- Email: [email protected]
A spatial reasoning approach to estimating paddy rice water demand in Hwanghaenam-do, North Korea
agricultural water management 89 (2007) 185–198
available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/agwat
A spatial reasoning approach to estimating paddy rice water demand in Hwanghaenam-do, North Korea Min-Won Jang a, Jin-Yong Choi b,*, Jeong-Jae Lee b a
Research Institute for Agriculture & Life Sciences, Seoul National University, Seoul, Republic of Korea Department of Rural Systems Engineering, Research Institute for Agriculture & Life Sciences, Seoul National University, Seoul, Republic of Korea
b
article info
abstract
Article history:
The primary objective of this study was to estimate the agricultural water demand of paddy
Accepted 22 January 2007
fields in Hwanghaenam-do, North Korea. Three Landsat TM images, GIS data including
Published on line 7 March 2007
digital elevation maps, a Thiessen network and administration maps of North Korea, and meteorological data were synthesized for this study. In order to estimate water demand for
Keywords:
agricultural use, the FAO Blaney–Criddle method and 10-day crop coefficients of the north-
Agricultural water demand
ern areas of South Korea were used. To classify the Landsat images, supervised and
Geographic information system
unsupervised classification methods were conducted. Topographical constraints based
Paddy field
on paddy rice growing conditions, which are under 7% slope and 200 m above sea level,
Remote sensing
were taken into account. The results showed an annual net water demand of 611.7 mm/year
Spatial reasoning
(916.4 Mt/year) is required for the 150,079 ha of paddy fields and the average gross water
North Korea
demand and design water demand for paddy rice were estimated to be 939.6 mm/year (1408 Mt/year) and 1131.97 mm/year (1695.1 Mt/year), respectively. # 2007 Elsevier B.V. All rights reserved.
1.
Introduction
The Hwanghaenam-do province is a supplier of staple foods to North Korea. As such, the agricultural water demand of the region is considered to be the primary factor for assessing current water use and future water requirements. Water demand associated with agricultural water use in North Korea must be considered for near-future investment in agricultural production and to prepare for future Korean unification. Agricultural water demand is generally estimated using the water balance concept (Haque et al., 2004). The water balance is determined by first collecting regional data including observed meteorological data, cultivated land distribution, and farming practices. Secondly, water requirement is computed from daily evapotranspiration and effective rainfall using water balance models developed with the observed data. Lastly, gross water demand considering lot-management water requirement and
canal-system-management water requirement is calculated. South Korea has applied the water balance models based on field experimental data, which had been measured over the long-term, to the development and management of agricultural water resource facilities. Through the water balance methods, the Ministry of Agriculture and Forest (1999) predicted that the total water demand in 2010 would be 15,472 Mt/year for 1,100,000 ha of paddy fields in South Korea. However, for remote sites that cannot be easily accessed by researchers due to political, economical or natural conditions, the typical computational procedures cannot be conducted due to the lack of essential data of appropriate quality. These difficulties are especially relevant in places separated by political barriers, such as North Korea, as it is commonly difficult to reach relevant information from the outside world. Thus, different data collection methods must be applied to gain relevant information for the inaccessible regions.
* Corresponding author. Tel.: +82 28804583; fax: +82 28732087. E-mail address: [email protected] (J.-Y. Choi). 0378-3774/$ – see front matter # 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.agwat.2007.01.009
186
agricultural water management 89 (2007) 185–198
Many researches have shown that spatial reasoning using remote sensing and geographic information system (GIS) is adequate for ascertaining data in such limited conditions as those mentioned above. Spatial reasoning is a modeling process that comprises the formation of ideas through the spatial relationships between geographic objects (Crawford, 1992–1993). Spatial reasoning is a situation where the process and procedures of manipulating maps transcend the mere mechanics of GIS interaction (input, display and management), leading the user to think spatially using the language of spatial statistics, spatial process models, and spatial analysis functions in GIS (Berry, 1995). It is widely used for assessing land cover that is either too remote or extensive for researchers to access directly. Such advanced technologies make possible the affordable collection of a wide range of land cover information in the face of political and geographical barriers. In the case of North Korea, the land use information including cultivated lands is not clear because the North Korea government has never opened the statistics officially and the preceding studies showed considerably different results. For example, Food and Agriculture Organization [FAO] (2001) estimated the paddy area was about 1,000,000 ha in 1993, but it was changed into 600,000 ha in just 1 year. Also, the FAO (2005), which published a digital global map of irrigated areas over the world since 1999, declared that no map of irrigated areas had been available for North Korea. Largely, the total paddy area is evaluated as 580,000–600,000 ha but the spatial distribution is still obscure (Shin et al., 1998). In order to set up the water resources development plan for agriculture and also to predict the crop yields in secluded regions like North Korea, a reasonable estimation for arable area and the geographical distribution should be found first. The objective of this study was to determine the agricultural water demand for a remote site, Hwanghaenam-do in North Korea. To achieve this study goal, paddy rice cultivation areas were classified using a spatial reasoning method for satellite
images of Hwanghaenam-do, and water demand associated with irrigation requirement was estimated considering topographically suitable areas for paddy rice cultivation.
2.
Site selection and methods
2.1.
Site description
Hwanghaenam-do province has an area of about 8578 km2. Hwanghaenam-do is broadly representative of the cultivation areas of North Korea and includes the nation’s largest paddy fields, the Jaeryeong plain (1300 km2). Hwanghaenam-do serves as the principal agricultural base of North Korea. Its 3300 km2 of agricultural areas comprise about 17% of the whole cultivation area of North Korea, and the paddy rice cultivation areas comprise about 26% (1500 km2) of the total North Korean paddy area. Hwanghaenam-do is located very close to South Korea (Fig. 1) and has similar conditions to northern South Korea in terms of agro-climate and farming methods. The annual mean accumulated temperature is 3400–3600 8C, and the mean precipitation is about 1238 mm per year in Hwanghaenamdo (Korea Rural Economic Institute, 1996; Cheon et al., 2003). The cultivation calendar of Hwanghaenam-do can be inferred from that of Kyungki-do, which is the South Korean province closest to North Korea.
2.2.
Methods
Two procedures were carried out in parallel. One was to classify paddy areas from satellite images and the other was to compute agricultural water demand per unit area using meteorological data (Fig. 2). Image processing and classification methods were used to extract the distribution of paddy fields in the study area. Spatial analysis performed based on
Fig. 1 – Study site (Hwanghaenam-do Province, North Korea).
agricultural water management 89 (2007) 185–198
187
Fig. 2 – Flow chart for agricultural water demand calculation.
typical land suitability criteria because suitable groundtruthing data were not available.
Actual crop evapotranspiration is calculated using the following equation (Chung, 2005):
2.2.1.
AET ¼ Kc PET
Estimation of agricultural water demand
Crop water demand depends on various conditions including water management methods, cultivation types, topographic and soil conditions, groundwater level, and meteorological conditions (Raine et al., 2005). In paddy rice cultivation, the water demand is basically determined as a function of evapotranspiration, infiltration, effective rainfall, drainage, and management. Agricultural water demand is defined as the net water demand for paddy rice cultivation and the value determined from daily crop evapotranspiration, effective rainfall and deep percolation. Thus, the water balance of a paddy field can be calculated using the following equation: WR ¼ AET þ DP ER
(1)
where WR is the net water demand (mm), AET the actual crop evapotranspiration (mm), DP the deep percolation (mm), and ER is the effective rainfall (mm).
(2)
where Kc is the crop coefficient and PET is the potential evapotranspiration (mm). Daily potential evapotranspiration (PET) is estimated using meteorological data. Climatic data could not be obtained directly from North Korea, but were available through the GTS (Global Telecommunication System) network of the WMO (World Meteorological Organization. In South Korea, PET is typically calculated from the FAO Modified Penman method or the FAO Blaney–Criddle method. The combination models like the Penman method show the best overall fit among all evapotranspiration estimation models (Jensen, 1974), but it requires a variety of meteorological data. If sufficient meteorological data are not available, an alternative method which depends on less meteorological data should be chosen (Shin, 1984). Comparing to the FAO Modified Penman method using nine climatic components to calculate daily PET, the Blaney–Criddle method requires only temperature data to give
188
agricultural water management 89 (2007) 185–198
a 10-day or monthly mean PET (Doorenbos and Pruitt, 1977). In this study, the FAO Blaney–Criddle method was chosen to estimate PET due to the limitation of available meteorological data of North Korea. The method is widely accepted and is easy to use. Eq. (3) is the Blaney–Criddle method for calculating PET. PET ¼ C½Pð0:46T þ 8:13Þ
(3)
where C is the correction coefficient (generally 1.0), P the percent of daily mean daytime of annual daytime and T is the mean temperature (8C). Deep percolation (DP) is water moving down through the saturated soil layer. In the case of paddy lands, infiltration means deep percolation because all layers of soil are assumed as saturated condition. Infiltration varies with soil characteristics including particle properties, salinity, and groundwater level. Various experiments have shown that about 4–6 mm/ day of water enter into paddy soils during irrigation seasons (Lee, 1988). Through field experiments conducted since 1970, the Rural Research Institute of KRC (Korea Rural Community and Agriculture Corporation) has determined that the average infiltration rate ranges from 5.05.6 mm/day to 5.6 mm/day in several basins (Korea Water Resources Cooperation, 2004). In this study, daily deep percolation (DP) of paddy fields was assumed to be 5.0 mm/day. Finally effective rainfall (ER) is defined as the amount of water available for crop growth from rainfall excluding surface runoff. Effective rainfall amount during irrigation seasons depends on rainfall amount, rainfall intensity, topography, soil infiltration rate, antecedent soil moisture, and water management practices. It is difficult to estimate effective rainfall because the infiltration rate changes with time and soil conditions, and because of the spatial and temporal variability of rain (Malano et al., 2004). Because water management and climate information was limited for North Korea, the fixed effective ratio method, which is simple and has been used for agricultural water facility design by the KRC, was selected. The method gives an ER of 0 mm when rainfall is under 5 mm, 80% of rainfall between 5 mm and 80 mm rainfall, and 64 mm over 80 mm rainfall (Chung et al., 2006).
2.2.2.
Extraction of the distribution of paddy fields
Satellite images make it possible to classify land cover of remote sites when there is a deficiency of ground data that allow researchers to pay considerable attention to classifica-
tion procedures. To increase the accuracy of the classification of paddy fields with topographical criteria, a supervised classification can be fused with an unsupervised classification and a land suitability analysis (Fig. 2). Topographical constraints can be also considered in the evaluation of land suitability for paddy rice, and are easily identified with GIS. For paddy rice cultivation, a topographic condition with a slope and an elevation of less than 7% and 200 m above sea level, respectively, is recommended (NICS; National Institute of Crop Science). The NICS suggests an elevational constraint of less than 100 m above sea level for northern South Korea due to potential damage from cold climate. The classified paddy areas were divided by dominant boundaries obtained from a Thiessen network analysis. The volumetric water demand for paddy fields was computed by multiplying the paddy area of each Thiessen polygon and water demand per square meter.
2.3.
Data collection
2.3.1.
Climate data
The KMA (Korea Meteorological Administration) provides climate data of North Korea through the GTS network of the WMO. Nine meteorological stations from within and near Hwanghaenam-do were selected in this study (Table 1). The data periods covered 26 years from 1974 to 1999.
2.3.2.
GIS data
Four different digital maps were created to rectify the geographical distortion of the satellite images and to analyze the suitability of paddy rice cultivation. Obtainable elevation data were converted to ArcInfo grid format (ESRI, USA) at a cell resolution of 100 m. A location map of meteorological observation stations of North Korea and a Thiessen network map were created based on a KMA report. All maps including satellite images were converted to the UTM Zone 52N projection. Detailed specifications are shown in Table 2.
2.3.3.
Satellite data
The primary satellite data sources were seven spectral band Landsat Thematic Mapper (TM) images (Table 3). The TM scenes used in this study include rows 33 and 34 in paths 117 and 116 (Fig. 3). The images were acquired on August and September, when the rice paddies were near full development. The acquisition of images showing fully developed paddies is critical for the accurate classification of agricultural land cover.
Table 1 – Specifications of nine meteorological observation stations Observation
Code
Latitude
Longitude
Sea water level (m)
Periods
Yangduk Pyungyang Nampo Sariwon Singe Yongyeon Haeju Gaesung Pyunggang
47,052 47,058 47,060 47,065 47,067 47,068 47,069 47,070 47,075
398100 398020 388430 388310 388300 388120 388020 378580 388240
1268500 1258470 1258220 1258460 1268320 1248530 1258420 1268430 1278180
279 38 47 52 100 5 81 70 371
1981–1999 1974–1999 1981–1999 1974–1999 1981–1999 1981–1999 1974–1999 1981–1999 1981–1999
Observation items (1) Mean temperature; (2) min. relative humidity; (3) mean wind velocity; (4) dew point; (5) precipitation
189
agricultural water management 89 (2007) 185–198
Table 2 – Specifications of digital maps Map
Format
Feature
Coordinate System
Administration Map of North Korea Digital Elevation Map Location Map of Met. Observation Stations
Coverage Grid Coverage
Polygon Cell Point
UTM 52N (WGS84)
Remarks 1:50,000 scale 100 m 100 m resolution
Table 3 – Specifications of acquired satellite images Satellite and sensors Landsat5 Thematic Mapper
Path/row
Observed date
Cloud (%)
Sun azimuth
116/34 117/33 117/34
1 September 1996 16 August 1999 16 August 1999
1 0 0
128.008 129.488 126.368
3.
Paddy rice cultivation area mapping
3.1.
Image pre-processing
The image registration procedure involves the selection of accurate GCPs (Ground Control Points). Reference information for Hwanghaenam-do was derived from 1:200,000 scale maps produced by the former Soviet Union, which were served by the UNEP (United Nations Environment Programme) GRID (Global Resource Information Database), Sioux Falls, South Dakota, USA. The 1:200,000 maps contain latitude/longitude grids that were used for the 117/33 and 117/34 image registration. For the 116/34 image, which mostly covers the Kyungki-do and Chungcheong-do provinces of South Korea, 1:25,000 scale topographic maps in the Transverse Mercator projection (Bessel 1841 spheroid) were referenced, which were produced by the NGII (National Geographic Information Institute). All scenes were then reprojected to UTM zone 52N (WGS 84 spheroid) and resampled to a pixel size of 30 m. Initial total RMS (Root Mean Square) error was 1.1 pixel, 1.0
50.008 56.678 56.338
pixel, and 0.5 pixel for path/row 117/33, 117/34, and 116/34, respectively. Suitable GCPs should be selected to obtain better accuracy, but the reading precision of coordinate values varies according to the scale of reference maps. To compensate for the low spatial accuracy in the North Korean parts of the target site, the image-to-image registration technique was performed on the overlap areas between the rectified images. With this method, total RMS errors were decreased to 0.7 pixels (about 21 m). There is no solid standard about the tolerable total RMS error. Several researchers recommended a maximum tolerable RMS error value of <0.5 pixels (Jensen, 1996), but others like Rogan and Chen (2004) stated that the acceptable RMSE depends on the type of satellite images and referring topographic maps. For instance, it can be over five pixels in case of that small scale topographic maps are referenced. In this study, the registrated images were well matched with the digital administration boundaries maps although the scale of the reference maps, 1:200,000, was small. Therefore, the total RMS error, 0.7 pixels, was evaluated to be satisfactory. The pre-processing phase was concluded by conducting the tasseled cap transformation. In an effort to increase the accuracy of the image classification algorithm, the tasseled cap transformation was used to isolate the wetness component of the image subset. The tasseled cap transformation is a linear transformation that rotates data onto new axes directly correlating to the physical characteristics of vegetation as absorption by vegetation in the mid-infrared bands is caused primarily by moisture content, it is related to soil moisture and usually is called wetness (Crist and Cicone, 1984). Hence, the tasseled cap transformation was adopted to produce wetness maps that are strongly correlated with vegetation canopy and soil moisture contents.
3.2.
Fig. 3 – Landsat 5 TM images (RGB 7, 4, 3) for the study area.
Sun elevation
Unsupervised classification
Unsupervised classification was performed with the wetness maps produced using the tasseled cap transformation. In this process, the ISODATA (Iterative Self-Organizing Data Analysis Technique Algorithm) was chosen which uses a minimum spectral distance algorithm to assign a cluster for each candidate pixel. The process begins with a specified number of cluster means and then it processes the image data repetitively, whereby each pixel is assigned to a cluster mean. After each iteration, the clusters can be split or merged
190
agricultural water management 89 (2007) 185–198
depending on the size and spread of the data points in the clusters (Smith and Brown, 1997). Each image was divided into 100 clusters, and known paddy fields were identified on the basis of information from existing land-use maps, field surveys, and experienced knowledge. A mosaic of the three classified images showed a total of 151,670 ha of paddy fields (Fig. 4).
3.3.
Supervised classification
Supervised classification was performed on the water masked images by employing the Bayesian Maximum Likelihood Classifier (MLC) without applying the tasseled cap transformation. The MLC, a parametric decision rule, is a welldeveloped method from statistical decision theory that has been applied to the classification of image data (Settle and Briggs, 1987). Training signatures for identifiable classes are built on the basis of field information. For the section of the study area in North Korea, geographic information of representative large-scale plains such as the Pyeongyang plain (95,000 ha), the Jaeryeong plain (130,000 ha), and the Yeonbaek plain (115,000 ha) were obtained from the Ministry of Unification and the Ministry of Agriculture and Forestry of South Korea. After ensuring that individual classes were suitably discernable, the final classification was carried out, and the resultant images were composed as Fig. 4. Supervised classification showed 189,270 ha of paddy fields, an overestimation as compared to the results of the unsupervised classification. This is because the spectral signature of wetted paddy fields is commonly confused with that of shadowed forest areas although the determination of training signatures is the most sensitive phase of supervised classification. Thus, topographical constraints for paddy rice cultivation were employed to diminish the illogical classification.
3.4.
Post-processing with spatial reasoning
3.4.1.
Topographical constraints for paddy rice cultivation
Due to spectral similarity of rice paddy areas and shadowed, steeply sloped areas, some locations were classified as paddy fields despite surfaces with a steep slope or high altitude. Although it is possible that some of the areas exceeding topographical limits actually comprised paddy area, the pixels were few enough to warrant their exclusion from the area classified as rice paddies. In order to find topographic constraints on cultivation of paddy rice, paddy areas in Kyungki-do in South Korea, which is close to Hwanghaenamdo but separated by the DMZ (Demilitarized Zone), were analyzed using digital elevation data and land use maps published by the Ministry of Environment. In Kyungki-do, about 98.5% (121,200 ha) of all paddy fields (123,100 ha) are distributed under a maximum slope of 7% and an elevation of 200 m above sea level as Fig. 9, which the NICS recommended. Nearly all of the paddy areas that do not meet such conditions are rain-fed and terraced paddy fields in uphill lands (95.6%). Furthermore, because the annual mean temperature is 2 8C lower in Hwanghaenam-do than in Kyungki-do, topographical constraints for paddy rice cultivation are even more limiting. Therefore, it is reasonable to exclude areas classified as paddy fields in Hwanghaenam-do above an elevation of 200 m above sea level. Eventually, a maximum slope of 7% and an elevation of 200 m above sea level were selected as the topographic constraints on cultivation of paddy rice in this study.
3.4.2.
Paddy area distribution
A target image was first created by unioning the results of both classifications so as to not to rule out possible paddy field candidate pixels. Then, two grid maps of elevation and slope were generated using a 100 m 100 m DEM. Maximum and
Fig. 4 – Results of paddy fields classified with unsupervised classification and supervised classification, showing a total of 189,270 ha and a total of 151,670 ha of paddy fields, respectively.
agricultural water management 89 (2007) 185–198
191
Fig. 5 – Hillshade and elevation map of Hwanghaenam-do province. The average elevation is 71.6 m, and about 91.1% of the total territory areas are distributed below 200 m above sea level, and the average slope is 10.5%, and the terrains below slope of 7% cover about 53.7% of the total study areas.
average elevation values were 943 m at Guweol Mountain and 71.6 m, respectively, and the average slope was 10.5% (Fig. 5). The grid maps were overlaid with the previously unioned image, and topographically unfavorable districts were isolated through map algebra. Post-processing resulted in the classification of 150,079 ha of paddy fields (Fig. 6), which is close to
areas from the FAO/WFP report (2001), 147,200 ha in 2001 and 150,300 ha in 2002. The final classified area was acceptable considering that the target satellite images were acquired 2 or 5 years earlier than the report and that the statistical data for North Korea is not accurate. Fig. 6 clearly reveals the Jaeryeong and Yeonbaek plains which are the predominant plains in
Fig. 6 – The finally achieved distribution of paddy fields after excluding topographically unfavorable terrains with a slope and an elevation of more than 7% or 200 m above sea level from the unioned image of both classification, showing a total of 150,079 ha of paddy fields.
192
agricultural water management 89 (2007) 185–198
Fig. 7 – The paddy distribution by elevation showing that 99% of the whole paddy areas of Hwanghaenam-do are located below sea level 100 m.
North Korea. The distribution of paddy fields by altitude is shown in Fig. 7. In Kyungki-do, the cumulative paddy areas below 20 m and 100 m were determined to be 32% and 85%, respectively, while those of Hwanghaenam-do were 84% and 99%.
4. Estimation of agricultural water demand for paddy 4.1. Water demand based on meteorological observation stations First, potential evapotranspiration (PET) was computed using the Blaney–Criddle equation from 10-day average tempera-
Fig. 8 – A result of 10-day water balance analysis for the Nampo station.
ture, and then actual crop evapotranspiration was calculated by multiplying 10-day crop coefficients to 10-day PET values. Crop coefficients were based on standard coefficients determined by field observations made in the northern districts of South Korea over 5 years from 1982 (Table 4). Because paddy rice cultivation in the northern districts begins earlier than in the southern due to the risk of cold damage, the crop coefficients in the northern districts were given from late May to August, whereas in the southern districts they were setup from June to early September. As a result, the average annual actual crop evapotranspiration was about 87.1% of the average annual PET for all target stations. Each 10-day effective rainfall was also decided according to the amount of daily rainfall. Finally, the annual net water demand was calculated using Eq. (1) as shown in Table 5. The
Table 4 – Ten days crop coefficient as determined by the Blaney–Criddle equation (Rural Development Corporation, 1997) Region
Ten days May
Northern Southern
June
July
August
September
Late
Early
Mid
Late
Early
Mid
Late
Early
Mid
Late
Early
0.98 –
0.83 0.83
0.68 0.70
0.87 0.64
1.13 0.77
0.90 0.9
0.82 0.9
0.96 1.11
0.84 0.95
0.72 0.91
– 0.92
Table 5 – Agricultural water demand during the paddy rice growing season (from May to August) by observation station Observation station Yangduk Pyeongyang Nampo Sariwon Singye Yongyeon Haeju Gaesung Pyunggang
Potential evapotranspiration (mm/year)
Actual crop evapotranspiration (mm/year)
Rainfall (mm/year)
Effective rainfall (mm/year)
524.3 542.0 516.1 547.1 526.2 484.2 503.6 503.8 484.4
457.2 472.5 449.7 477.1 458.2 421.9 437.7 438.9 422.3
579.7 565.7 496.3 528.4 643.1 522.1 615.7 727.9 678.8
401.8 436.1 364.1 384.9 482.1 364.7 442.2 522.6 500.9
Net water demand (mm/year) 612.6 613.6 655.6 650.0 594.5 613.9 589.4 551.0 540.1
agricultural water management 89 (2007) 185–198
193
Fig. 9 – A contour map of average annual rainfall (unit: mm/year).
average annual effective rainfall was almost 72.8% of the average annual rainfall during the paddy rice growing season, and the ratio was highest for the Pyeongyang station (77.1%). Among the target stations, potential evapotranspiration and actual crop evapotranspiration were greatest for the Sariwon station, and were smallest for Yongyeon and Pyunggang. However, annual rainfall and effective rainfall were greatest for the Gaesung station, and were smallest for Nampo. The Nampo station maintained the largest net water demand, 655.6 mm/year because the amount of rainfall was absolutely low during the growing season, while the Pyonggang station required the smallest quantity of agricultural
water, 540.1 mm/year because the evapotranspiration was almost smallest but the rainfall was comparatively sufficient and effective. Fig. 8 shows temporal changes of rainfall, consumptive use, and water requirement by 10-day in the Nampo station during paddy rice growing seasons from 1994 to 1999. The change of water requirement was contrary to that of rainfall because the more rainfall an area has, basically the less irrigation water is needed. The radical quantities of rainfall in 1996 and 1999 matched the historical disaster records (Shin et al., 1998) that had occurred in this region before. Rainfall in late July 1996 was extremely intensive enough that the crop yields fell in southwest areas of North
Fig. 10 – A contour map of average annual actual evapotranspiration (unit: mm/year).
194
agricultural water management 89 (2007) 185–198
Fig. 11 – A contour map of average annual net water demand (unit: mm/year).
Korea, and the place was not an exception of worldwide global warming in 1999 in addition to severe drought in 1997. In 1996 when rainfall was almost over 860 mm/year during the paddy rice growing season, the amount of water requirement was also significant. This is because the appropriate amount of rainfall at the right time according to the crop’s need is more effective rather than the high amount of total rainfall. The point-based results, such as rainfall, effective rainfall, and crop evapotranspiration, were interpolated with the inverse distance weighted (IDW) technique. Figs. 9–11 show the spatial distribution rainfall, effective rainfall, and crop evapotranspiration. Annual rainfall in the southeast portion of the target region was higher than in the northwest. However, crop evapotranspiration was higher in the northeast portion of the target region due to a greater number of sunny days and a higher average temperature, relative to the other target areas, during the paddy rice growing season. This finding demonstrates that the northwest portions of the target area have a greater annual net water demand than the southeast regions. That is, it suggests that the Pyeongyang and Jaeryeong plains in the northern sites of the target area need more attention to water resources and irrigation facilities than the Yeonback plain close to the DMZ. Fig. 12 shows average values of annual effective rainfall and annual net water demand. The greater effective rainfall resulted in the less water demand. Water demand in paddy fields was shown to greatly exceed effective rainfall in 1977, 1982, 1989, 1991 and 1999, which may have resulted in the absolute need for irrigation. It coincided with the historical records that North Korea suffered from severe drought. In addition, even in 1995 and 1996, when severe flood damages ensued from abnormally intensive rainfall, the annual net water demands were substantial because of discord between the water supply by rainfall and the water demand by crop.
4.2.
Water demand based on administrative boundaries
4.2.1.
Net water demand
Annual net water demand on each administrative unit, si or gun, similar to city or county of western countries, was extracted using an administration boundaries map and the classified paddy area map of Fig. 6, and volumetric net water demand for paddy was calculated by overlaying the interpolation map of Fig. 11 between annual net water demands by meteorological observations on the classified paddy area map. Each paddy area and water demand for 20 si or gun units are shown in Table 6. In terms of average water depth, the annual net water demand was greatest at Enucheon (648.7 mm/year), Eunyool (639.8 mm/year), Anak (639.0 mm/ year), and Jaeryeong (638.9 mm/year) where the Jaeryeong
Fig. 12 – Annual fluctuations in effective rainfall and net water demand. Hwanghaenam-do was assumed to experience severe drought in 1977, 1982, 1989, 1991 and 1999 when the net water demand was greatly higher than the effective rainfall.
195
agricultural water management 89 (2007) 185–198
Table 6 – Agricultural water demand during the paddy rice growing season (from May to August) by administration units (si or gun) Administration (si/gun)
Anak Baecheon Bongcheon Byeokseong Cheongdan Euncheon Eunyool Gangryeong Gwail Haeju Jaeryeong Jangyeon Samcheon Sincheon Sinwon Songhwa Taetan Woongjin Yeonan Yongyeon Average (total)
Administration area (ha)
Paddy area (ha)
37,484 46,003 53,016 44,209 47,281 43,849 47,026 53,758 36,550 17,622 32,347 39,319 36,229 49,837 52,686 19,207 36,110 66,699 43,923 54,662 857,818
16,210 15,193 3562 6417 13,767 10,394 5597 6737 1941 2037 10,468 2061 3690 11,084 1776 907 4287 10,833 18,568 4550 (150,079)
plain is located, and was smallest at Baecheon (573.7 mm/ year), Bongcheon (587.2 mm/year), and Yeonan (589.7 mm/ year) where the Yeonback plain is located (Fig. 13). On the other hand, as shown in Fig. 14, the total volumetric net water demand was greatest at Yeonan (109.5 Mt/year), Anak (103.6 Mt/year), and Baecheon (87.2 Mt/year). Baecheon and Yeonan showed the smallest depth of annul net water demand, but the volumetric amounts of water demand were greatest because the corresponding paddy areas were considerable as 15,193 ha and 18,568 ha, respectively. Conversely
Net water demand
Gross water demand
By depth (mm/year)
By volume (Mt/year)
By depth (mm/year)
639.0 573.7 587.2 596.5 592.3 648.7 639.8 597.2 621.7 590.1 638.9 614.2 622.9 627.7 604.2 621.3 609.5 605.9 589.7 612.5 611.7
103.6 87.2 20.9 38.3 81.5 67.5 35.8 40.2 12.1 12.0 66.8 12.7 23.0 69.5 10.7 5.6 26.1 65.6 109.5 27.8 (916.4)
973.8 892.1 909.0 920.6 915.4 985.9 974.8 921.5 952.2 912.6 973.6 942.8 953.6 959.6 930.3 951.6 936.9 932.4 912.2 940.7 939.6
By volume (Mt/year) 157.8 135.5 32.4 59.1 126.0 102.5 54.6 62.1 18.5 18.6 101.9 19.4 35.2 106.4 16.5 8.6 40.2 101.0 169.4 42.8 (1408.4)
Euncheon and Eunyool whose annual net water demand by depth was highest demonstrated comparatively small amounts of water demand. In conclusion, Hwanghaenamdo needed a total of 611.7 mm/year or 916.4 Mt/year of annual irrigation water for paddy fields on the average except lotmanagement requirement such as land preparation and transplanting. Concerning the lot-management requirement which generally is considered as 140 mm/day in South Korea, the amount of net water demand was 751.7 mm/year in the target site.
Fig. 13 – Average annual net water demand (mm/year) by administrative units showing the average was 611.7 mm/year for the whole paddy areas, 150,079 ha, of Hwanghaenam-do.
196
agricultural water management 89 (2007) 185–198
Fig. 14 – Volumetric annual net water demand (Mt/year) by administrative units showing the total demand was 916.4 Mt/ year for the whole paddy areas, 150,079 ha, of Hwanghaenam-do.
4.2.2.
Gross water demand
4.2.3.
Gross water demand (GW) for paddy fields derives from the sum of net water demand (WR) and canal-systemmanagement water demand (CMW) as the following equation:
GW ¼ WR þ CMW ¼
WR 1 WLRC
(4)
where WR includes lot-management water requirement and WLRC is water loss ratio in canal-system-management. CMW consists of conveyance losses, delivery management water and maintenance water requirement. Conveyance losses come out through evaporation and seepage from canals during conveying water from water sources to paddy fields, and are mainly determined by the type of canal and the length. The MAF (1998) recommended it as from 15% to 25% and from 5% to 7% of net water demand for earth lining canals and concrete lining canals, respectively. Delivery water requirement is the use of bypass water in paddy field irrigation to help maintain desired water levels in irrigation canals and to distribute water to paddy plots in a uniform manner (Kim et al., 2005), and maintenance water requirement is water flow to maintain canal systems during non-irrigation periods. Chung et al. (2006) revealed in their college textbook that the water loss rate in canal-systemmanagement (WLRC) was from 20% to 30% of gross water demand, and the MAF (1999) applied 20% to predict the water demand for paddy fields of South Korea in 2011. In this study, a constant rate, 20%, was also chosen to calculate the gross water demand of the target site. The calculation details were summarized in Table 6. The estimated gross water demand in Hwanghaenam-do province was 939.6 mm per year on the average, and was totally 1408.4 Mt/year.
Design water demand
Design water demand for agricultural use is defined as an amount of gross water demand that would be expected to recur on the average of every 10 years, and is referred to the design of agricultural water facilities to satisfy water requirement at least at the drought of 10-year return period. A time series of annual gross water requirement was first set up for each si or gun administration unit ranging from 1981 to 1999, and six different cumulative distribution functions (Gaussian,
Table 7 – Design water demand for a 10-year return period by administration units (si or gun) Administration (si/gun)
By depth (mm/year)
By volume (Mt/year)
Anak Baecheon Bongcheon Byeokseong Cheongdan Euncheon Eunyool Gangryeong Gwail Haeju Jaeryeong Jangyeon Samcheon Sincheon Sinwon Songhwa Taetan Woongjin Yeonan Yongyeon
1170.9 1085.2 1100.0 1103.1 1096.9 1208.4 1191.2 1098.2 1159.7 1097.5 1165.9 1145.8 1147.4 1150.6 1113.2 1148.2 1119.8 1109.5 1091.9 1136.3
189.8 164.9 39.2 70.8 151.0 125.6 66.7 74.0 22.5 22.4 122.0 23.6 42.3 127.5 19.8 10.4 48.0 120.2 202.7 51.7
Total
1132.0
1695.1
agricultural water management 89 (2007) 185–198
Log-Gaussian, Gamma, Log-Gamma, Gumbel and Weibull) were then examined to determine the best fit of the theoretical probability distribution with the gross water requirement data of 20 si or gun administration districts. In order to make time series data by si or gun administration districts, the annual gross water demand map was spatially interpolated using a geographic information system after the 19-year gross water requirement data were determined from each weather observation station. Eventually, a Gaussian probability distribution function was selected for all target districts using the KolmogorovSmirnov fitting test. The maximum deviation ranged from 0.088 to 0.154, which is less than the lower limits of the 95% confidence level value for sample sizes of 19, 0.301. By the Gaussian function the design water demand for a 10-year return period was determined to be 1131.97 mm/year (1695.1 Mt/year) on average in Hwanghaenam-do, about 20% more than the gross water demand, 939.6 mm/year. Calculation details are summarized in Table 7. The order of districts in terms of the quantity of design water demand almost coincided with that by the net water demand.
197
rainfall is less than in other regions, the water demand was determined to be relatively greater. In addition, the gross water demand was found to be 939.6 mm/year (1408.4 Mt/ year) considering lot-management water requirement and water loss rate in canal systems, 20%, and the design water demand for a 10-year return period was estimated as 1131.97 mm/year (1695.1 Mt/year). The results of this study are expected to be used as the basis of investigations into water use patterns and to establish and implement future agricultural projects.
Acknowledgements The US government retains ownership of Landsat data. Space Imaging and JAXA (the former NASDA) supported Min W. Jang in the acquisition of satellite data at a marginal cost. Topographic maps of North Korea, created by the former Soviet Union, were obtained with the assistance of Dr. Gene Fosnight with the UNEP GRID in Sioux Falls, South Dakota, USA.
references
5.
Summary and conclusion
This study was conducted to estimate agricultural water requirement of paddy rice in North Korea using the GIS technique with limited access to climate data and information on water management practices due to the political barriers in place. This study obtained paddy rice cultivation area using Landsat TM satellite images and GIS maps, and the cultivation area was incorporated into the estimation of paddy rice water demand to achieve the estimation of water demand for paddy rice cultivation in Hwanghaenam-do. Because paddy fields require flat lands for efficient irrigation and low temperatures to avoid crop damage, the classification accuracy was increased by incorporating spatial reasoning technique with topographical constraints including a slope and elevation of below 7% and 200 m above sea level. Image classification and subsequent spatial reasoning using GIS resulted in 150,079 ha of paddy fields, an area validated with 2001 area values from the FAO report (2001). When planning an irrigation project, it should be prior to estimate water volume needed inside a project area. However, in the case of Hwanghaenam-do, North Korea, it had to be assumed regarding the experiences in other project areas, Kyungki-do of South Korea, nearby and similar to the target area on account of the lack of accessible information. The water balance analysis for paddy fields was conducted using meteorological data. The FAO Blaney–Criddle method was adopted to determine potential evapotranspiration of paddy rice, and consumptive use was calculated using daily meteorological data from the GTS and 10-day crop coefficients for the northern portion of South Korea. Six observation stations in North Korea were used including Nampo, Sariwon, Yongyeon, Haeju, Gaesung, and Singye, where daily data were collected from 1975 to 1999. As a result, the annual net water demand was estimated to be 611.7 mm/year (916.4 Mt/year). In the northwestern areas of Hwanghaenam-do, where effective
Berry, J.K., 1995. Spatial Reasoning for Effective GIS. GIS World Inc., Colorado, USA. Cheon, M.G., Kim, S.J., Cheong, S.K., 2003. A study on Agricultural Water Use System of North Korea and the Improvement Methods. Korea Water Resources Association, Seoul, Republic of Korea (in Korean). Chung, H.W., 2005. Estimation of Crop Water Requirement in Korea. Hanchang Publisher, Seoul, Republic of Korea (in Korean). Chung, H.W., Kim, S.J., Kim, J.S., Noh, J.K., Park, K.U., Son, J.K., Yoon, K.S., Lee, K.H., Lee, N.H., Chung, S.O., Choi, J.D., Choi, J.Y., 2006. Irrigation and Drainage Engineering, 2nd ed. Dongmyeong Publisher, Seoul, Republic of Korea, (in Korean). Crawford, C., 1992–1993. Spatial versus verbal reasoning. The Journal of Computer Game Design 6. Online at http://www.erasmatazz.com/library/jcgd_volume_6/ spatial-_vs_verbal.html. Crist, E.P., Cicone, R.C., 1984. A physically based transformation of Thematic Mapper data: the TM Tasseled Cap. IEEE Geosci. Remote 22, 256–263. Doorenbos, J., Pruitt, W.O., 1977. Guidelines for Predicting Crop Water Requirements. Food and Agriculture Organization, Rome, Italy. Food and Agriculture Organization, 2001. Special Report: FAO/ WFP crop and food supply assessment mission to the Democratic People’s Republic of Korea. Online at http://www.reliefweb.int. Food and Agriculture Organization, 2005. Global map of irrigated areas. Online at http://www.fao.org/ag/AGL/ AGLW/aquastat/irrigationmap. Haque, M.A., Najim, M.M.M., Lee, T.S., 2004. Modeling irrigation water delivery schedule for rice cultivation in east coast Malaysia. Trop. Agric. Res. 16, 204–213. Jensen, J.R., 1996. Introductory Digital Image Processing: A Remote Sensing Perspective. Prentice-Hall, New York, United States. Jensen, M.E., 1974. Consumptive Use of Water and Irrigation Water Requirements. The American Society of Civil Engineers, New York, United States.
198
agricultural water management 89 (2007) 185–198
Kim, J.S., Oh, S.Y., Oh, K.Y., Cho, J.W., 2005. Delivery management water requirement for irrigation ditches associated with large-sized paddy plots in Korea. Paddy Water Environ. 3, 57–62. Korea Rural Economic Institute, 1996. Classification of Agricultural Region and Potential Agricultural Production in North Korea. Rural Development Administration, Seoul, Republic of Korea (in Korean, with English abstract). Korea Water Resources Cooperation, 2004. Dam Design Manual. Korea Water Resources Cooperation, Taejeon, Republic of Korea (in Korean). Lee, N.H., 1988. Simulating Daily Operation of Water Management System Of Irrigation Districts. Ph.D. Thesis. Seoul National University. Seoul, Republic of Korea (in Korean, with English abstract). Malano, H.M., George, B.A., Davidson, B., 2004. A framework for improving the management of irrigation schemes in Vietnam. In: Proceedings of a workshop of Australian Centre for International Agricultural Research, November 28, 2003, Hochiminh, Vietnam, pp. 16–18. Ministry of Agriculture and Forestry, 1998. Design Standard Of Agricultural Improvement Project: Irrigation. Ministry of Agriculture and Forestry, Seoul, Republic of Korea, (in Korean). Ministry of Agriculture and Forestry, 1999. Comprehensive Plan for Agricultural and Rural Water. Ministry of Agriculture and Forestry, Seoul, Republic of Korea (in Korean).
National Institute of Crop Science, 2005. Cultivation Methods by Direct Sowing for Paddy Rice. Online at http:// www.nces.go.kr/R_study/Rice/Cultural/Direct/driect01.htm (in Korean). Raine, S.R., Meyer, W.S., Rassam, D.W., Hutson, J.L., Cook, F.J., 2005. Soil–water and salt movement associated with precision irrigation systems: Research investment opportunities. Final report to the National Program for Sustainable Irrigation. CRCIF technical Report number 3.13/ 1. Cooperative Research Centre for Irrigation Futures, Toowoomba, Australia. Rogan, J., Chen, D., 2004. Remote sensing technology for mapping and monitoring land-cover and land-use change. Prog Plann. 61, 301–325. Rural Development Corporation, 1997. A study on the Water Requirement Variation with the Farming Conditions in the Paddy Field. Ministry of Agriculture and Forestry, Suwon, Republic of Korea (in Korean). Settle, J.J., Briggs, S.S., 1987. Fast maximum likelihood classification of remotely sensed imagery. Int. J. Remote Sens. 8, 723–734. Shin, D.F., Park, S.H., Park, K.Y., Kim, J.H., Lee, G.S., Lee, S.P., 1998. Agricultural Technology of North Korea. Osung Publisher, Seoul, Republic of Korea (in Korea). Shin, S.F., 1984. Data requirements for evapotranspiration estimation. J. Irrig. Drain. 110, 263–373. Smith, C., Brown, N., 1997. IMAGINE 8.3 Guide. ERDAS Inc., Atlanta, GA, USA.
available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/agwat
A spatial reasoning approach to estimating paddy rice water demand in Hwanghaenam-do, North Korea Min-Won Jang a, Jin-Yong Choi b,*, Jeong-Jae Lee b a
Research Institute for Agriculture & Life Sciences, Seoul National University, Seoul, Republic of Korea Department of Rural Systems Engineering, Research Institute for Agriculture & Life Sciences, Seoul National University, Seoul, Republic of Korea
b
article info
abstract
Article history:
The primary objective of this study was to estimate the agricultural water demand of paddy
Accepted 22 January 2007
fields in Hwanghaenam-do, North Korea. Three Landsat TM images, GIS data including
Published on line 7 March 2007
digital elevation maps, a Thiessen network and administration maps of North Korea, and meteorological data were synthesized for this study. In order to estimate water demand for
Keywords:
agricultural use, the FAO Blaney–Criddle method and 10-day crop coefficients of the north-
Agricultural water demand
ern areas of South Korea were used. To classify the Landsat images, supervised and
Geographic information system
unsupervised classification methods were conducted. Topographical constraints based
Paddy field
on paddy rice growing conditions, which are under 7% slope and 200 m above sea level,
Remote sensing
were taken into account. The results showed an annual net water demand of 611.7 mm/year
Spatial reasoning
(916.4 Mt/year) is required for the 150,079 ha of paddy fields and the average gross water
North Korea
demand and design water demand for paddy rice were estimated to be 939.6 mm/year (1408 Mt/year) and 1131.97 mm/year (1695.1 Mt/year), respectively. # 2007 Elsevier B.V. All rights reserved.
1.
Introduction
The Hwanghaenam-do province is a supplier of staple foods to North Korea. As such, the agricultural water demand of the region is considered to be the primary factor for assessing current water use and future water requirements. Water demand associated with agricultural water use in North Korea must be considered for near-future investment in agricultural production and to prepare for future Korean unification. Agricultural water demand is generally estimated using the water balance concept (Haque et al., 2004). The water balance is determined by first collecting regional data including observed meteorological data, cultivated land distribution, and farming practices. Secondly, water requirement is computed from daily evapotranspiration and effective rainfall using water balance models developed with the observed data. Lastly, gross water demand considering lot-management water requirement and
canal-system-management water requirement is calculated. South Korea has applied the water balance models based on field experimental data, which had been measured over the long-term, to the development and management of agricultural water resource facilities. Through the water balance methods, the Ministry of Agriculture and Forest (1999) predicted that the total water demand in 2010 would be 15,472 Mt/year for 1,100,000 ha of paddy fields in South Korea. However, for remote sites that cannot be easily accessed by researchers due to political, economical or natural conditions, the typical computational procedures cannot be conducted due to the lack of essential data of appropriate quality. These difficulties are especially relevant in places separated by political barriers, such as North Korea, as it is commonly difficult to reach relevant information from the outside world. Thus, different data collection methods must be applied to gain relevant information for the inaccessible regions.
* Corresponding author. Tel.: +82 28804583; fax: +82 28732087. E-mail address: [email protected] (J.-Y. Choi). 0378-3774/$ – see front matter # 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.agwat.2007.01.009
186
agricultural water management 89 (2007) 185–198
Many researches have shown that spatial reasoning using remote sensing and geographic information system (GIS) is adequate for ascertaining data in such limited conditions as those mentioned above. Spatial reasoning is a modeling process that comprises the formation of ideas through the spatial relationships between geographic objects (Crawford, 1992–1993). Spatial reasoning is a situation where the process and procedures of manipulating maps transcend the mere mechanics of GIS interaction (input, display and management), leading the user to think spatially using the language of spatial statistics, spatial process models, and spatial analysis functions in GIS (Berry, 1995). It is widely used for assessing land cover that is either too remote or extensive for researchers to access directly. Such advanced technologies make possible the affordable collection of a wide range of land cover information in the face of political and geographical barriers. In the case of North Korea, the land use information including cultivated lands is not clear because the North Korea government has never opened the statistics officially and the preceding studies showed considerably different results. For example, Food and Agriculture Organization [FAO] (2001) estimated the paddy area was about 1,000,000 ha in 1993, but it was changed into 600,000 ha in just 1 year. Also, the FAO (2005), which published a digital global map of irrigated areas over the world since 1999, declared that no map of irrigated areas had been available for North Korea. Largely, the total paddy area is evaluated as 580,000–600,000 ha but the spatial distribution is still obscure (Shin et al., 1998). In order to set up the water resources development plan for agriculture and also to predict the crop yields in secluded regions like North Korea, a reasonable estimation for arable area and the geographical distribution should be found first. The objective of this study was to determine the agricultural water demand for a remote site, Hwanghaenam-do in North Korea. To achieve this study goal, paddy rice cultivation areas were classified using a spatial reasoning method for satellite
images of Hwanghaenam-do, and water demand associated with irrigation requirement was estimated considering topographically suitable areas for paddy rice cultivation.
2.
Site selection and methods
2.1.
Site description
Hwanghaenam-do province has an area of about 8578 km2. Hwanghaenam-do is broadly representative of the cultivation areas of North Korea and includes the nation’s largest paddy fields, the Jaeryeong plain (1300 km2). Hwanghaenam-do serves as the principal agricultural base of North Korea. Its 3300 km2 of agricultural areas comprise about 17% of the whole cultivation area of North Korea, and the paddy rice cultivation areas comprise about 26% (1500 km2) of the total North Korean paddy area. Hwanghaenam-do is located very close to South Korea (Fig. 1) and has similar conditions to northern South Korea in terms of agro-climate and farming methods. The annual mean accumulated temperature is 3400–3600 8C, and the mean precipitation is about 1238 mm per year in Hwanghaenamdo (Korea Rural Economic Institute, 1996; Cheon et al., 2003). The cultivation calendar of Hwanghaenam-do can be inferred from that of Kyungki-do, which is the South Korean province closest to North Korea.
2.2.
Methods
Two procedures were carried out in parallel. One was to classify paddy areas from satellite images and the other was to compute agricultural water demand per unit area using meteorological data (Fig. 2). Image processing and classification methods were used to extract the distribution of paddy fields in the study area. Spatial analysis performed based on
Fig. 1 – Study site (Hwanghaenam-do Province, North Korea).
agricultural water management 89 (2007) 185–198
187
Fig. 2 – Flow chart for agricultural water demand calculation.
typical land suitability criteria because suitable groundtruthing data were not available.
Actual crop evapotranspiration is calculated using the following equation (Chung, 2005):
2.2.1.
AET ¼ Kc PET
Estimation of agricultural water demand
Crop water demand depends on various conditions including water management methods, cultivation types, topographic and soil conditions, groundwater level, and meteorological conditions (Raine et al., 2005). In paddy rice cultivation, the water demand is basically determined as a function of evapotranspiration, infiltration, effective rainfall, drainage, and management. Agricultural water demand is defined as the net water demand for paddy rice cultivation and the value determined from daily crop evapotranspiration, effective rainfall and deep percolation. Thus, the water balance of a paddy field can be calculated using the following equation: WR ¼ AET þ DP ER
(1)
where WR is the net water demand (mm), AET the actual crop evapotranspiration (mm), DP the deep percolation (mm), and ER is the effective rainfall (mm).
(2)
where Kc is the crop coefficient and PET is the potential evapotranspiration (mm). Daily potential evapotranspiration (PET) is estimated using meteorological data. Climatic data could not be obtained directly from North Korea, but were available through the GTS (Global Telecommunication System) network of the WMO (World Meteorological Organization. In South Korea, PET is typically calculated from the FAO Modified Penman method or the FAO Blaney–Criddle method. The combination models like the Penman method show the best overall fit among all evapotranspiration estimation models (Jensen, 1974), but it requires a variety of meteorological data. If sufficient meteorological data are not available, an alternative method which depends on less meteorological data should be chosen (Shin, 1984). Comparing to the FAO Modified Penman method using nine climatic components to calculate daily PET, the Blaney–Criddle method requires only temperature data to give
188
agricultural water management 89 (2007) 185–198
a 10-day or monthly mean PET (Doorenbos and Pruitt, 1977). In this study, the FAO Blaney–Criddle method was chosen to estimate PET due to the limitation of available meteorological data of North Korea. The method is widely accepted and is easy to use. Eq. (3) is the Blaney–Criddle method for calculating PET. PET ¼ C½Pð0:46T þ 8:13Þ
(3)
where C is the correction coefficient (generally 1.0), P the percent of daily mean daytime of annual daytime and T is the mean temperature (8C). Deep percolation (DP) is water moving down through the saturated soil layer. In the case of paddy lands, infiltration means deep percolation because all layers of soil are assumed as saturated condition. Infiltration varies with soil characteristics including particle properties, salinity, and groundwater level. Various experiments have shown that about 4–6 mm/ day of water enter into paddy soils during irrigation seasons (Lee, 1988). Through field experiments conducted since 1970, the Rural Research Institute of KRC (Korea Rural Community and Agriculture Corporation) has determined that the average infiltration rate ranges from 5.05.6 mm/day to 5.6 mm/day in several basins (Korea Water Resources Cooperation, 2004). In this study, daily deep percolation (DP) of paddy fields was assumed to be 5.0 mm/day. Finally effective rainfall (ER) is defined as the amount of water available for crop growth from rainfall excluding surface runoff. Effective rainfall amount during irrigation seasons depends on rainfall amount, rainfall intensity, topography, soil infiltration rate, antecedent soil moisture, and water management practices. It is difficult to estimate effective rainfall because the infiltration rate changes with time and soil conditions, and because of the spatial and temporal variability of rain (Malano et al., 2004). Because water management and climate information was limited for North Korea, the fixed effective ratio method, which is simple and has been used for agricultural water facility design by the KRC, was selected. The method gives an ER of 0 mm when rainfall is under 5 mm, 80% of rainfall between 5 mm and 80 mm rainfall, and 64 mm over 80 mm rainfall (Chung et al., 2006).
2.2.2.
Extraction of the distribution of paddy fields
Satellite images make it possible to classify land cover of remote sites when there is a deficiency of ground data that allow researchers to pay considerable attention to classifica-
tion procedures. To increase the accuracy of the classification of paddy fields with topographical criteria, a supervised classification can be fused with an unsupervised classification and a land suitability analysis (Fig. 2). Topographical constraints can be also considered in the evaluation of land suitability for paddy rice, and are easily identified with GIS. For paddy rice cultivation, a topographic condition with a slope and an elevation of less than 7% and 200 m above sea level, respectively, is recommended (NICS; National Institute of Crop Science). The NICS suggests an elevational constraint of less than 100 m above sea level for northern South Korea due to potential damage from cold climate. The classified paddy areas were divided by dominant boundaries obtained from a Thiessen network analysis. The volumetric water demand for paddy fields was computed by multiplying the paddy area of each Thiessen polygon and water demand per square meter.
2.3.
Data collection
2.3.1.
Climate data
The KMA (Korea Meteorological Administration) provides climate data of North Korea through the GTS network of the WMO. Nine meteorological stations from within and near Hwanghaenam-do were selected in this study (Table 1). The data periods covered 26 years from 1974 to 1999.
2.3.2.
GIS data
Four different digital maps were created to rectify the geographical distortion of the satellite images and to analyze the suitability of paddy rice cultivation. Obtainable elevation data were converted to ArcInfo grid format (ESRI, USA) at a cell resolution of 100 m. A location map of meteorological observation stations of North Korea and a Thiessen network map were created based on a KMA report. All maps including satellite images were converted to the UTM Zone 52N projection. Detailed specifications are shown in Table 2.
2.3.3.
Satellite data
The primary satellite data sources were seven spectral band Landsat Thematic Mapper (TM) images (Table 3). The TM scenes used in this study include rows 33 and 34 in paths 117 and 116 (Fig. 3). The images were acquired on August and September, when the rice paddies were near full development. The acquisition of images showing fully developed paddies is critical for the accurate classification of agricultural land cover.
Table 1 – Specifications of nine meteorological observation stations Observation
Code
Latitude
Longitude
Sea water level (m)
Periods
Yangduk Pyungyang Nampo Sariwon Singe Yongyeon Haeju Gaesung Pyunggang
47,052 47,058 47,060 47,065 47,067 47,068 47,069 47,070 47,075
398100 398020 388430 388310 388300 388120 388020 378580 388240
1268500 1258470 1258220 1258460 1268320 1248530 1258420 1268430 1278180
279 38 47 52 100 5 81 70 371
1981–1999 1974–1999 1981–1999 1974–1999 1981–1999 1981–1999 1974–1999 1981–1999 1981–1999
Observation items (1) Mean temperature; (2) min. relative humidity; (3) mean wind velocity; (4) dew point; (5) precipitation
189
agricultural water management 89 (2007) 185–198
Table 2 – Specifications of digital maps Map
Format
Feature
Coordinate System
Administration Map of North Korea Digital Elevation Map Location Map of Met. Observation Stations
Coverage Grid Coverage
Polygon Cell Point
UTM 52N (WGS84)
Remarks 1:50,000 scale 100 m 100 m resolution
Table 3 – Specifications of acquired satellite images Satellite and sensors Landsat5 Thematic Mapper
Path/row
Observed date
Cloud (%)
Sun azimuth
116/34 117/33 117/34
1 September 1996 16 August 1999 16 August 1999
1 0 0
128.008 129.488 126.368
3.
Paddy rice cultivation area mapping
3.1.
Image pre-processing
The image registration procedure involves the selection of accurate GCPs (Ground Control Points). Reference information for Hwanghaenam-do was derived from 1:200,000 scale maps produced by the former Soviet Union, which were served by the UNEP (United Nations Environment Programme) GRID (Global Resource Information Database), Sioux Falls, South Dakota, USA. The 1:200,000 maps contain latitude/longitude grids that were used for the 117/33 and 117/34 image registration. For the 116/34 image, which mostly covers the Kyungki-do and Chungcheong-do provinces of South Korea, 1:25,000 scale topographic maps in the Transverse Mercator projection (Bessel 1841 spheroid) were referenced, which were produced by the NGII (National Geographic Information Institute). All scenes were then reprojected to UTM zone 52N (WGS 84 spheroid) and resampled to a pixel size of 30 m. Initial total RMS (Root Mean Square) error was 1.1 pixel, 1.0
50.008 56.678 56.338
pixel, and 0.5 pixel for path/row 117/33, 117/34, and 116/34, respectively. Suitable GCPs should be selected to obtain better accuracy, but the reading precision of coordinate values varies according to the scale of reference maps. To compensate for the low spatial accuracy in the North Korean parts of the target site, the image-to-image registration technique was performed on the overlap areas between the rectified images. With this method, total RMS errors were decreased to 0.7 pixels (about 21 m). There is no solid standard about the tolerable total RMS error. Several researchers recommended a maximum tolerable RMS error value of <0.5 pixels (Jensen, 1996), but others like Rogan and Chen (2004) stated that the acceptable RMSE depends on the type of satellite images and referring topographic maps. For instance, it can be over five pixels in case of that small scale topographic maps are referenced. In this study, the registrated images were well matched with the digital administration boundaries maps although the scale of the reference maps, 1:200,000, was small. Therefore, the total RMS error, 0.7 pixels, was evaluated to be satisfactory. The pre-processing phase was concluded by conducting the tasseled cap transformation. In an effort to increase the accuracy of the image classification algorithm, the tasseled cap transformation was used to isolate the wetness component of the image subset. The tasseled cap transformation is a linear transformation that rotates data onto new axes directly correlating to the physical characteristics of vegetation as absorption by vegetation in the mid-infrared bands is caused primarily by moisture content, it is related to soil moisture and usually is called wetness (Crist and Cicone, 1984). Hence, the tasseled cap transformation was adopted to produce wetness maps that are strongly correlated with vegetation canopy and soil moisture contents.
3.2.
Fig. 3 – Landsat 5 TM images (RGB 7, 4, 3) for the study area.
Sun elevation
Unsupervised classification
Unsupervised classification was performed with the wetness maps produced using the tasseled cap transformation. In this process, the ISODATA (Iterative Self-Organizing Data Analysis Technique Algorithm) was chosen which uses a minimum spectral distance algorithm to assign a cluster for each candidate pixel. The process begins with a specified number of cluster means and then it processes the image data repetitively, whereby each pixel is assigned to a cluster mean. After each iteration, the clusters can be split or merged
190
agricultural water management 89 (2007) 185–198
depending on the size and spread of the data points in the clusters (Smith and Brown, 1997). Each image was divided into 100 clusters, and known paddy fields were identified on the basis of information from existing land-use maps, field surveys, and experienced knowledge. A mosaic of the three classified images showed a total of 151,670 ha of paddy fields (Fig. 4).
3.3.
Supervised classification
Supervised classification was performed on the water masked images by employing the Bayesian Maximum Likelihood Classifier (MLC) without applying the tasseled cap transformation. The MLC, a parametric decision rule, is a welldeveloped method from statistical decision theory that has been applied to the classification of image data (Settle and Briggs, 1987). Training signatures for identifiable classes are built on the basis of field information. For the section of the study area in North Korea, geographic information of representative large-scale plains such as the Pyeongyang plain (95,000 ha), the Jaeryeong plain (130,000 ha), and the Yeonbaek plain (115,000 ha) were obtained from the Ministry of Unification and the Ministry of Agriculture and Forestry of South Korea. After ensuring that individual classes were suitably discernable, the final classification was carried out, and the resultant images were composed as Fig. 4. Supervised classification showed 189,270 ha of paddy fields, an overestimation as compared to the results of the unsupervised classification. This is because the spectral signature of wetted paddy fields is commonly confused with that of shadowed forest areas although the determination of training signatures is the most sensitive phase of supervised classification. Thus, topographical constraints for paddy rice cultivation were employed to diminish the illogical classification.
3.4.
Post-processing with spatial reasoning
3.4.1.
Topographical constraints for paddy rice cultivation
Due to spectral similarity of rice paddy areas and shadowed, steeply sloped areas, some locations were classified as paddy fields despite surfaces with a steep slope or high altitude. Although it is possible that some of the areas exceeding topographical limits actually comprised paddy area, the pixels were few enough to warrant their exclusion from the area classified as rice paddies. In order to find topographic constraints on cultivation of paddy rice, paddy areas in Kyungki-do in South Korea, which is close to Hwanghaenamdo but separated by the DMZ (Demilitarized Zone), were analyzed using digital elevation data and land use maps published by the Ministry of Environment. In Kyungki-do, about 98.5% (121,200 ha) of all paddy fields (123,100 ha) are distributed under a maximum slope of 7% and an elevation of 200 m above sea level as Fig. 9, which the NICS recommended. Nearly all of the paddy areas that do not meet such conditions are rain-fed and terraced paddy fields in uphill lands (95.6%). Furthermore, because the annual mean temperature is 2 8C lower in Hwanghaenam-do than in Kyungki-do, topographical constraints for paddy rice cultivation are even more limiting. Therefore, it is reasonable to exclude areas classified as paddy fields in Hwanghaenam-do above an elevation of 200 m above sea level. Eventually, a maximum slope of 7% and an elevation of 200 m above sea level were selected as the topographic constraints on cultivation of paddy rice in this study.
3.4.2.
Paddy area distribution
A target image was first created by unioning the results of both classifications so as to not to rule out possible paddy field candidate pixels. Then, two grid maps of elevation and slope were generated using a 100 m 100 m DEM. Maximum and
Fig. 4 – Results of paddy fields classified with unsupervised classification and supervised classification, showing a total of 189,270 ha and a total of 151,670 ha of paddy fields, respectively.
agricultural water management 89 (2007) 185–198
191
Fig. 5 – Hillshade and elevation map of Hwanghaenam-do province. The average elevation is 71.6 m, and about 91.1% of the total territory areas are distributed below 200 m above sea level, and the average slope is 10.5%, and the terrains below slope of 7% cover about 53.7% of the total study areas.
average elevation values were 943 m at Guweol Mountain and 71.6 m, respectively, and the average slope was 10.5% (Fig. 5). The grid maps were overlaid with the previously unioned image, and topographically unfavorable districts were isolated through map algebra. Post-processing resulted in the classification of 150,079 ha of paddy fields (Fig. 6), which is close to
areas from the FAO/WFP report (2001), 147,200 ha in 2001 and 150,300 ha in 2002. The final classified area was acceptable considering that the target satellite images were acquired 2 or 5 years earlier than the report and that the statistical data for North Korea is not accurate. Fig. 6 clearly reveals the Jaeryeong and Yeonbaek plains which are the predominant plains in
Fig. 6 – The finally achieved distribution of paddy fields after excluding topographically unfavorable terrains with a slope and an elevation of more than 7% or 200 m above sea level from the unioned image of both classification, showing a total of 150,079 ha of paddy fields.
192
agricultural water management 89 (2007) 185–198
Fig. 7 – The paddy distribution by elevation showing that 99% of the whole paddy areas of Hwanghaenam-do are located below sea level 100 m.
North Korea. The distribution of paddy fields by altitude is shown in Fig. 7. In Kyungki-do, the cumulative paddy areas below 20 m and 100 m were determined to be 32% and 85%, respectively, while those of Hwanghaenam-do were 84% and 99%.
4. Estimation of agricultural water demand for paddy 4.1. Water demand based on meteorological observation stations First, potential evapotranspiration (PET) was computed using the Blaney–Criddle equation from 10-day average tempera-
Fig. 8 – A result of 10-day water balance analysis for the Nampo station.
ture, and then actual crop evapotranspiration was calculated by multiplying 10-day crop coefficients to 10-day PET values. Crop coefficients were based on standard coefficients determined by field observations made in the northern districts of South Korea over 5 years from 1982 (Table 4). Because paddy rice cultivation in the northern districts begins earlier than in the southern due to the risk of cold damage, the crop coefficients in the northern districts were given from late May to August, whereas in the southern districts they were setup from June to early September. As a result, the average annual actual crop evapotranspiration was about 87.1% of the average annual PET for all target stations. Each 10-day effective rainfall was also decided according to the amount of daily rainfall. Finally, the annual net water demand was calculated using Eq. (1) as shown in Table 5. The
Table 4 – Ten days crop coefficient as determined by the Blaney–Criddle equation (Rural Development Corporation, 1997) Region
Ten days May
Northern Southern
June
July
August
September
Late
Early
Mid
Late
Early
Mid
Late
Early
Mid
Late
Early
0.98 –
0.83 0.83
0.68 0.70
0.87 0.64
1.13 0.77
0.90 0.9
0.82 0.9
0.96 1.11
0.84 0.95
0.72 0.91
– 0.92
Table 5 – Agricultural water demand during the paddy rice growing season (from May to August) by observation station Observation station Yangduk Pyeongyang Nampo Sariwon Singye Yongyeon Haeju Gaesung Pyunggang
Potential evapotranspiration (mm/year)
Actual crop evapotranspiration (mm/year)
Rainfall (mm/year)
Effective rainfall (mm/year)
524.3 542.0 516.1 547.1 526.2 484.2 503.6 503.8 484.4
457.2 472.5 449.7 477.1 458.2 421.9 437.7 438.9 422.3
579.7 565.7 496.3 528.4 643.1 522.1 615.7 727.9 678.8
401.8 436.1 364.1 384.9 482.1 364.7 442.2 522.6 500.9
Net water demand (mm/year) 612.6 613.6 655.6 650.0 594.5 613.9 589.4 551.0 540.1
agricultural water management 89 (2007) 185–198
193
Fig. 9 – A contour map of average annual rainfall (unit: mm/year).
average annual effective rainfall was almost 72.8% of the average annual rainfall during the paddy rice growing season, and the ratio was highest for the Pyeongyang station (77.1%). Among the target stations, potential evapotranspiration and actual crop evapotranspiration were greatest for the Sariwon station, and were smallest for Yongyeon and Pyunggang. However, annual rainfall and effective rainfall were greatest for the Gaesung station, and were smallest for Nampo. The Nampo station maintained the largest net water demand, 655.6 mm/year because the amount of rainfall was absolutely low during the growing season, while the Pyonggang station required the smallest quantity of agricultural
water, 540.1 mm/year because the evapotranspiration was almost smallest but the rainfall was comparatively sufficient and effective. Fig. 8 shows temporal changes of rainfall, consumptive use, and water requirement by 10-day in the Nampo station during paddy rice growing seasons from 1994 to 1999. The change of water requirement was contrary to that of rainfall because the more rainfall an area has, basically the less irrigation water is needed. The radical quantities of rainfall in 1996 and 1999 matched the historical disaster records (Shin et al., 1998) that had occurred in this region before. Rainfall in late July 1996 was extremely intensive enough that the crop yields fell in southwest areas of North
Fig. 10 – A contour map of average annual actual evapotranspiration (unit: mm/year).
194
agricultural water management 89 (2007) 185–198
Fig. 11 – A contour map of average annual net water demand (unit: mm/year).
Korea, and the place was not an exception of worldwide global warming in 1999 in addition to severe drought in 1997. In 1996 when rainfall was almost over 860 mm/year during the paddy rice growing season, the amount of water requirement was also significant. This is because the appropriate amount of rainfall at the right time according to the crop’s need is more effective rather than the high amount of total rainfall. The point-based results, such as rainfall, effective rainfall, and crop evapotranspiration, were interpolated with the inverse distance weighted (IDW) technique. Figs. 9–11 show the spatial distribution rainfall, effective rainfall, and crop evapotranspiration. Annual rainfall in the southeast portion of the target region was higher than in the northwest. However, crop evapotranspiration was higher in the northeast portion of the target region due to a greater number of sunny days and a higher average temperature, relative to the other target areas, during the paddy rice growing season. This finding demonstrates that the northwest portions of the target area have a greater annual net water demand than the southeast regions. That is, it suggests that the Pyeongyang and Jaeryeong plains in the northern sites of the target area need more attention to water resources and irrigation facilities than the Yeonback plain close to the DMZ. Fig. 12 shows average values of annual effective rainfall and annual net water demand. The greater effective rainfall resulted in the less water demand. Water demand in paddy fields was shown to greatly exceed effective rainfall in 1977, 1982, 1989, 1991 and 1999, which may have resulted in the absolute need for irrigation. It coincided with the historical records that North Korea suffered from severe drought. In addition, even in 1995 and 1996, when severe flood damages ensued from abnormally intensive rainfall, the annual net water demands were substantial because of discord between the water supply by rainfall and the water demand by crop.
4.2.
Water demand based on administrative boundaries
4.2.1.
Net water demand
Annual net water demand on each administrative unit, si or gun, similar to city or county of western countries, was extracted using an administration boundaries map and the classified paddy area map of Fig. 6, and volumetric net water demand for paddy was calculated by overlaying the interpolation map of Fig. 11 between annual net water demands by meteorological observations on the classified paddy area map. Each paddy area and water demand for 20 si or gun units are shown in Table 6. In terms of average water depth, the annual net water demand was greatest at Enucheon (648.7 mm/year), Eunyool (639.8 mm/year), Anak (639.0 mm/ year), and Jaeryeong (638.9 mm/year) where the Jaeryeong
Fig. 12 – Annual fluctuations in effective rainfall and net water demand. Hwanghaenam-do was assumed to experience severe drought in 1977, 1982, 1989, 1991 and 1999 when the net water demand was greatly higher than the effective rainfall.
195
agricultural water management 89 (2007) 185–198
Table 6 – Agricultural water demand during the paddy rice growing season (from May to August) by administration units (si or gun) Administration (si/gun)
Anak Baecheon Bongcheon Byeokseong Cheongdan Euncheon Eunyool Gangryeong Gwail Haeju Jaeryeong Jangyeon Samcheon Sincheon Sinwon Songhwa Taetan Woongjin Yeonan Yongyeon Average (total)
Administration area (ha)
Paddy area (ha)
37,484 46,003 53,016 44,209 47,281 43,849 47,026 53,758 36,550 17,622 32,347 39,319 36,229 49,837 52,686 19,207 36,110 66,699 43,923 54,662 857,818
16,210 15,193 3562 6417 13,767 10,394 5597 6737 1941 2037 10,468 2061 3690 11,084 1776 907 4287 10,833 18,568 4550 (150,079)
plain is located, and was smallest at Baecheon (573.7 mm/ year), Bongcheon (587.2 mm/year), and Yeonan (589.7 mm/ year) where the Yeonback plain is located (Fig. 13). On the other hand, as shown in Fig. 14, the total volumetric net water demand was greatest at Yeonan (109.5 Mt/year), Anak (103.6 Mt/year), and Baecheon (87.2 Mt/year). Baecheon and Yeonan showed the smallest depth of annul net water demand, but the volumetric amounts of water demand were greatest because the corresponding paddy areas were considerable as 15,193 ha and 18,568 ha, respectively. Conversely
Net water demand
Gross water demand
By depth (mm/year)
By volume (Mt/year)
By depth (mm/year)
639.0 573.7 587.2 596.5 592.3 648.7 639.8 597.2 621.7 590.1 638.9 614.2 622.9 627.7 604.2 621.3 609.5 605.9 589.7 612.5 611.7
103.6 87.2 20.9 38.3 81.5 67.5 35.8 40.2 12.1 12.0 66.8 12.7 23.0 69.5 10.7 5.6 26.1 65.6 109.5 27.8 (916.4)
973.8 892.1 909.0 920.6 915.4 985.9 974.8 921.5 952.2 912.6 973.6 942.8 953.6 959.6 930.3 951.6 936.9 932.4 912.2 940.7 939.6
By volume (Mt/year) 157.8 135.5 32.4 59.1 126.0 102.5 54.6 62.1 18.5 18.6 101.9 19.4 35.2 106.4 16.5 8.6 40.2 101.0 169.4 42.8 (1408.4)
Euncheon and Eunyool whose annual net water demand by depth was highest demonstrated comparatively small amounts of water demand. In conclusion, Hwanghaenamdo needed a total of 611.7 mm/year or 916.4 Mt/year of annual irrigation water for paddy fields on the average except lotmanagement requirement such as land preparation and transplanting. Concerning the lot-management requirement which generally is considered as 140 mm/day in South Korea, the amount of net water demand was 751.7 mm/year in the target site.
Fig. 13 – Average annual net water demand (mm/year) by administrative units showing the average was 611.7 mm/year for the whole paddy areas, 150,079 ha, of Hwanghaenam-do.
196
agricultural water management 89 (2007) 185–198
Fig. 14 – Volumetric annual net water demand (Mt/year) by administrative units showing the total demand was 916.4 Mt/ year for the whole paddy areas, 150,079 ha, of Hwanghaenam-do.
4.2.2.
Gross water demand
4.2.3.
Gross water demand (GW) for paddy fields derives from the sum of net water demand (WR) and canal-systemmanagement water demand (CMW) as the following equation:
GW ¼ WR þ CMW ¼
WR 1 WLRC
(4)
where WR includes lot-management water requirement and WLRC is water loss ratio in canal-system-management. CMW consists of conveyance losses, delivery management water and maintenance water requirement. Conveyance losses come out through evaporation and seepage from canals during conveying water from water sources to paddy fields, and are mainly determined by the type of canal and the length. The MAF (1998) recommended it as from 15% to 25% and from 5% to 7% of net water demand for earth lining canals and concrete lining canals, respectively. Delivery water requirement is the use of bypass water in paddy field irrigation to help maintain desired water levels in irrigation canals and to distribute water to paddy plots in a uniform manner (Kim et al., 2005), and maintenance water requirement is water flow to maintain canal systems during non-irrigation periods. Chung et al. (2006) revealed in their college textbook that the water loss rate in canal-systemmanagement (WLRC) was from 20% to 30% of gross water demand, and the MAF (1999) applied 20% to predict the water demand for paddy fields of South Korea in 2011. In this study, a constant rate, 20%, was also chosen to calculate the gross water demand of the target site. The calculation details were summarized in Table 6. The estimated gross water demand in Hwanghaenam-do province was 939.6 mm per year on the average, and was totally 1408.4 Mt/year.
Design water demand
Design water demand for agricultural use is defined as an amount of gross water demand that would be expected to recur on the average of every 10 years, and is referred to the design of agricultural water facilities to satisfy water requirement at least at the drought of 10-year return period. A time series of annual gross water requirement was first set up for each si or gun administration unit ranging from 1981 to 1999, and six different cumulative distribution functions (Gaussian,
Table 7 – Design water demand for a 10-year return period by administration units (si or gun) Administration (si/gun)
By depth (mm/year)
By volume (Mt/year)
Anak Baecheon Bongcheon Byeokseong Cheongdan Euncheon Eunyool Gangryeong Gwail Haeju Jaeryeong Jangyeon Samcheon Sincheon Sinwon Songhwa Taetan Woongjin Yeonan Yongyeon
1170.9 1085.2 1100.0 1103.1 1096.9 1208.4 1191.2 1098.2 1159.7 1097.5 1165.9 1145.8 1147.4 1150.6 1113.2 1148.2 1119.8 1109.5 1091.9 1136.3
189.8 164.9 39.2 70.8 151.0 125.6 66.7 74.0 22.5 22.4 122.0 23.6 42.3 127.5 19.8 10.4 48.0 120.2 202.7 51.7
Total
1132.0
1695.1
agricultural water management 89 (2007) 185–198
Log-Gaussian, Gamma, Log-Gamma, Gumbel and Weibull) were then examined to determine the best fit of the theoretical probability distribution with the gross water requirement data of 20 si or gun administration districts. In order to make time series data by si or gun administration districts, the annual gross water demand map was spatially interpolated using a geographic information system after the 19-year gross water requirement data were determined from each weather observation station. Eventually, a Gaussian probability distribution function was selected for all target districts using the KolmogorovSmirnov fitting test. The maximum deviation ranged from 0.088 to 0.154, which is less than the lower limits of the 95% confidence level value for sample sizes of 19, 0.301. By the Gaussian function the design water demand for a 10-year return period was determined to be 1131.97 mm/year (1695.1 Mt/year) on average in Hwanghaenam-do, about 20% more than the gross water demand, 939.6 mm/year. Calculation details are summarized in Table 7. The order of districts in terms of the quantity of design water demand almost coincided with that by the net water demand.
197
rainfall is less than in other regions, the water demand was determined to be relatively greater. In addition, the gross water demand was found to be 939.6 mm/year (1408.4 Mt/ year) considering lot-management water requirement and water loss rate in canal systems, 20%, and the design water demand for a 10-year return period was estimated as 1131.97 mm/year (1695.1 Mt/year). The results of this study are expected to be used as the basis of investigations into water use patterns and to establish and implement future agricultural projects.
Acknowledgements The US government retains ownership of Landsat data. Space Imaging and JAXA (the former NASDA) supported Min W. Jang in the acquisition of satellite data at a marginal cost. Topographic maps of North Korea, created by the former Soviet Union, were obtained with the assistance of Dr. Gene Fosnight with the UNEP GRID in Sioux Falls, South Dakota, USA.
references
5.
Summary and conclusion
This study was conducted to estimate agricultural water requirement of paddy rice in North Korea using the GIS technique with limited access to climate data and information on water management practices due to the political barriers in place. This study obtained paddy rice cultivation area using Landsat TM satellite images and GIS maps, and the cultivation area was incorporated into the estimation of paddy rice water demand to achieve the estimation of water demand for paddy rice cultivation in Hwanghaenam-do. Because paddy fields require flat lands for efficient irrigation and low temperatures to avoid crop damage, the classification accuracy was increased by incorporating spatial reasoning technique with topographical constraints including a slope and elevation of below 7% and 200 m above sea level. Image classification and subsequent spatial reasoning using GIS resulted in 150,079 ha of paddy fields, an area validated with 2001 area values from the FAO report (2001). When planning an irrigation project, it should be prior to estimate water volume needed inside a project area. However, in the case of Hwanghaenam-do, North Korea, it had to be assumed regarding the experiences in other project areas, Kyungki-do of South Korea, nearby and similar to the target area on account of the lack of accessible information. The water balance analysis for paddy fields was conducted using meteorological data. The FAO Blaney–Criddle method was adopted to determine potential evapotranspiration of paddy rice, and consumptive use was calculated using daily meteorological data from the GTS and 10-day crop coefficients for the northern portion of South Korea. Six observation stations in North Korea were used including Nampo, Sariwon, Yongyeon, Haeju, Gaesung, and Singye, where daily data were collected from 1975 to 1999. As a result, the annual net water demand was estimated to be 611.7 mm/year (916.4 Mt/year). In the northwestern areas of Hwanghaenam-do, where effective
Berry, J.K., 1995. Spatial Reasoning for Effective GIS. GIS World Inc., Colorado, USA. Cheon, M.G., Kim, S.J., Cheong, S.K., 2003. A study on Agricultural Water Use System of North Korea and the Improvement Methods. Korea Water Resources Association, Seoul, Republic of Korea (in Korean). Chung, H.W., 2005. Estimation of Crop Water Requirement in Korea. Hanchang Publisher, Seoul, Republic of Korea (in Korean). Chung, H.W., Kim, S.J., Kim, J.S., Noh, J.K., Park, K.U., Son, J.K., Yoon, K.S., Lee, K.H., Lee, N.H., Chung, S.O., Choi, J.D., Choi, J.Y., 2006. Irrigation and Drainage Engineering, 2nd ed. Dongmyeong Publisher, Seoul, Republic of Korea, (in Korean). Crawford, C., 1992–1993. Spatial versus verbal reasoning. The Journal of Computer Game Design 6. Online at http://www.erasmatazz.com/library/jcgd_volume_6/ spatial-_vs_verbal.html. Crist, E.P., Cicone, R.C., 1984. A physically based transformation of Thematic Mapper data: the TM Tasseled Cap. IEEE Geosci. Remote 22, 256–263. Doorenbos, J., Pruitt, W.O., 1977. Guidelines for Predicting Crop Water Requirements. Food and Agriculture Organization, Rome, Italy. Food and Agriculture Organization, 2001. Special Report: FAO/ WFP crop and food supply assessment mission to the Democratic People’s Republic of Korea. Online at http://www.reliefweb.int. Food and Agriculture Organization, 2005. Global map of irrigated areas. Online at http://www.fao.org/ag/AGL/ AGLW/aquastat/irrigationmap. Haque, M.A., Najim, M.M.M., Lee, T.S., 2004. Modeling irrigation water delivery schedule for rice cultivation in east coast Malaysia. Trop. Agric. Res. 16, 204–213. Jensen, J.R., 1996. Introductory Digital Image Processing: A Remote Sensing Perspective. Prentice-Hall, New York, United States. Jensen, M.E., 1974. Consumptive Use of Water and Irrigation Water Requirements. The American Society of Civil Engineers, New York, United States.
198
agricultural water management 89 (2007) 185–198
Kim, J.S., Oh, S.Y., Oh, K.Y., Cho, J.W., 2005. Delivery management water requirement for irrigation ditches associated with large-sized paddy plots in Korea. Paddy Water Environ. 3, 57–62. Korea Rural Economic Institute, 1996. Classification of Agricultural Region and Potential Agricultural Production in North Korea. Rural Development Administration, Seoul, Republic of Korea (in Korean, with English abstract). Korea Water Resources Cooperation, 2004. Dam Design Manual. Korea Water Resources Cooperation, Taejeon, Republic of Korea (in Korean). Lee, N.H., 1988. Simulating Daily Operation of Water Management System Of Irrigation Districts. Ph.D. Thesis. Seoul National University. Seoul, Republic of Korea (in Korean, with English abstract). Malano, H.M., George, B.A., Davidson, B., 2004. A framework for improving the management of irrigation schemes in Vietnam. In: Proceedings of a workshop of Australian Centre for International Agricultural Research, November 28, 2003, Hochiminh, Vietnam, pp. 16–18. Ministry of Agriculture and Forestry, 1998. Design Standard Of Agricultural Improvement Project: Irrigation. Ministry of Agriculture and Forestry, Seoul, Republic of Korea, (in Korean). Ministry of Agriculture and Forestry, 1999. Comprehensive Plan for Agricultural and Rural Water. Ministry of Agriculture and Forestry, Seoul, Republic of Korea (in Korean).
National Institute of Crop Science, 2005. Cultivation Methods by Direct Sowing for Paddy Rice. Online at http:// www.nces.go.kr/R_study/Rice/Cultural/Direct/driect01.htm (in Korean). Raine, S.R., Meyer, W.S., Rassam, D.W., Hutson, J.L., Cook, F.J., 2005. Soil–water and salt movement associated with precision irrigation systems: Research investment opportunities. Final report to the National Program for Sustainable Irrigation. CRCIF technical Report number 3.13/ 1. Cooperative Research Centre for Irrigation Futures, Toowoomba, Australia. Rogan, J., Chen, D., 2004. Remote sensing technology for mapping and monitoring land-cover and land-use change. Prog Plann. 61, 301–325. Rural Development Corporation, 1997. A study on the Water Requirement Variation with the Farming Conditions in the Paddy Field. Ministry of Agriculture and Forestry, Suwon, Republic of Korea (in Korean). Settle, J.J., Briggs, S.S., 1987. Fast maximum likelihood classification of remotely sensed imagery. Int. J. Remote Sens. 8, 723–734. Shin, D.F., Park, S.H., Park, K.Y., Kim, J.H., Lee, G.S., Lee, S.P., 1998. Agricultural Technology of North Korea. Osung Publisher, Seoul, Republic of Korea (in Korea). Shin, S.F., 1984. Data requirements for evapotranspiration estimation. J. Irrig. Drain. 110, 263–373. Smith, C., Brown, N., 1997. IMAGINE 8.3 Guide. ERDAS Inc., Atlanta, GA, USA.