and then the Although the loss of biodiversity and natural forests are global institutions for and these mostly in large China, which have and political we estimate that mostly natural forest cover from to of mainland China, and forest and woodland cover from 18.7% to From 2000 to 2010, this to a annual loss rate and a of years for forest and to a annual loss rate and a for forest and of the deforestation in the non-NFPP provinces rate in in China (Figs. 1 & while the annual rate across the 17 NFPP provinces was 3.3 times at Moreover, within the NFPP provinces, most of the deforestation was in provinces, and in China which were designated in the NFPP as “key state-owned forest was as in this as in the of the Yangtze and Yellow deforestation was higher in the non-NFPP provinces, in China rate in than in the NFPP provinces However, NFPP provinces, and through which the Yangtze River and which a high of China's biodiversity et al. also high levels of 2). Although a of our suggest that an important of forest loss across all provinces is by tree plantations (Li 2004; Li et al. 2007; Wang et al. & Zhang 2010; Xu Eucalyptus, to high for 2001; 2005; Zhang et al. 2012). In and which have to plantations are also an important (Li et al. and have had on forests from in and living et al. 2012). also appears that was to forests in and in many more than whether were included in the so we deforestation in the NFPP provinces to the of that However, our estimates for annual rates of and deforestation in the NFPP provinces = = 0.62%, and more than achieve the target of reducing annual deforestation to 1.1% 2). In addition, Hansen et Landsat-based of global forest cover to us to that China more than achieved its NFPP Hansen et estimates are with in of forest cover in 2000 1.765 million km2, tree Hansen et al. million km2, tree and the of cover and However, Hansen et al. a lower deforestation rate annual rate for for tree cover than we (1.05% for which is by the pixel in Landsat images m = to MODIS m = deforestation in a MODIS pixel is as to of the deforestation in China by Hansen et al. was pixel in size. Thus, Hansen et our is and that China its NFPP deforestation reduction target for both forest and woodland the forest cover was natural or due to plantations our the in deforestation in NFPP provinces because we that all forest was the forest cover had been then the NFPP would have been even more effective than we The NNR has also reduced deforestation in most provinces and in approximately two-thirds of NNRs (Figs. 3 & NNRs have deforestation outright. Moreover, the total avoided deforestation km2, from the that China's NNR is from forest & et al. However, our of NNR effectiveness is for We natural from of forest we national-level reserves and had to and reserves due to polygon Thus, of our pixels have been in these would have reduced the effectiveness of the The effectiveness of the NNRs across provinces NNR that forest cover inside its borders to in effectiveness as the deforestation rate outside its borders Thus, estimated effectiveness was in non-NFPP provinces, in of deforestation and We that the deforestation rate inside NNRs was with deforestation in the as a = = = = We that in the of forest-protection governance at the level being an NFPP have led to lower levels of protection inside NNRs as The of the effectiveness of NNRs with Hansen et Landsat data as for the MODIS the MODIS data deforestation to the Landsat that the NNRs at the level were at protecting forest cover. The provinces that the most between were and because had the greatest extent of forests, which are to the higher Our the first assessment of China's most important forest-protection policies. the of the NFPP in 2000, the annual rate of forest loss has at a across the the annual forest and woodland loss rate is higher in non-NFPP and in NFPP provinces, and which are biodiversity in The of resolution Landsat data that of MODIS data both of deforestation. Our results are with many analyses in China (Li 2004; Li et al. 2007; Wang et al. & Zhang 2010; Xu 2011; Zhai et al. that natural forests are being by and it appears it is mostly woodland that is being The NFPP has been for and it be to its (SFA PRC 2000, 2010a). The NNRs in China have been is because most nature reserves in China have a of all borders is a conservation for is to or nature reserves in and China, is deforestation rates are and natural forests are et al. Our to the so on effectiveness that has recently been by Joppa and and et al. We also that with effectiveness most of the in China is of forest ecosystems. that to effectiveness, our results suggest that forest governance be with this is is an important for this from of China's National Natural China's of Protection China's of and of of and at of and We Clark for on random forests and and for We and Global Forest Change and Google Earth for data of The is for the or of by the than be to the for the