(Lal, 2000; Prokop and Poreba, 2012; Xiao and Ximing, 2011),and c) agricultural land abandonment (Dhanmanjiri, 2011). The country currently faces the dilemma of needing to increase agricultural productivity while highly productive agricultural lands are being converted to urban uses (Brahmanad et al., 2013; Chadchan and Shankar; 2012; Fazal, 2001; Kalmkar, 2009). The sustainable management of agricultural lands is a prime concern amongst the science and policy communities (Dhanmanjiri, 2011; Fazal, 2001), and a first step towards finding a solution is to develop a scientific understanding of agricultural land loss due to urban expansion.
Two common approaches to examine the area of agricultural land lost to urban growth are to use data from national yearbooks and estimates from remote sensing (Kaufmann and Seto,2001;Seto et al., 2000). Data from yearbooks lack detailed spatially explicit information (Boucher and Seto, 2009; Kaufmann and Seto, 2001). Studies have utilized multi-temporal remote sensing images to investigate urban growth and have reported significant loss of agricultural lands in different Indian cities including Vadodara (Sandhya Kiran and Joshi, 2013), Saharanpur (Fazal, 2001),Hyderabad (Rahman et al., 2011 ; Wakode et al., 2013),and Aligarh (Farooq and Ahmad, 2008). Most studies using multi-temporal remote sensing images analyze land-use change over a defined period of five years (Farooq and Ahmad,2008;Wakode et al., 2013), ten years (Fazal, 2001; Wakode et al., 2013),or other time periods,depending upon the availability of images. Due to a lack of temporal detail, such studies are limited in identifying exactly when land-use change occurred and thus commonly report land-use changes over an entire period. In this context, coarse resolution high frequency data from sensors such as the Moderate Resolution Image Spectroradiometer (MODIS), Defense Mateorological Satellite Program/Operational Linescan System (DMSP/OLS), and SPOT Vegetation (SPOT-VGT) are valuable for determining when land-use changes occur. Knowing when agricultural land conversion took place is important to examine the effectiveness of policies and their implementation (Boucher and Seto, 2009; Kaufmann and Seto, 2001).
The purpose of this research is to examine the urban conversion of agricultural lands in India, including where and when the land conversions occurred. Our primary objectives are to analyze agricultural land loss throughout the country and to explore the degree and consistency of agreement between estimates from agricultural census and satellite data. This study analyzes agricultural land loss in India between 2001 and 2010 using remote sensing data from multiple sensors. The study also highlights the advantages and limitations of using remote sensing versus census-derived estimates to assess agricultural land loss in India.
2.Material and methods
2.1 Data description
2.1.1 Satellite data
We used the MODIS MOD13Q1 time series dataset (10 tiles each containing 253 time series images) for the period June, 2000 to May, 2011. We preprocessed the data on a tile-by-tile basis with the TIMESAT program (Jonsson and Eklundh, 2004) and applied adaptive Savitzky-Golay filter to minimize negatively biased noise commonly encountered in the NDVI datasets due to cloud and haze cover. In addition, we used the pixel reliability parameter included in the MOD13Q1 dataset to assign higher weights (1) to good quality data points and lower weights (0.25) to marginal quality data points in the time series during preprocessing. Even though preprocessing with the adaptive Savitzky-Golay filter minimized the residual noise, anomalous spikes in the NDVI time series remained for some pixels. Hence we applied principal component based projective filtering (Small, 2012) on the NDVI time-series (fig. 1).
In addition to the NDVI dataset, we also used nighttime lights (NTL) data-collected by the DMSP/OLS-to characterize urban conversion of agricultural lands (http://ngdc.noaa.gov/eog/dmsp/downloadV4composites.html). NTL data measure lights on Earth's surface, including street lights, and has been shown to be correlated with urban economic activity (chen and Nordhaus, 2011: Henderson et al., 2012: Sutton, 2007), population (Sutton et al., 2001; Sutton, 2003; Zhuo et al., 2009) and the built environment (Elvidge et al., 2001, 2007: Lu et al., 2008; Ma et al., 2012). A recent study shows that NTL data has mixed accuracy in identifying urbanization. In general, it does well identifying urbanization in developed countries but not as well in developing countries (Zhang and Seto, 2013). The lower accuracy in developing countries is likely due to a lack of infrastructure, especially in the outdoor street lighting, as well as limited access to or availability of electricity (Zhang and Seto, 2013). We used a second-order polynomial regression to inter-calibrate the time series NTL (Elvidge et al., 2009), and then calculated the median NTL for a given year by including images for+-1 year. The rationale for this is to accommodate for any inconsistencies that remain even after inter-calibration and to accurately detect change pixels. We computed the median NTL for 2000 and 2010. To compute median NTL for the year 2000, we used inter-calibrated images of year 1999, 2000 and 2001. The NTL image for 2011 had significant over glow issues even after calibration. Therefore, we computed the median NTL for the year 2010 using calibrated images for 2008, 2009 and 2010. Using visual examination, we observed that the median NTL image were able to capture urban expansion for our study period. It is important to note here that the processed NTL data (median images) used were for the years 2000 and 2010, whereas the MODIS NDVI data were from 2000 to 2011. MODIS NDV time-series was used until 2011 to accommodate the temporal signature of the crops that were sown in the winter season of 2010 and harvested in spring season of 2011. We resampled the median NTL to 250 m to match the spatial resolution of the NDVI data.
2.1.2.1. Agricultural census data
We acquired state-wise agricultural land use statistics compiled by the Directorate of Economics and statistics, Ministry of Agriculture, Govt. of Indian (http://lus.dacnet.nic.in). The statistics include nine land-use categories: 1) forests: 2 ) area under nonagricultural uses, 3 ) barren and un-culturable land, 4) permanent pastures and other grazing lands, 5 ) land under miscellaneous tree, crops, etc., 6 ) Fallow lands other than current follow, 7 ) current fallow, 8 ) culturable waste land (such as fallow or covered with shrubs, but not put to use during the last five year), and 9 ) net area sown. Of the nine categories, the conversion of agricultural lands for non-agricultural purposes is captured in the category "area under non-agricultural uses”. However, this category also includes land under water such as rivers and canals. The agricultural census does not explicitly label agricultural lands that are irreversibly lost to urban development. Analysis of the agricultural census data shows inconsistencies in the time-series estimates of area under non-agricultural uses. For example, for some states, the total area in “non-agricultural uses” was less than the preceding year, suggesting that the total land occupied by buildings and roads were less than the change in land area under water, which is unlikely and suggests errors in the data. To correct for these inconsistencies in the area estimates over time, we normalized the area under non-agricultural uses with the reporting area used time series differencing to obtain a normalized estimate of change in area under non-agricultural uses. Although we removed all the values for
(Lal, 2000; Prokop and Poreba, 2012; Xiao and Ximing, 2011),and c) agricultural land abandonment (Dhanmanjiri, 2011). The country currently faces the dilemma of needing to increase agricultural productivity while highly productive agricultural lands are being converted to urban uses (Brahmanad et al., 2013; Chadchan and Shankar; 2012; Fazal, 2001; Kalmkar, 2009). The sustainable management of agricultural lands is a prime concern amongst the science and policy communities (Dhanmanjiri, 2011; Fazal, 2001), and a first step towards finding a solution is to develop a scientific understanding of agricultural land loss due to urban expansion. Two common approaches to examine the area of agricultural land lost to urban growth are to use data from national yearbooks and estimates from remote sensing (Kaufmann and Seto,2001;Seto et al., 2000). Data from yearbooks lack detailed spatially explicit information (Boucher and Seto, 2009; Kaufmann and Seto, 2001). Studies have utilized multi-temporal remote sensing images to investigate urban growth and have reported significant loss of agricultural lands in different Indian cities including Vadodara (Sandhya Kiran and Joshi, 2013), Saharanpur (Fazal, 2001),Hyderabad (Rahman et al., 2011 ; Wakode et al., 2013),and Aligarh (Farooq and Ahmad, 2008). Most studies using multi-temporal remote sensing images analyze land-use change over a defined period of five years (Farooq and Ahmad,2008;Wakode et al., 2013), ten years (Fazal, 2001; Wakode et al., 2013),or other time periods,depending upon the availability of images. Due to a lack of temporal detail, such studies are limited in identifying exactly when land-use change occurred and thus commonly report land-use changes over an entire period. In this context, coarse resolution high frequency data from sensors such as the Moderate Resolution Image Spectroradiometer (MODIS), Defense Mateorological Satellite Program/Operational Linescan System (DMSP/OLS), and SPOT Vegetation (SPOT-VGT) are valuable for determining when land-use changes occur. Knowing when agricultural land conversion took place is important to examine the effectiveness of policies and their implementation (Boucher and Seto, 2009; Kaufmann and Seto, 2001). The purpose of this research is to examine the urban conversion of agricultural lands in India, including where and when the land conversions occurred. Our primary objectives are to analyze agricultural land loss throughout the country and to explore the degree and consistency of agreement between estimates from agricultural census and satellite data. This study analyzes agricultural land loss in India between 2001 and 2010 using remote sensing data from multiple sensors. The study also highlights the advantages and limitations of using remote sensing versus census-derived estimates to assess agricultural land loss in India. 2.Material and methods2.1 Data description2.1.1 Satellite data We used the MODIS MOD13Q1 time series dataset (10 tiles each containing 253 time series images) for the period June, 2000 to May, 2011. We preprocessed the data on a tile-by-tile basis with the TIMESAT program (Jonsson and Eklundh, 2004) and applied adaptive Savitzky-Golay filter to minimize negatively biased noise commonly encountered in the NDVI datasets due to cloud and haze cover. In addition, we used the pixel reliability parameter included in the MOD13Q1 dataset to assign higher weights (1) to good quality data points and lower weights (0.25) to marginal quality data points in the time series during preprocessing. Even though preprocessing with the adaptive Savitzky-Golay filter minimized the residual noise, anomalous spikes in the NDVI time series remained for some pixels. Hence we applied principal component based projective filtering (Small, 2012) on the NDVI time-series (fig. 1). In addition to the NDVI dataset, we also used nighttime lights (NTL) data-collected by the DMSP/OLS-to characterize urban conversion of agricultural lands (http://ngdc.noaa.gov/eog/dmsp/downloadV4composites.html). NTL data measure lights on Earth's surface, including street lights, and has been shown to be correlated with urban economic activity (chen and Nordhaus, 2011: Henderson et al., 2012: Sutton, 2007), population (Sutton et al., 2001; Sutton, 2003; Zhuo et al., 2009) and the built environment (Elvidge et al., 2001, 2007: Lu et al., 2008; Ma et al., 2012). A recent study shows that NTL data has mixed accuracy in identifying urbanization. In general, it does well identifying urbanization in developed countries but not as well in developing countries (Zhang and Seto, 2013). The lower accuracy in developing countries is likely due to a lack of infrastructure, especially in the outdoor street lighting, as well as limited access to or availability of electricity (Zhang and Seto, 2013). We used a second-order polynomial regression to inter-calibrate the time series NTL (Elvidge et al., 2009), and then calculated the median NTL for a given year by including images for+-1 year. The rationale for this is to accommodate for any inconsistencies that remain even after inter-calibration and to accurately detect change pixels. We computed the median NTL for 2000 and 2010. To compute median NTL for the year 2000, we used inter-calibrated images of year 1999, 2000 and 2001. The NTL image for 2011 had significant over glow issues even after calibration. Therefore, we computed the median NTL for the year 2010 using calibrated images for 2008, 2009 and 2010. Using visual examination, we observed that the median NTL image were able to capture urban expansion for our study period. It is important to note here that the processed NTL data (median images) used were for the years 2000 and 2010, whereas the MODIS NDVI data were from 2000 to 2011. MODIS NDV time-series was used until 2011 to accommodate the temporal signature of the crops that were sown in the winter season of 2010 and harvested in spring season of 2011. We resampled the median NTL to 250 m to match the spatial resolution of the NDVI data.2.1.2.1. Agricultural census data We acquired state-wise agricultural land use statistics compiled by the Directorate of Economics and statistics, Ministry of Agriculture, Govt. of Indian (http://lus.dacnet.nic.in). The statistics include nine land-use categories: 1) forests: 2 ) area under nonagricultural uses, 3 ) barren and un-culturable land, 4) permanent pastures and other grazing lands, 5 ) land under miscellaneous tree, crops, etc., 6 ) Fallow lands other than current follow, 7 ) current fallow, 8 ) culturable waste land (such as fallow or covered with shrubs, but not put to use during the last five year), and 9 ) net area sown. Of the nine categories, the conversion of agricultural lands for non-agricultural purposes is captured in the category "area under non-agricultural uses”. However, this category also includes land under water such as rivers and canals. The agricultural census does not explicitly label agricultural lands that are irreversibly lost to urban development. Analysis of the agricultural census data shows inconsistencies in the time-series estimates of area under non-agricultural uses. For example, for some states, the total area in “non-agricultural uses” was less than the preceding year, suggesting that the total land occupied by buildings and roads were less than the change in land area under water, which is unlikely and suggests errors in the data. To correct for these inconsistencies in the area estimates over time, we normalized the area under non-agricultural uses with the reporting area used time series differencing to obtain a normalized estimate of change in area under non-agricultural uses. Although we removed all the values for
การแปล กรุณารอสักครู่..

( ลาว , 2000 ; โพรเคิป และ poreba , 2012 ; เสี่ยว และ ximing , 2011 ) , และ C ) การละทิ้งที่ดินเพื่อการเกษตร ( dhanmanjiri , 2011 ) ประเทศในปัจจุบันยังคงไม่ต้องเพิ่มผลผลิตทางการเกษตร ขณะที่พื้นที่เกษตรกรรมประสิทธิภาพสูงจะถูกแปลงไปใช้ในเมือง ( brahmanad et al . , 2013 ; และ chadchan Shankar ; 2012 ; fazal , 2001 ; kalmkar , 2009 )การจัดการอย่างยั่งยืนของเกษตร เป็นปัญหาสำคัญในวิทยาศาสตร์และนโยบายชุมชน ( dhanmanjiri 2011 ; fazal , 2001 ) และก้าวแรกในการหาทางออกเพื่อพัฒนาความเข้าใจทางวิทยาศาสตร์ของที่ดินเพื่อการเกษตรที่สูญเสียเนื่องจากการขยายตัวของเมือง
สองวิธีการทั่วไปเพื่อตรวจสอบพื้นที่การเกษตรเสียเพื่อการเจริญเติบโตของเมือง จะใช้ข้อมูลจากความถี่แห่งชาติ และการประเมินจากระยะไกล ( ที่สุดกับเซโตะ เซโตะ , 2001 ; et al . , 2000 ) ข้อมูลจากหนังสือรุ่น ขาดรายละเอียดที่ชัดเจน ( และเปลี่ยนข้อมูล Boucher เซโตะ , 2009 ; ที่สุด และ เซโตะ , 2001 )การศึกษาได้ใช้หลายภาพชั่วคราวระยะไกลเพื่อศึกษาการเจริญเติบโตของเมืองและมีรายงานพบการสูญเสียของที่ดินทางการเกษตรในเมืองอินเดียที่แตกต่างกันรวมทั้ง Vadodara ( sandhya กิราน และ Joshi , 2013 ) ( fazal Saharanpur , 2001 ) , ไฮเดอราบัด ( Rahman et al . , 2011 ; wakode et al . , 2013 ) และ Aligarh ( Farooq และ อาหมัด , 2008 )การศึกษาส่วนใหญ่ใช้ multi ชั่วคราวระยะไกลภาพวิเคราะห์การเปลี่ยนผ่านกำหนดระยะเวลา 5 ปี และ นายฟา , 2008 ; wakode et al . , 2013 ) , สิบปี ( fazal , 2001 ; wakode et al . , 2013 ) , หรืออื่น ๆช่วงเวลา ขึ้นอยู่กับความพร้อมของภาพ เนื่องจากไม่มีรายละเอียดเวลาการศึกษาดังกล่าวจะถูก จำกัด ในการระบุว่าเมื่อเปลี่ยนแปลงการใช้ที่ดินและการเปลี่ยนแปลงการใช้ที่ดินจึงเกิดขึ้นโดยทั่วไปรายงานมากกว่าระยะเวลาทั้งหมด ในบริบทนี้ หยาบ ละเอียด ความถี่สูง เช่น ข้อมูลจากเซ็นเซอร์ภาพความละเอียดปานกลาง spectroradiometer ( โมดิส ) , ป้องกัน mateorological ดาวเทียมโปรแกรม / ระบบงาน linescan ( dmsp / OLS )และจุด ( spot-vgt ) เป็นพืชที่มีคุณค่าสำหรับการกำหนดเวลาการเปลี่ยนแปลงการใช้ที่ดินเกิดขึ้น ทราบว่าเมื่อแปลงที่ดินเอาสถานที่สำคัญเพื่อศึกษาประสิทธิผลของนโยบายและการดำเนินงานของพวกเขาและดึงดูดเซโตะ , 2009 ; คอฟแมน และ เซโตะ , 2001 ) .
วัตถุประสงค์ของงานวิจัยนี้คือ เพื่อศึกษาการแปลงเมืองของที่ดินทางการเกษตรในอินเดียรวมที่ไหนและเมื่อที่ดินแปลงที่เกิดขึ้น วัตถุประสงค์หลักของเราคือเพื่อวิเคราะห์การสูญเสียที่ดินทั่วทั้งประเทศ และเพื่อศึกษาระดับและความสอดคล้องของข้อตกลงระหว่างการประเมินจากการสำรวจสำมะโนประชากร และข้อมูลดาวเทียมเพื่อการเกษตร การศึกษานี้วิเคราะห์การสูญเสียที่ดินทางการเกษตรในอินเดียระหว่างปี 2001 และ 2010 โดยใช้ข้อมูลระยะไกลจากเซ็นเซอร์ต่างๆ
การแปล กรุณารอสักครู่..
