730 Bambang Sulis (Proposed Model).cdr PROPOSED ON LEVELS OF MODEL DEGRADED LAND AT MERAWU WATERSHED, BANJARNEGARA REGENCY, CENTRAL JAVA PROVINCE, INDONESIA BAMBANG SULISTYO1*, TOTOK GUNAWAN , HARTONO , PROJO DANOEDORO2 2 2 3and ROCHMAT MARTANTO 1 as 38371Faculty of Agriculture, Universit Bengkulu, Bengkulu , Indonesia 2 as Sleman, 55281Faculty of Geography, Universit Gadjah Mada, Yogyakarta , Indonesia 3 College 55293National Land , Sleman, Yogyakarta , Indonesia Received 30 November 2016/Accepted 10 June 2017 ABSTRACT Conservation of degraded land in Indonesia requires maps of degraded land. The maps were established based on a model developed in 1998 by the then Indonesia Department of Forestry. The model has 2 weaknesses i.e. 1. high level of uncertainty due to vector-based data used to build the thematic maps and 2. parameters redundancy or duplication from the model. This research was aimed to build up a proposed model on levels of degraded land at Merawu Watershed using fully raster-based data supported with remote sensing and GIS techniques. Parameters analyzed were Slope, Erosivity (R), Erodibility (K), Slope Length and Steepness (LS), Cover and Management (C), Support Practice (P) and Percentage of Canopy Cover. These data were presented in fully raster format. Management parameter was not explicitly used in this research because management parameter was already represented by the C and P parameters. Five parameters were directly obtained using fully raster format, i.e. Slope, LS, C, P and Percentage of Canopy Cover. The other 2 parameters went through spatial interpolation process before being presented as fully raster format. Correlation analysis among parameters was carried out. Parameters having high correlation coefficient (r ≥ 0.8) were excluded from the model to avoid redundancy. The proposed model only used parameters having low correlation coefficient. The research result showed that the determination of levels of degraded land was more accurate when using only erosion parameters, formulated as: Level of Degraded Land (LoD ) ≈ Erosion ≈ R x K x LS x C x PL . Keywords: Degraded land, GIS, raster, remodel, remote sensing INTRODUCTION Soil erosion management is a dilemmatic problem for developing countries such as Indonesia. Erosion is accelerating from year to year due to the uncontrollable pressure of population and economic growth Fulazzaky & ( Gany 2009; Sulistyo . 2009; Bohre & Chaubey et al 2014; Sulistyo 2015a). There are six types of land degradation i.e. water erosion, wind erosion, soil fertility decline, salinitation, water logging and the lowering of water table (FAO 1994). In most of tropical countries such as Indonesia, degraded land is mostly caused by water erosion resulted from high intensity of rainfall (Abdurachman & Sutono 2002), especially when land covers were removed. Soil of the bare land would be washed away by the runoff water to the lowest places or estuaries (Arsyad 2000). Identification of degraded land can be done by investigating watershed In the condition of areas. Indonesia, the number of critical highly s and eroded watershed areas increasing from year are to year (Hidayat . 2014; Kartodihardjo 2008).et al T ,o conserve the degraded land it is important to identify the objectives of the conservation and to disseminate these objectives to the stakeholders in the form of maps of degraded land (Tarigan 2012; & ). These Gibbs Salmon 2015 maps are usually developed from a model consisted of several chosen parameters, to reduce cost et al(Sulistyo . 2017). *Corresponding author: bambangsulistyounib@gmail. com; bsulistyo@unib.ac.id BIOTROPIA 4 3 7 220 229 Vol. 2 No. , 201 : - DOI: 10.11598/btb.201 .2 . .7 4 3 730 220 In Indonesia, the model and guidelines are published through Decree No. 041/Kpts/V/ 1998 of 21 April 1998 by the Directorate General of Reforestation and Land Rehabilitation, Depar tment of Forestr y (Depar temen Kehutanan 1998). This decree is updated through the Regulation of the Director General of Land Rehabilitation and Social Forestry on Guidelines for Monitoring and Evaluation of Watersheds, Number: P.04/V-Set/2009, dated 5 March 2009, but still using the same concept ( eKement rian Kehutanan The model implemented 2009). is nationwide. Mathematically, the level of degraded land (abbreviated as LoDL for further discussion) is formulated as: LoDL = w Slope + w Erosion + w Percentage 1 2 3 of Canopy Cover + w Management .. (1)4 where w , w , w and w are weighting factors .1 2 3 4 The weighting factors are determined based on the criteria of the LoDL that published and are presented in tables. Determination of the LoDL is initiated by computing estimated erosion using USLE the ( ) developed by Universal Soil Loss Equation model Wischmeier and Smith (1978), and formulated as: A = R K LS C P ......................................... x x x x (2) where: A = mean annual soil erosion rate in (ton/hectare/year) R = rainfall erosivity factor (R factor) in (MJ mm/ha/h/year) K = soil erodibility factor (K factor) in (ton hectare/MJ/mm) LS = slope length and steepness factor (LS factor) (dimensionless) C = cover and management factor (C factor) (dimensionless) P = support practice factor (P factor) (dimensionless) By replacing Erosion in equation (1) with equation (2), a new equation becomes: (3) LoDL = w Slope + w ( x x x x ) + w R K LS C P1 2 3 Percentage of Canopy Cover + w 4 Management ....................................... (3) Some weaknesses in the model are: 1) the input data are mostly simplified vector-based thematic maps leading to high uncertainty; and 2) the existing formula or model contains redundant or duplicate parameters. DeMers (2008) stated that the presentation of earth information in vector data model assumes the existence of homogeneity in a mapping unit as a result of classification and simplification, so that the variation of earth information is reduced, leading to uncertainties. Those uncertainties can be minimized by applying a fully-raster-based model which is able to keep the variation and objectivity of earth information ; (Aronoff 1989 Hadmoko 2007). The fully-raster-based model is among technological advances in Remote Sensing (RS) and Geographical Information System (GIS) fields. Sulistyo a studied land unit elimination (2015 ) analysis to determine LoDL using vector-based format. Land unit elimination causes a decrease in erosion rate, affects LoDL and finally influences the recommendations for conserving degraded land. Originally, USLE developed in is small agricultural areas in North America having characteristics of moderate climate (low in rain intensity), slope 3 - 18%, with consistent cropping and management system (Asdak 2007). Several modifications are needed when implementing USLE in loca t ions hav ing d i f f e ren t characteristics, especially modifications on parameters used to provide more reliable erosion estimation. Fistikoglu and Harmancioglu (2002) concluded that erosion model developed using USLE was more reliable when the analysis was carried out using small raster-based data. Sulistyo et al proved. (2009) that erosion modeling using fully raster-based data very - provided high accuracy compared to the actual erosion. A fully- raster-based erosion model uses raster-formatted data input. The raster-formatted data input are not products of Vector to Raster Conversion algorithm. Three USLE parameters (LS, C and P) can directly produce raster-formatted data, while data provided by the other two USLE parameters (R and K) has to be transformed using spatial interpolation to obtain raster-formatted data. Spatial interpolation is the process of using points with known values, to estimate the values of the other points (Chang 2008). Several parameters of USLE were analyzed respectively i.e. R, C, LS, and K (Sulistyo 2011a; Sulistyo 2011b; Sulistyo et al. 2011; Sulistyo 221 Proposed model on levels of degraded land at Merawu Watershed, Indonesia - Sulistyo et al. 2015b). Percentage of Canopy Cover was also analyzed (Sulistyo et al. 2013). All those parameters were analyzed using raster-based data format. This study was aimed to develop alternative model for LoDL at Merawu Watershed, using raster-based data format to avoid parameters redundancy or duplication. MATERIALS AND METHODS This research is the ultimate research on degraded land by the first author. Several preliminary studies on this subject were already published by the first author, i.e. Sulistyo et al. 2009; Sulistyo 2011a; Sulistyo 2011b; Sulistyo et al. 2011; Sulistyo et al. 2013; Sulistyo 2015b; Sulistyo 2016 and Sulistyo et al. 2017. Therefore, the data, hardware, software, research areas and methods used in this research were similar to those preliminary studies. Study Area The study area was located at Merawu Watershed, geographically lied at 10941'24" – 10950'24" E and 710'12" – 722'12" S, administratively located in Banjarnegara District, Central Java Province, Indonesia. Merawu Watershed covered an area of ±22,734 ha with 3 BIOTROPIA Vol. 24 No. 3, 2017 222 Figure 1 Study area main rivers flowing through the area from north to south i.e. Merawu, Urang and Penaraban Rivers (Fig. 1). Data and Methods Data used were topographical map, landform + map, remotely-sensed data of Landsat 7 ETM , rainfall data recorded at Merawu Watershed and its surroundings, as well as other data and reports related to the study. GIS software used were ILWIS (Integrated Land and Water Information System) version 3.4 and ArcView version 3.3. Several equipment used for field work were binoculars, compass, Hagameter, Munsell soil color chart, tape, soil sample ring kit, Auger soil boring tool, Global Positioning System tool and digital camera. Data from all parameters affecting LoDL were analyzed as fully-raster-based format. Pixel size used for this study was 30 x 30 m. LoDL model was developed using fully-raster-based data format (Fig. 2). The main research stages were creating all LoDL parameters, then implementing correlation analysis among parameters. To implement correlation analysis among parameters, all data had to be presented in quantitative form. To fulfil this requirement, the Management factor as qualitative factor had to be excluded from equation (3). Moreover, Management factor were represented by C and P factors. For further analyses, equation (3) was modified to: LoDL = w Slope + w (R x K x LS x C x P) + w 1 2 3 Percentage of Canopy Cover (4) ............. showedEquation (4) explicitly that the number of parameters Slope, R s LoDL were factor, K factor, LS factor, C factor, P factor, and Percentage of Canopy Cover . (7 parameters) Creatin arametersg LoDL P LoDL parameters were created by similar methods as the previous studies carried out by the first author (Sulistyo et al. 2009; Sulistyo 2011a; Sulistyo 2011b; Sulistyo et al. 2011; Sulistyo et al. 2013; Sulistyo 2015b; Sulistyo 2016; Sulistyo et al. 2017). The exception was the use of NDVI to create C factor; therefore, C factor was formulated as (Sulistyo et al. 2011): C = 0.6 – 0.77 x NDVI (5) ....................................... Percentage of Canopy Cover was formulated as (Sulistyo et al. 2013): % CC = 118.33 x NDVI – 17.605 (6) ..................... The NDVI can be extracted from satellite data to land cover according to obtain information on its density formulated as (Silleos et al.. NDVI was 2006): NDVI = (NIR – R) / (NIR + R) (7) ....................... where: NIR Near Infrared band= R Red band= 223 Proposed model on levels of degraded land at Merawu Watershed, Indonesia - Sulistyo et al. Figure 2 Flow diagram of the study Cor re lat ion Analysis among LoDL Parameters In relation to the raster-based degraded land modeling, each parameter contained spatial information pixel valueor . Pixel value for each parameter of the LoDL was then viewed as pixel value for each band or channel in remote sensing digital image. C , as one of multi-band orrelation analysis statistics in remote sensing analysis, was applied to determine the closeness of one band toward other bands (Ilwis User's Guide 2002; Lillesand et al. 2004). ndividual bands of a multi-spectral image I a i ying datare often highly correlated, mpl redundancy . and duplication Correlation matrix was implemented t o evaluate the degree of correlation between individual bands. Correlation matrix (a normali ed form of covariance matrix) z ha values in the range of -1 to 1 representing a s , strong negative correlation to a strong positive correlation respectively alues close to zero , . V represent correlation. weak Bands showing the least correlation, which implied the largest amount of image variations, were chosen to be included in the multi-band composite, prior to conducting multi-spectral classification analysis. LoDL parameters having the least correlation value were chosen to be included in the model. C orrelation value is the classified as “high” if correlation value is ≥ 0.8 Gordon et al. 1992). ( RESULTS AND DISCUSSION Maps of each LoDL parameters in raster- based data format are presented in Figures 3 to 9, while the correlation matrix of LoDL parameters is presented in Table 1. Table 1 shows that almost all LoDL parameters has correlation values i.e. between 0.01 (P low factor and K factor) and 0.44 (P factor and Slope). Percentage of Canopy Cover had high negative correlation value with C factor (r = -1.00), meaning that an increase in Percentage of Canopy Cover would cause a decrease in C factor. Slope had correlation value with LS factor (r high positive = 0.99), meaning that an increase in Slope would cause an increase in LS factor. et al to beRenard . (1997) formulated C factor calculated multiplying rior and se from P L U (PLU) ercentage of anopy over assessed and P C C for different over types (CC) and surface cover c (SC). inclusion of the ercentage This meant that P of anopy over in formula s C C LoDL wa redundant. Slope length and steepness factor (LS factor) consist of two subfactors the slope length (L) s , i.e. and the slope gradient (S). lope length (L) S is defined as: “the distance from the point of origin of the surface flow to the point where each slope gradient (S) decreases enough for the beginning of deposition or when the flow comes to concentrate in a defined channel” Wischmeier ( & Smith 1978). that inclusion of Slope This meant in formula redundant.LoDL was Based on wasthose results and reasons, it concluded that Percentage of Canopy Cover should not be used simultaneously with C factor in determining the LoDL Slope and that should not be used simultaneously with LS Factor. Therefore, should formula to determine LoDL only use parameters of USLE model because USLE considered all biophysical model has aspects affecting LoDL. revised model The to calculate LoDL is: LoDL ≈ Erosion ≈ x x x x (8)R K LS C P .......... The revised model is in line with the conclusion stated by Abdurachman and Sutono (2002) that the LoDL in Indonesia is mainly caused by erosion due to high amount and 224 BIOTROPIA Vol. 24 No. 3, 2017 Table 1 Correlation matrix of LoDL parameters C K LS P Slope % Canopy R C 1.00 0.09 -0.26 -0.32 -0.26 -1.00 -0.04 K 0.09 1.00 0.06 0.01 0.06 -0.09 0.21 LS -0.26 0.06 1.00 0.39 0.99 0.26 0.01 P -0.32 0.01 0.39 1.00 0.44 0.32 0.06 Slope -0.26 0.06 0.99 0.44 1.00 0.27 0.03 % Canopy -1.00 -0.09 0.26 0.32 0.27 1.00 0.04 R -0.04 0.21 0.01 0.06 0.03 0.04 1.00 225 Proposed model on levels of degraded land at Merawu Watershed, Indonesia - Sulistyo et al. Figure 3 Map of R factor Figure 4 Map of K factor Figure 5 Map of LS factor Figure 6 Map of C factor 226 BIOTROPIA Vol. 24 No. 3, 2017 Figure 7 Map of P factor Figure 8 Map of Slope Figure 9 Map of Percentage of Canopy Cover intensity of rainfall. The erosion terminology is used worldwide as an indicator of degraded land occurrence. This model is identical with the Level of Erosion Hazard (in Bahasa: Tingkat Bahaya Erosi or TBE) which is formulated in the Decree No. 041/Kpts/V/1998 of the then Indonesia's Ministry of Forestry. Level of Erosion Hazard was obtained by overlaying the erosion hazard map, derived from the USLE model, with soil depth map. The average of soil depth at Merawu W wa (Very Deep) Therefore atershed s 2.16 m . , the revised model of this study LoDL only co ed erosion . e s of nsider the occurrence Th result this revised LoDL model were simplified into 5 classes based on Sturges' Rule (Fig. 10; Table 2). Table 2 shows that 2,151 ha (9.5%) of Merawu Watershed were areas having vegetation index ≤ 0. Those areas were areas with water. Merawu Watershed area had 15,891 ha (70%) of Very Low and levels of degraded land. The other 4,689 Low ha (20.6%) had and levels Moderate, High Very High of degraded land. CONCLUSIONS Percentage of Canopy Cover should not be used simultaneously with C factor in determining the LoDL. Slope should not be used simultaneously with LS Factor. The best scenario was to use only erosion parameter derived from the USLE model to determine the LoDL model. The proposed LoDL model: LoDL ( ) ≈ Erosion ≈ R Level of Degraded Land x K x LS x C x P. Proposed model on levels of degraded land at Merawu Watershed, Indonesia - Sulistyo et al. Table 2 The area of each class of the LoDL at Merawu Watershed No Level Erosion (mm/year) Area (ha) Area (%) 1. Very low 0.01 to 1 12,470 54.9 2. Low 1.01 to 2 3,421 15.1 3. Moderate 2.01 to 3 1,714 7.5 4. High 3.01 to 4 982 4.3 5. Very high > 4 1,993 8.8 6. Areas having vegetation index 2,151 9.5 Total 22,734 100.0 (-) Figure 10 Results of the revised LoDL model 227 ≤ 0 ACKNOWLEDGEMENTS The authors thank the Indonesian Directorate General of Higher Education for providing grant to conduct this research. REFERENCES Abdurachman A, Sutono. 2002. Soil erosion control technology. Research and Development of Soil and Agro-climate. Bogor (ID): The Agency for Agricultural Research and Development. Aronoff S. 1989. Geographic Information Systems: a management perspective. Ottawa (CA): WDL Publication. Arsyad S 2000 Konservasi anah an ir IPB . . t d a . Bogor (ID): Press. Asdak C. 2007. Hidrologi dan engelolaan aerah liran p d a s Yogyakarta (ID):ungai. Gadjah Mada University Press. Bohre P Chaubey OP. 2014. Restoration of degraded lands , through plantation forests. 14:18-27.GJSFR Chang KT. 2008. Introduction to geographic information s New York (US) :ystems. McGraw-Hil l International. DeMers MN. 2008. Fundamental of eographic g i s . New York (US):nformation ystems John Wiley & Sons. Departemen Kehutanan 1998 Guideline for the . . a p l rrrangement of lanning for and ehabilitation in a w a . (ID)atershed rea Jakarta . F A O 1994 Land ood gricultural rganization [FAO]. . degradation in South Asia: its severity, causes and effects upon the people World Soil . Rome (IT): Resources Reports. Fulazzaky MA Gany AHA. 2009 Challenges of soil erosion , . and sludge management for sustainable development in Indonesia. J Environ Manage 90:2387-92. Fistikoglu O Harmancioglu NB 2002 Integration of GIS , . . with USLE in assessment of soil erosion. Water Resour Manag 16:447-67. Gibbs HK, Salmon JM. 2015. Mapping the world's degraded lands. Appl Geogr 57:12-21. Gordon ND, McMohan TA, Finlayson BL. 1992 Stream . h a i e . New York idrology: n ntroduction for cologist (US): . John Wiley & Sons Hadmoko DS. 2007. Toward GIS- ased integrated landslide b hazard assessment: critical overview. Indon J a es Geogr 39:55-77. Hidayat L, Susanto S, Sudira P, Jayadi R. 2014. Penilaian kinerja pengelolaan daerah aliran sungai berbasis model hidrologi elementer kasus: daerah tangkapan air Waduk Mrica. Agritech 34:337-46. International Institute for Aerospace Survey and Earth Sciences Ilwis User's Guide. Enschede. 2002. (NL): ITC-ILWIS. Kartodihardjo H. 2008. Management of natural resources and environment. Bengkulu (ID): Universitas Bengkulu. Kement rian Kehutanan 2009 Peraturan Direktur e . . Jenderal Rehabilitasi Lahan dan Perhutanan Sosial tentang Pedoman Monitoring dan Evaluasi Daerah Aliran Sungai, Nomor: P.04/V-Set/2009, Tanggal: 5 Maret 2009. Jakarta (ID). Lillesand TM, Kiefer RW, Chipman J. 2004 Remote ensing . s an mage nterpretationd i i . Fifth edition. New York (US): . John Wiley & Sons Renard KG, Forster GR, Weesies GA, McCool DK Yoder , DC. 1997. Predicting soil erosion by water: guide a to conservation planning with the Revised Universal Soil Loss Equation (RUSLE) . Washington DC (US): US D A epartment of griculture. Silleos NG, Alexandridis TK, Gitas IZ Perakis K. 2006. , Vegetation ndices: dvances ade in iomass i a m b e v m lstimation and egetation onitoring in the ast 30 y .ears Geocarto Int 21:21-8. Sulistyo B, Gunawan T, Hartono, Danoedoro P. 2009 . Toward a fully and absolutely raster-based erosion modeling by using RS and GIS Indones J Geogr. 41:149-70. Sulistyo B, Gunawan T, Hartono, Danoedoro P 2011 . . Pemetaan faktor C yang diturunkan dari berbagai indeks vegetasi data penginderaan jauh sebagai masukan pemodelan erosi di DAS Merawu (C actor f m d f v v iapping erived rom arious egetation ndices of r s d a i f e memotely ensed ata s nput or rosion odeling a . t Merawu Watershed) JML 18:68-78. Sulistyo B. 2011 Pengaruh rosivitas ujan yang iperoleh a. e h d dari umus yang erbeda terhadap emodelan rosi r b p e b r kerbasis aster (Studi asus di DAS Merawu, Banjarnegara, Jawa Tengah) (The ffect of ain e r e g d f arosivity enerated from ifferent ormulae on r b e .aster- ased rosion) Agritech 31:250-9. Sulistyo B. 2011 The fect of hoosing ifferent ontour b. ef c d c i o a f r b e mnterval n ully aster- ased rosion odeling (The ase n Merawu Watershed, Banjarnegara, c i Central Java) 16:258-66.. JTS Sulistyo B, Gunawan T, Hartono, Danoedoro P. 2013 . Pemodelan persentase tajuk di DAS Merawu yang diturunkan dari berbagai indeks vegetasi data penginderaan jauh (Modeling of ercentage of p c a c area d fanopy t Merawu atchment erived rom v v i r s d . arious egetation ndices of emotely ensed ata) For Geo 27:23-32. Sulistyo B. 2015a. Kajian perubahan tingkat kekritisan lahan sebagai akibat proses eliminasi unit lahan: studi kasus di kawasan pertambangan Danau Mas Hitam, 228 BIOTROPIA Vol. 24 No. 3, 2017 Provinsi Bengkulu (The study of land degradation level as a result of land unit elimination process: case study in mining region of Danau Mas Hitam, Bengkulu Province). National Seminar on Biodiversity, University Club. Yogyakarta (ID): University of Gadjah Mada. p. 828-33. Sulistyo B. 2015b. Pemodelan faktor K berbasis raster sebagai masukan pemodelan erosi Di DAS Merawu, Banjarnegara, Provinsi Jawa Tengah (Modeling of raster-based of K factor as input for erosion modeling at Merawu Catchment Area, Banjarnegara, Central Java Province). Jurnal Manusia dan Lingkungan 22:240-6. Sulistyo B. 2016. The effect of choosing three different C factor formulae derived from NDVI on a fully raster- based erosion modeling. IOP Conf Ser: Earth Environ Sci 47(1):1-10. doi:10.1088/1755- 1315/47/1/012030. Sulistyo B, Gunawan T, Hartono, Danoedoro P, Listyaningrum N. 2017. tAbsolute accuracy of he erosion model of DEM-NDVI nd t s a i ' modification. IJG 13:23-34. Tarigan SD. 2012. Methods for delineating degraded land at Citarum Watershed, West Java, Indonesia. J Tropical Soils 17:267-74. Wischmeier WH, Smith DD 1978 Predicting ainfall . . r e l a g c prosion osses: uide to onservation lanning. Washington DC (US): USDA Agriculture Handbook No. 37. 229 Proposed model on levels of degraded land at Merawu Watershed, Indonesia - Sulistyo et al. Page 1 Page 2 Page 3 Page 4 Page 5 Page 6 Page 7 Page 8 Page 9 Page 10