H26V4+Inner+Mongolia+and+Northeastern+China

Positive Vegetation Change in Inner Mongolia and Northeastern China. By Dong Yan

This tile I will work on is H26V4 which covers the major part of Inner Mongolia and Heilongjiang province while some areas of Hebei, Liaoning, Jilin as well as Mongolia and Russia are also included in this tile. The positive vegetation changes are widespread all over the tile. The location of this tile and the positive vegetation change (area outlined in Green) in this tile is shown in figure 1. Figure 1. The location of the MODIS landcover tile h26v04

Based on the University of Maryland classification scheme, the color coded landcover classes of tile h26v4 are shown in figure 2. Figure 2. Color coded landcover classes of MODIS landcover tile h26v04

The widespread positive vegetation change in this tile is suspected to be related to the large scale vegetation conservation programs launched by Chinese government since 1998 which aim to mitigate the degraded natural environment in China. The Natural Forest Conservation Program began in 1998 and it plans to conserve natural forest by mountain closure and afforestation(Zhang et al., 2000). The Grain to Green Program, launched in 1999, aims to convert cropland on steep slopes back to forest and grassland to reduce the amount of soil erosion(Liu et al., 2008). In addition to vegetation conservation programs, Fang&Wang(2010) also have reported weather change as a major factor in controlling inter-annual vegetation variation in this area. So I will also use weather data collected from weather observation sites to test whether there is significant correlation between inter-annual vegetation variation and weather change exists.

The following images shows the spatial and temporal variation of NDVI in Inner Mongolia and Northeastern China. The area shows the most significant NDVI temporal change locates in the eastern part of the image. The vegetation in this area is featured by forest and cropland. The calendar date for the spring NDVI image is Marth 30th and the high missing data rate in early spring is probably due to cloudy or snow condition in this area. As this area is the most important site of rice and corn production in China, during the summer, widespread high NDVI could be found in the summer NDVI image. The high NDVI in the east most part of image in the fall season is probably due to double season crop and evergeen forest. Figure 3. Spring NDVI in Inner Mongolia and Northeastern China (Julian Date:2010/089, Tile:h26v4, Nodata area is shown as black) Figure 4. Summer NDVI in Inner Mongolia and Northeastern China (Julian Date:2010/185, Tile:h26v4, Nodata area is shown as black) Figure 5. Fall NDVI in Inner Mongolia and Northeastern China (Julian Date:2010/273, Tile:h26v4, Nodata area is shown as black)


 * Lab 4**


 * Question1: Write down the image dimensions (e.g. number of lines, columns and bands) for the Anthromes file and the MODIS time series.**


 * =  ||= Sample ||= Line ||= Bands ||= Spatial Resolution (Degree) ||
 * = Anthromes File ||= 434 ||= 121 ||= 1 ||= 0.08333330 ||
 * = MODIS time series ||= 722 ||= 201 ||= 144 ||= 0.05 ||


 * Question2:How do the number of lines and columns compare for the time series and the Anthromes file? Are they the same or different, why? How many time steps does your time series have? How many observations per year?**

The number of sample and line for the Anthromes file are different from that for the MODIS time series due to the different spatial resolution for the two files. There are 144 time steps in the MODIS time series file and there are 16 observations per year from 2000 through 2007 and 4 observations in 2008.


 * Question3: Discuss why it would not be right to shrink the Anthrome pixels.**

As the Anthrome file are of categorical data, the resampling the categorical data would cause significant bias in the results.


 * Basic Class Statistics**
 * = # ||< Class ||= # pixels in area ||= % pixels in area ||
 * = 11 ||< Urban ||= 179 ||= 0.341% ||
 * = 12 ||< Dense settlements ||= 269 ||= 0.512% ||
 * = 21 ||< Rice villages ||= 17 ||= 0.032% ||
 * = 22 ||< Irrigated villages ||= 363 ||= 0.691% ||
 * = 23 ||< Cropped pastoral villages ||= 446 ||= 0.849% ||
 * = 24 ||< Pastoral villages ||= 1,229 ||= 2.340% ||
 * = 25 ||< Rainfed villages ||= 2,305 ||= 4.389% ||
 * = 26 ||< Rainfed mosaic villages ||= 449 ||= 0.855% ||
 * = 31 ||< Residential irrigated cropland ||= 867 ||= 1.651% ||
 * = 32 ||< Residential rainfed mosaic ||= 5,158 ||= 9.822% ||
 * = 33 ||< Populated irrigated cropland ||= 50 ||= 0.095% ||
 * = 34 ||< Populated rainfed cropland ||= 1,011 ||= 1.925% ||
 * = 35 ||< Remote croplands ||= 18 ||= 0.034% ||
 * = 41 ||< Residential rangelands ||= 5,303 ||= 10.098% ||
 * = 42 ||< Populated rangelands ||= 6,357 ||= 12.105% ||
 * = 43 ||< Remote rangelands ||= 10,572 ||= 20.132% ||
 * = 51 ||< Populated forests ||= 3,251 ||= 6.191% ||
 * = 52 ||< Remote forests ||= 4,514 ||= 8.596% ||
 * = 61 ||< Wild forests ||= 2,211 ||= 4.210% ||
 * = 62 ||< Sparse trees ||= 10 ||= 0.019% ||
 * = 63 ||< Barren ||= 664 ||= 1.264% ||


 * Five Dominant Classes**

1.Remote rangelands

2.Populated rangelands

3.Residential rangelands

4.Residential rainfed mosaic

5.Remote forests


 * Question4: How much area is covered in your region by the 5 most dominant classes?**

The five dominant classes have covered 60.75% (1941355.763 square kilometers) of the total area.


 * Figures**

Figure 6 Overall mean NDVI by five dominant classes Figure7 Coefficient of variation by five dominant classes


 * Question5: Discuss the differences between NDVI and the coefficient of variation for the 5 classes.**

The NDVI and coefficient of variation generally change with the human populations. The less the human population, the higher the NDVI as well as the lower coefficient of variation, which means that more human population can bring more disturbance to the area and more human population can lower the stability of the ecosystem. The remote rangelands is an exception to the previous statements, this may be due to the fact that the variability of the remote rangelands is more subject to climate change rather than human population.

Figure8 NDVI time series for remote rangelands 2000 to 2007



Figure9 NDVI time series for populated rangelands 2000 to 2007

Figure10 NDVI time series for residential rangelands 2000 to 2007

Figure11 NDVI time series for Residential rainfed mosaic 2000 to 2007

Figure12 NDVI time series for Remote forests 2000 to 2007

It is possible to determine the starts and ends of the growing season by the NDVI time series, the points on the upward and downward NDVI curve for each year with the biggest slope are starts and ends of growing season respectively.
 * Question6: Can you determine when the growing season starts and ends for each class? Does it change by year?**

Given that there is no severe weather conditions like drought or flood or biological disturbances such as moth invasion, the phenology would exhibit higher stability which means the high variability induced by human populations such as crop phenology has been removed.
 * Question7: How do you think the phenology in this biome would have looked if there were no humans?**


 * Question8: Do you find anything remarkable in the time series? Are there trends?**

The peak values of the time series for remote forests and residential rainfed mosaic last longer than those for the rangelands classes. As the NDVI of remote forests and residential rainfed mosaic is generally higher than that of the rangelands classes, higher vegetation cover should be expected in these two classes and NDVI tends to be saturated in these area which makes a longer peak value. The linear interannual trend is not significant over the study period(very low slope of the trend line as well as value of R square), this may be due to the NDVI sample used in trend estimated is spatially averaged NDVI value and vegetation trends may exist in some areas.


 * Lab 5**

Figure13 Growing season length anomaly from 2001 and 2009

The growing season anomaly map is derived from MODIS/Terra Land Cover Dynamics Yearly L3 global 1km data (MOD12Q2). The growing season length for each year from 2001 to 2009 is derived from the difference between the onset greenness increase date and the onset greenness decrease date for the same year. I assume the date of the greenness increase is the start of season while the date of the greenness decrease is the end of season. I choose the growing season length of 2001 as a base year and the difference between the growing season length of 2001 and that of another year is the growing season length anomaly in each year. The above map is derived by summing up the anomalies of the seven years. The positive sum indicates a lengthening growing season length while the negative sum represents the growing season length is becoming shorter.



Using the 'compute global spatial statistics" from ENVI, I derive the time series of growing season trend. 'The spatially averaged growing season length is showing a slightly increasing trend. This may indicate that the widespread positive vegetation change may be due to the lengthening of growing season over the period from 2001 to 2009.


 * References**

Fang, H., Wang, Ping.(2010). Vegetation Change of Ecotone in West of Northeast China Plain Using Time-series Remote Sensing Data. // Chinese Geographical Sciences //, 20(2), 167-175. Liu, J.G., Li, S.X., Ouyang, Z.Y.,Tam, C., Chen, X.D. (2008). Ecological and socioeconomic effects of China's policies for ecosystem services. // Proceedings of the National Academy of Sciences of the United States of America, // 105, 9477-9482. Zhang,P.C., Shao,G.F., Zhao, G., Le Master, D.C., Parker, G.R., Dunning, J.J.B., et al.(2000). China's Forest Policy for the 21st Century. //Science//, 288, 2135 – 2136.