Forest cover change in Northeast China during the period of 1977-2007 and its driving forces

Motivated by asking the question whether or not the large Natural Forest Protection Program (NFPP) had been effective in protecting the natural forests in northeast China. Ten adjacent counties were selected in Sanjiang Plain area of Heilongjiang, upon which region the NFPP had been heavily concentrated. The three chief hypotheses are: (1) the region had undergone severe deforestation and forest degradation before the implementation of NFPP; (2) while the decline of forest cover might have been slowed down following the initiation of NFPP, it would take a longer time to see any significant gain; (3) farmland expansion is the dominant driver of deforestation, whereas population increase, economic growth, and management policy are among the more fundamental forces. Thus the specific tasks were set to detect the regional LUCC over a period of 30 years (1977-2007) and to explore the demographic, economic, political, and other determinants of the detected changes.;Landsat images for six periods were acquired to derive the Land Use Land Cover (LUCC) information. With minor classes being merged, classification resulted in four classes---forestland, farmland, built-up land and other (wetland being a main component). Rule-based rationality evaluation and formal accuracy assessment both proved the classification results are acceptable. The detection results show that: (1) farmland and forestland are the two predominant classes of the regional land use; (2) farmland and built-up land increased persistently during the 30 years; (3) forestland suffered an extended, heavy loss before the end of last century and the decline slowed down significantly thereafter; and (4) "other" land declined continuously. Detailed examination based on extended conversion matrixes reveals that although forestland experienced the most loss, while wetland suffered the largest proportional reduction. Moreover, the calculated landscape diversity and integrity indexes show that the distribution of land-cover types became more uneven, and land-use patches became more interspersed.;During the investigation the effects of various forces driving deforestation based on series of single equation models, it was found that directly taking farmland as regressor suffer problems, e.g. endogeneity. Thus instrument variables analysis and simultaneous equation modelling were employed to remedy the endogeneity problem and to incorporate the interaction and feedback effects between different land uses.;The outcomes of using the instrumental variable (IV) method were much improved---the coefficients of NFPP is significant, implying that the program has played a positive role in protecting local forests. In addition, the coefficient of the "Forestland-Farmland-Wetland" system are generally consistent with those derived from the IV method. The area of wetland is negatively correlated with the area of forestland, indicating a mutual substitution in farmland expansion; likewise, farmland is negatively correlated with wetland. The significantly positive coefficient of built-up area in the farmland equation suggests a strong link between farming activities and residential construction. The significant negative coefficient of irrigation confirms that wetland loss is adversely affected by the change in local cropping structure. However, due to the limitations of small sample data, estimates could possibly suffer an upward bias while inferences are not reliable.