2004 
UNCERTAINTY AND EFFECTS OF RESOLUTION OF DIGITAL ELEVATION MODEL AND 
ITS DERIVED FEATURES: CASE STUDY OF SUMBERJAYA, SUMATERA, INDONESIA 
A. Widayati , B. Lusiana, D. Suyamto, B. Verbist 
World Agroforestry Centre, ICRAF Southeast Asia, PO Box 161, Bogor 16001, Indonesia - 
a.widayati@cgiar.org; b.lusiana@cgiar.org; d.suyamto@cgiar.org; b.verbist@cgiar.org 
KEY WORDS: DEM, error, hydrology, resolution, reliability 
ABSTRACT 
Hydrological and erosion studies in Way Besai watershed in Sumberjaya, West-Lampung, Sumatera, Indonesia, require assessment 
of slope and flow pathways. À Digital Elevation Model (DEM) was generated from aerial photographs (1:24.000) using a softcopy 
photogrammetry approach. Error of elevation in a DEM affects derived slope, topographic index and catchment boundaries. 
Propagation of elevation error was evaluated on slopes both in a relatively flat terrain as well as in an undulating one. True elevation 
is unknown, but error can be estimated from uncertainty assessed with statistical methods following Monte Carlo simulation 
approach. Statistics grids as well as single parameters resulted were analyzed. Effects of DEM uncertainty on the derived slope is 
more pronounced in the flat terrain than in the undulating one. Within the scope and assumptions of this study, the effect of 
resolution on the slope uncertainty shows that higher resolution DEM creates larger slope uncertainty. Incorporation of spatial 
dependence in the assessment of crror propagation has strong effects on the apparent error of slope. Therefore, spatial dependence of 
DEM uncertainty should be considered when assessing error in spatial data, especially if slope is to be derived from the DEM. 
1. INTRODUCTION 
1.1 Background 
A good assessment of topography is of major importance in 
quantifying processes of erosion, sedimentation and water flow. 
Topography is commonly represented in a Digital Elevation 
Model, which can be produced from various sources and 
through different methods, and is used as a basis for modeling 
dynamic processes. 
To support hydrological and erosion studies conducted in Way 
Besai watershed, West-Lampung, Sumatera, Indonesia, two 
sources of Digital Elevation Model were used for the various 
levels of those studies. The whole watershed is covered by a 
DEM derived from Topographical Maps of 1:50,000, while for 
the sub-catchment level studies, a more detailed DEM was 
generated from aerial photographs of 1:24,000, which covers 
about 60% of the Way Besai watershed. The latter was able to 
capture micro relief in riparian areas that didn’t show on the 
topographic maps, but altered the views on slope and erosion. 
Error refers to difference between observed or recorded values 
and the corresponding true values (Hunter & Goodchild, 1997), 
supposing the latter is known. Root Mean Square Error (RMSE) 
is a valid statistical measure to estimate errors with the 
assumption that errors are random and normally distributed. 
In the production of spatial data, the true values are often 
unknown, or difficult to obtain. When the true value is 
unknown, uncertainty can substitute for the estimation of error. 
Uncertainty refers to lack of knowledge about the reliability of a 
Measurement in its representation of the true value (Wechsler, 
2000) or simply lack of knowledge of the true value (Hunter & 
Goodchild, 1997). 
  
* 
Corresponding author 
Because DEM error propagates to the derived products such as 
slope, DEM uncertainty also causes uncertainty in the derived 
products. For hydrological purposes, assessment of the 
uncertainty of e.g. the derived slope, topographic index and 
catchment boundaries is more important than uncertainty of the 
elevation itself. 
1.2 Objective of the Study 
The objectives of this study were to assess: 
1. Effects of DEM uncertainty on the uncertainty of the 
derived slope, 
2. Effects of the DEM cell resolution on the uncertainty of 
slope, 
3. Uncertainty of the derived slope when spatial dependence is 
incorporated compared to that when error is considered 
spatially-independent. 
2. DATA AND STUDY SITE 
The study site is part of Way Besai watershed, which almost 
coincides with the sub-district of Sumberjaya, located in 
Lampung Province, Sumatera, Indonesia. It covers a watershed, 
where large forest areas have been transformed over the past 
three decades into mosaics of smallholder coffee fields on 
slopes and rice paddies in the valleys. Ongoing studies of 
options for managing watershed functions include involvement 
by local governments and other stakeholders (Verbist, 2003). 
A 5km * 5km subset of the watershed was used for this study in 
Bodong area covering Way Ringkik subcatchment, where 
various hydrological and erosion studies are conducted. 
1013