Anthropogenic Changes on Land Cover and Its Impact on Actual Evapotranspiration

Tulisan ini memaparkan Perubahan distribusi vegetasi akibat kegiatan manusia serta dampaknya terhadap Perubahan evapotranspirasi aktual di Monsoon Asia. Perbandingan antara vegetasi aktual dan potensial menjadi indikator dari dampak Perubahan akibat kegiatan manusia. Kondisi vegetasi akual diidentifikasi dengan menggunakan citra satelit, sedangkan vegetasi potensial diekstrak dengan menggunakan data iklim. Dengan membandingkan distribusi vegetasi antara potensial dan aktual, ternyata bahwa Perubahan banyak terjadi di India, China, Indonesia dan Malaysia. Selanjutnya, dengan menggunakan analisis neraca air dilakukan perhitungan evapotransipirasi aktual untuk kedua kondisi tersebut dengan menggunakan data iklim yang sama, tetapi dengan nilai albedo yang berbeda sebagai penciri perbedaan antara kondisi vegetasi potensial dan actual. Perubahan a E berkisar antara 0-12% per tahun. Nilai 0 untuk mencirikan daerah yang tidak mengalami Perubahan akibat kegiatan manusia. Penurunan a E sebesar kurang dari 5% teridentifikasi di daerah yang mengalami Perubahan dari evergreen broadleaf forest (seasonal) ke padi sawahataupun dari hutan subtropikal menjadi lahan pertanian, seperti yang terjadi di Shandong (China), Uttar Pradesh (India). Penurunan a E mencapai 9% teridentifikasi pada saat hutan sub tropis berubah menjadi padi sawah, seperti yang terjadi di Assam (India), serta Guangdong dan Guangxi (China). Penurunan sebesar 12% terjadi pada saat hutan tropis berubah menajdi lahan pertanian seperti yang terjadi di Kalimantan Selatan (Indonesia) and Pahang (Malaysia).


INTRODUCTION
Knowledge of the evapotranspiration is important in global change research since this parameter is essential to the hydrological and climatic processes between the earth and atmosphere, which are performed by heat and water balance equations. Therefore, many researches have been focused not only on the microclimate aspects, but also in the regional, continental, and global level.
Relating to study of the land cover change and its impact on hydrologic changes especially for evapotranspiration parameter recently has been seriously observed, as reported by Running et al. (1996): Dickinson and Henderson-Sellers in 1988 simulated the Amazon basin with full forest cover, and then replaced with grasslands. The degraded grasslands reduced evapotranspiration so much that surface temperatures were predicted to increase by 3-5qC; Walker et al. in 1995 found that precipitation had been reduced by 1.2 mm/day due to reductions in evapotranspiration of 18% by land cover changes. Running et al. (1996) also pointed out that the global hydrologic changes resulting from land cover changes are predominantly of two kinds. The evapotranspiration was decreased when forested areas were changed to cropland and it was increased when the deserts were modified to irrigation land. Kondoh (1995) pointed out that over Monsoon Asia the potential evapotranspiration was changed around 0-200 mm/year because of land cover changes by human activities.
Following those research evidences, the purpose of this study is to investigate the land cover/vegetation changes that have occurred due to human activities and its impact on actual evapotranspiration in Monsoon Asia (-20°S -60°N, 60°E-160°E).

METHODS
To allow for land cover and a E comparisons, the method was defined in two ways ( Figure   1). The first is the current vegetation procedure that indicates the current vegetated surfaced determined with satellite data. The second way is potential vegetation that exists the vegetation distribution without human disturbance determined with climatic data. The difference between the two results of vegetation distributions and a E calculations are then mapped for comparisons.

Current vegetation classification
An alternative way to obtain the current vegetation distributions is by means of satellite data. Solar radiation in the visible and near-LQIUD UHG ZDYH EDQGV UHIOHFWHG E\ WKH (DUWK ¶V VXUIDFH DQG FROOHFWHG by remote sensing, can be combined into a spectral vegetation index such as the Normalized Difference Vegetation Index (NDVI) and related to the physical properties of the vegetation. The NDVI is generated from red and near-infrared (NIR) reflectance of satellite sensor by the following equation: (NIR-red)/(NIR+red). High values of this index are obtained for areas covered by green vegetation and low values for unvegetated areas and cloud covered.
There were some attempts to use the NDVI value for deriving land cover grouping at a continental or global scale. Some researcher applied the single or multi temporal datasets of the NDVI to land cover classification for Africa , Asia (Malingreau, 1986;Tateishi, 1997), Southeast Asia (Achard and Estrequil, 1995), Africa (Townshend and Justice, 1986), and global (DeFries and Townshend, 1994).
However, in this study, we combine the NDVI and the climatic data, since we consider the differences in climatic condition are able to show either the latitudinal or longitudinal variations associated with the distribution of vegetation. Therefore, the twelve monthly of each NDVI, temperature and precipitation data were applied together using unsupervised and maximum likelihood algorithm in order to obtain the homogeneous phenology classes for current land cover classification.

Potential vegetation classification
The next procedure after determination of the current vegetation distribution is to analyze the potential vegetation for the same vegetation class. By assuming the same phenology class resulted from previous step will have the same hydrological function, the temperature, cloudiness, precipitation, radiation, humidity index, as well as elevation of each class have been extracted from available ground-based global datasets. These extracted climatic data were applied to determine the distribution of potential vegetation by using Kira (1948) and Budyko (1974) methods. The two maps resulted were then compared with Leemans (1990) map to conclude the potential vegetation of each class. Kira (1945) introduced the Warmth Index (WI) concept to divide the Monsoon Asia region into seven main regions. Kira`s WI was defined as: WI =¦(Ti-5), where, i=1 to 12, T i > 5, and T i is the monthly temperature (qC). The WI value of each class was then associated with the following criteria to determine the type of its own forest. Thus, less than 0 is generally associated with the polar frost zone; between 0 and 18 with polar (tundra zone); between 18 and 45 with sub polar; between 45 and 85 with cool temperate; between 85 and 180 with warm temperate; between 180 and 240 with sub tropical, and more than 240 with tropical zones. Budyko (1974) used both the radiation balance and moisture condition in geographic zonality. A moisture condition in similarity with humidity index was calculated by dividing the HYDSRUDWLRQ DQG SUHFLSLWDWLRQ 7KRUQWKZDLWH ¶V (1948) method was applied to calculate the HYDSRUDWLRQ ZKLFK XVHV DLU WHPSHUDWXUH DV PLQLPXP LQSXW UHTXLUHPHQW %XG\NR ¶V PHWKRG GLYLGHG D zone in to 8 main groups or 22 sub regions. Detailed procedure of the classification method could be referred to Budyko (1974). Leemans (1990) created the Global Holdridge Life Zone Classification using biotemperature, mean annual precipitation, and 0 E , resulted in 38 different kinds of vegetation forms for whole the world. These image data could be extracted from NOAA/NGDC and EPA Disc A (1992) directly.

a E estimation
The a E has been calculated as a residual in water balance equations from estimates of potential evapotranspiration, , 0 E using a soil moisture reduction function as shown in Figure 2 that modified from Ahn (1995). There exist many methods for computing , 0 E which have been developed from different kind of climatic input data (Kondoh, 1994, Jensen et al, 1990. Because of the limited of available spatial image datasets, the 0 E in this study was calculated by using Priestly-Taylor method (1972). One assumption of this method is in the absence of advection condition, 0 E (mm), can be estimated from:

Elevation (m)
Water surplus (mm) Water deficit (mm)  Where n R and G are net radiation and heat flux density (W/m 2 ), respectively; r E is rate of net radiation (W/m 2 ), D is constant (1.26); ' is the gradient of the saturated vapor pressure at a certain air temperature; J is the psychometric constant; The weighting factor J ' ' is unit less; eq ET as the equilibrium evapotranspiration (mm) is defined as follows: As shown in equation (1) and (2), net radiation, n R , is a major variable of Priestly and Taylor method. Therefore, the reliability of the 0 E estimation is depended on accuracy of this variable. Ahn and Tateishi (1994) used the Linacre method to calculate n R : where s R is total (direct and diffuse) short wave solar radiation (W/m2), N n is the ratio of actual and maximum possible sunshine hours (%). T is temperature (oC) and r is albedo (%). The result of monthly 0 E analysis has been used together with monthly precipitation and soil water holding capacity ( c W ) on the next step to calculate a E , as shown in Figure 2.
Firstly, when the monthly 0 E is reduced with precipitation in mm/month ( P ), it is possible to obtain a clearer understanding of the times (as staring point i) of moisture deficiency and access, and the relative moistness or aridity of the climate. In those stations where precipitation in every month is greater than the potential evapotranspiration, the soil always remains full of water and a water surplus occurs. At other stations where precipitation is always less than potential evapotranspiration, there is not enough water for the vegetation to use and a moisture deficit occurs.
Secondly, the soil moisture status of each month, i , in mm/month, was calculated with accumulated potential water loss AWL and water holding capacity ( c W ), as follow: Thirdly, the a E can be derived from the rates of P , 0 E , and m S in following way: Detail procedure of the a E estimation used in this study could be referred to Thornthwaite and Mather (1957) and Ahn (1995).

DATASETS
To achieve the goal of this study, the data used in this study as following: e. Soil water holding capacity was generated by Bouwman Bouwman et al. (1993). The data was supplied by UNEP/GRID-Geneva.
To be consistent with the whole datasets, all of the image data were resampled to ten minutes or about sixteen kilometer grid dataset covering Asian region (-20°S -60°N, 60°E-160°E) and were reprojected on a linear latitude longitude grid.     Figure 3b shows the distribution of potential vegetation, which is dominated by forest. The tropical rain forest was distributed in almost whole region of Indonesia, Papua New Guinea, and Brunei. It also appeared in some parts of Philippines, India, and Laos. The tropical seasonal forest was largely distributed in India region and peninsula of Malaya. The largest area of sub tropical forest was distributed in southern China and Myanmar (Sagaing, and Kachin State), and Laos. Evergreen broadleaf forest was mainly distributed in lowland of eastern China, Japan (Chubu, and Kinki), and Myanmar (Chin State and Kachin State). Evergreen needleleaf forest was covered in a small part of Southeastern China (Fujian, Zhejang, Jiangxi, and Jiangxi) beside Kyushu of Japan. The Grassland (low sparse and cold), woody savanna, semi desert shrub, semi desert and bare desert were the same with the current vegetation, since these formation almost no changed.

Estimation of a E
In order to calculate the amount of a E change due to the land cover changes by human only or without climatic influences, the a E of both current and potential vegetations has been calculated with the same climatic input data, except the albedo variable to represent the land cover conditions. Different albedo layer has been utilized here to indicate the differences of both land cover classifications. Hence, the annual albedo of each type of vegetation form was selected and GHULYHG IURP 0DWKHZ ¶V 6HDVRQDO $OEHGR DQG Katoda`s monthly Albedo (1986). From Table 1, September is determined as initial month or starting point i, since the August was the last month of the period with precipitation more than potential evapotranspiration. Based on Thornthwaite and Mather (1957) because of the sum of all the 0 E P values is positive (221) the value of AWL with which to start accumulating the negative values of 0 E P is 0 (after August). From September to November the precipitation was less than potential evapotranspiration and soil moisture was utilized in evapotranspiration. We expected the runoff did not occur in this period. From December to August, precipitation is greater than potential evapotranspiration. At the time, soil moisture was recharged and runoff begins to occur.  T is air temperature; N n is cloudiness; E is elevation; r is albedo; P is annual precipitation; c W is water holding capacity; AWL is water accumulation potential water loss; m S is soil moisture; 0 E is potential evapotranspiration, and a E is actual evapotranspiration.  By applied this approach to 255 classes, the twelve monthly datasets of potential and actual evapotranspiration maps were produced to obtain the annual value for current and potential vegetation conditions. In general, there are some differences between both maps. The latitudinal banding of estimated 0 E results primarily from a large-scale climatic condition. The lowest values are distributed in Russia region (around 200 mm/year), and increased until equator region (more than 1800 mm/year). While based on the a E map, the value is distributed following the distribution of vegetation type. The lowest values (around 150 mm/year) are distributed in wide desert area of Russia, China, Mongolia, and Kazakhstan. The equatorial zones are bounded tropical regions such as Indonesia, Malaysia, Philippines, Papua New Guinea were associated with the tropical forest with the highest values per year (more than 1600 mm/year). Figure 5 shows also that the 0 E value is higher than a E . It is ranging from 150 mm/year for tropical rain forest, 386

Figure 4 and
mm/year for cropland, 273 mm/year for rice paddy, and 800 mm/year for bare desert area.

Land cover changed
Areal estimates of current and potential vegetation on the same class have been compared in Figure 9a to indicate which ecosystems have been modified by human activities. The largest focus of the alteration as was expected to be in almost the whole area of India (such as Madya Pradesh, Maharashtra, Andhra Pradesh, Uttar Pradesh, and Rajasthan), Russia (Novosibirsk, Altay, and Omsk), and some parts of China (such as Guangxi, Sichuan, Hunan, Hubei, Nei Mongol, Henan, Heilongjiang, and Shanxi). This also appeared in Myanmar (Pegu, Irrawaddy, and Magwe), Indonesia (such as Sumatera Selatan, Jawa Timur, Jawa Barat, and Kalimantan Selatan), Kazakhstan (Kustanay, Kokchetav, and Pavlodar), and Pakistan (Punjab and Sind). In other countries such as Papua New Guinea, Philippines, Korea, and Japan, also denoted a little modification.
Considering the type of land cover which has been changed by human activities, we realized that tropical rain forest has been modified to crops was occurred in Indonesia, such as in Kalimantan Selatan and Sumatera Selatan. Some places in Malaysia (Perak), Indonesia (Jawa Barat), and Sri Lanka (Central) show the changes from tropical rain forest to rice paddy area. The tropical seasonal forest that has been changed to crops was occurred in wide area of India (Uttar Pradesh, Madya Pradesh, etc). In Myanmar, I also realized a large area changed from tropical seasonal forest to rice paddy as well in Bangladesh (Khulna), Vietnam (Kien Giang), Cambodia (Preah Vihear), Sri Lanka (Uva, North western), India (Tamil Nadu, Andhra Pradesh). For the sub tropical forest that has been modified to crops was occurred in India (Punjab), while the modified to rice paddy was happened in China (Guangdong, and Guangxi), Bangladesh (Chittagong), and also a little part in Nepal (Bagmati). The evergreen broadleaf forest has been change to crops could be mainly recognized in China (Sichuan, Kiaoning and Guizhou). For other area like Jiangxi, Hunan, and Henan in China, I found that the modification from evergreen broadleaf forest to rice paddy area. I also identified a wide change from deciduous needleleaf forest to grass/crop in Russia (Vovosibirsk). In Kazahkstan such as Kustanay province, the low sparse grassland has been change to grass/crop.

Actual evapotranspiration changed
The amount of a E changes between potential and current vegetations is 0 to 180 mm/year or 0 to 12% per year, as shown in Figure 6 and Table 2. The 0 value indicates such area where the a E has no changed, while the positive value indicates the a E of current condition has been decreased from potential one. Figure 6. Changes of annual a E (%) due to land cover changes. Table 2 presents the recapitulation of mean annual a E value for some classes of vegetation types, which have been changed by human activities. The highest value (12%) was occurred when the tropical rain forest was changed to rice paddy. The decreased by 9% was happened when the sub tropical rain forest changed to rice paddy. The tropical rain forest which has been changed to cropland caused the a E decreased by 7% as well as the changed from tropical seasonal forest to rice paddy. The lower value (less than 5%) was happened when evergreen broadleaf forest (seasonal) changed to rice paddy. It also occurred when the subtropical rain forest has been changed to cropland.

CONCLUSION
This paper investigated the land cover changes over the Asian region, which has indicated by comparing current vegetation as imaged by current phenology satellite against a hydro climatic defined potential vegetation that would theoretically exist without human disturbance. By comparing those two yielded maps, we realized that India and China as the center of land cover changes. It also appears in tropics such as Indonesia, Kazakhstan, and Thailand. Some places in Japan, Korea and Mongolia were denoted that the cover changes also occurred but in a relative small area.
Based on a E calculation, the changes of a E between current and potential vegetation is around 0 to 12% per year. The 0 value indicates the area where the a E and land cover has no changed. The lower value (less than 5%) was happened when evergreen broadleaf forest (seasonal) changed to rice paddy. It also occurred when the subtropical rain forest has been changed to cropland. In addition, when the sub tropical rain forest changed to rice paddy, the a E was decreased by 9%. The highest decreased value (12%) was occurred when the tropical rain forest was changed to rice paddy.
I consider that further research maybe necessary to improved the outcome of this research by using more accurate datasets and research methods in order to obtain better understanding concerning the anthropogenic land cover changes and its influences on hydrological processes.