Navigating the Deluge_ Dubai Floods and the Resilience of Dubai International...
Hydrological mapping of the vegetation using remote sensing products
1. Bulletin of the Transilvania University of Braşov • Vol. 3 (52) - 2010
Series II: Forestry • Wood Industry • Agricultural Food Engineering
HYDROLOGICAL MAPPING OF THE
VEGETATION USING REMOTE SENSING
PRODUCTS
M.D. NI Ă1 I. CLINCIU1
Abstract: This paper proposes a new method based on FCD (Forest Cover
Density) model, which maps the vegetation taking into consideration the
canopy density. The critical analysis of well-known methods from Romanian
literature, concluded that the age and stand consistency are the main factors
in hydrological mapping of the vegetation. In the new technique we
integrated several vegetation indexes, which made possible a hydrological
mapping based entirely on remote sensing products. We have applied the new
method in Upper Tărlung Watershed, and the results obtained revealed a
strong correlation with the results offered by classical methods.
Key words: hydrological mapping, remote sensing, vegetation index.
1. Introduction evaluation and quantification of retention
and infiltration at a specific rain, under
The hydrological mapping of forest conditions encountered in watershed [1].
vegetation is an activity used for some The characteristic coefficient for the
hydrological parameters determination. hydrological categories established by the
These parameters characterize the vegetation hydrological mapping system proposed by
for mathematical quantification of the Apostol in 1972 plays a very important
hydrological processes which happen within role in the estimation of retention. This
the hydrographic watersheds. mapping system has been completed in
The most common synthetic quantification 1973-1987 period by M. Ionescu, P.
of the hydrologic mapping of the vegetation Dumitrescu and N. Lazăr, and it was
is represented by the flow coefficients. adapted by Păcurar for computer application
These coefficients are calculated after the in the GWBASIC programming language
analysis of the factors that influence the [2]. In the next period, the dichotomic
hydrological balance between the flow and schematics used in the structure mentioned
the rain [1]. in [1] allowed the development of
According to the peak discharge computer based mapping modules that can
calculation methodology of small, be found in today’s geographic
predominantly forested watersheds, the information systems like IDRISI [2] or
flow coefficient estimation is based on the ESRI ArcMap [5].
1
Dept. of Forest Management Planning and Terrestrial Measurement, Transilvania University of Braşov.
2. 74 Bulletin of the Transilvania University of Braşov • Vol. 3 (52) - 2010 • Series II
The mapping system proposed by represented by the precision of the
Apostol and the method proposed by measurement of subcompartment borders.
Păcurar require as input data certain
information about the watershed conditions 2. Objectives
which can be found in the Forest
Management of that area. The system The calculus of flow coefficients is
proposed in 1972 did not offer a clear influenced directly by the hydrological
methodology of using the mapping system, mapping of vegetation. Irrespective of the
the schematics “having a qualitative nature methods they employ - such as using a
and include descriptive elements for each specific term (ex. S term in SCS-CN
category and subcategory” [2]. Therefore, method, which represents the maximum
the method proposed by Păcurar offers a value of the sum between the total retention
clear structure, developed for computer and the infiltration quantum); or using
application, based on quantitative facts. representative land cover categories (ex.
The Păcurar method was created Apostol method); or using simplified
specifically to be applied in programs procedures such as the Frevert method or
which work with punctual data as well as the Mo oc method - the mapping procedures
in GIS environments that work with are governed by the same developing
gridded data. directions:
The mapping systems mentioned above • Each method takes as main data input
use as input data information about the the presence or absence of the vegetation;
status of the soil, the stand age, the stand • The land cover is categorized in the
consistency and the production class. With following main categories: forests, pastures,
the aid of the last criteria, these systems agricultural lands and bare soil lands;
succeed in characterizing the forest • Qualitative and quantitative information
influence to the flow process, better than is used for the definition of further categories
the mapping system proposed by Frevert. and/or subcategories.
The spatial and temporal precision of the In the present paper the main objective is
simulation as well as the quality of the the realization and validation of a new
results are in direct link with the data used method of hydrological mapping of the
as input. Therefore, by using the data vegetation based on information from
retrieved from the forest management plan, remote sensing products.
the method can encounter several temporal The characterizing elements presented
and spatial difficulties: above influence each other; however, in
• First, the forest management plan is many cases the stand age and the
updated once at every 10 years, and during consistency influence decisively the other
this period modifications may appear in the elements. Therefore, the usage of satellite
structure of the vegetation within the images, aerophotograms or remote sensing
studied watershed; products in hydrological mapping of the
• Second, provided the application of a vegetation can contribute significantly to
silvicultural work within a large sub- the estimation of principal characterization
department is known, the development of elements (age and consistency).
the work cannot be characterized and Temporal and spatial problems that
determined thoroughly (ex. the position of appear in traditional methods can be
stripes from striped clear-cuts, the position minimized with the help of the new
of patches from progressive cuts); technique, which is based on remote
• Finally, another spatial issue is sensing products.
3. Ni ă, M.D., et al.: Hydrological Mapping of the Vegetation Using Remote Sensing Products 75
3. Material and Methods The crown arrangement found in the
forest stand leads to a shadow pattern that
The hydrological mapping method of the affects the spectral responses. The shadow
vegetation is based on the FCD model index SI quantifies this pattern and is able
structure, which was developed [3], [4] on to differentiate between young stands and
data input taken from Landsat TM satellite mature stands. The young stands have a
images. The FCD model analyzes the lower canopy shadow index (SI) compared
vegetation on the basis of four vegetation to the mature natural forest stands. The
indexes: Advanced Vegetation Index older forest stands show flat and low
(AVI), Bare Soil Index (BI), shadow Index spectral axis in comparison with the open
or scaled Shadow Index (SI, SSI) and area. SI has been calculated using Eq. (3):
Thermal Index (TI).
The results obtained after the model run SI = 3 (256 − B1) ⋅ (256 − B 2) ⋅ (256 − B 3 ) ,
offers information for monitoring the
canopy density, indicating the forest status (3)
in the acquisition time of satellite image.
The advanced vegetation index (AVI) is where: B1, B2, B3 - the 1st, 2nd, 3rd band of
related to the NDVI index, but is better LANDSAT image.
that the latter in highlighting subtle For the mapping method application a
differences in canopy density. AVI uses LANDSAT TM, with the acquisition year
the power degree of the infrared response 2009, was downloaded from LANDSAT
and this characteristic is making AVI more repository.
sensitive to forest density and physiognomic
vegetation classes. In our research, AVI 4. Results and Discussions
has been calculated using the Equation (1):
For validation, the method for hydrological
mapping of the vegetation was applied in
AVI = 3 (B 4 + 1) ⋅ (256 − B 3) ⋅ B 43 , (1)
Upper Tărlung Watershed, upstream the
Săcele lake.
where: B3, B4 - the 3rd and the 4th band of The procedure of hydrological mapping
Landsat image; B43 - the difference between method contains the following five steps:
B4 and B3. 1. Satellite image calibration;
Bare soil areas, fallow lands and 2. Vegetation indexes calculation: AVI,
vegetation with marked background BI and SI;
response are enhanced using the bare soil 3. Division of the vegetation index values
index. Similar to the concept of AVI, the in three distinct classes (Low, Medium, High),
bare soil index (BI) is a normalized index taking into consideration the value variance;
which separates the vegetation with 4. Combination of the vegetation index
different background visibility in classes;
completely bare, sparse canopy and dense Vegetation index combination identification
canopy. In our research we have used the and vegetation categorization in the
following formula to calculate BI: hydrological categories proposed by Apostol
(Table 1).
(B 5 + B 3) − (B 4 + B1) The vegetation indexes have different
BI = ⋅ 100 + 100 , (2)
(B 5 + B 3) − (B 4 + B1) behaviour as it follows (Figure 1):
• AVI offer information about all
where: B1, B3, B4, B5 - the 1st, 3rd, 4th and vegetation elements, both for forested areas
5th band of LANDSAT image. and pastures;
4. 76 Bulletin of the Transilvania University of Braşov • Vol. 3 (52) - 2010 • Series II
Composed vegetation index classes grouping in hydrologic categories Table 1
Hydrologic category
Specifications
A B C D1 D2
L-M-L; L-M- M-M-M; M-
L-L-L; L-L-
M; L-H-M; M-L; H-L-M;
AVI – SI – BI H-H-L; H-H- L-H-L; M-H- M; L-L-H; L-
M-L-L; M-L- H-L-H; H-M-
combination M; H-H-H L; M-H-M M-H; L-H-H;
M; M-M-L; L; H-M-M;
M-L-H
M-H-H H-M-H
where: H = High; M = Medium; L=Low
• SI takes higher values with the
decrease of the forest density;
• The increase in the percent of bare
soil determines a subsequent increase in
the value of BI.
Fig. 2. Vegetation index class division
The result is a thematic raster which
Fig. 1. Vegetation indexes characteristics contains on every pixel the hydrological
against the forest condition [4] class of the vegetation located in those
coordinates (Figure 3).
In method application, five hydrological After cropping the Landsat image on the
classes are obtained according to those region of interest, the pixel frequency
proposed by Apostol. distribution of the product offered by the
In order to combine the vegetation proposed method shows a high percent of
indexes, they were divided in three classes: forest vegetation in the area. This is one of
low, medium and high, according to the the advantages of the results offered by the
statistical distribution shown in Figure 2. method, which fulfils the first condition of
After reclassification, the final step in classical methods, namely to identify the
mapping the vegetation is the hydrological presence or the absence of the vegetation.
class extraction. The method is developed Knowing, at pixel level, the land cover
on a raster calculation algorithm, which category offers the first information both
creates complex codes to every pixel, from quantitative and qualitative on forest
simple codes (Table 1). degree in the studied watershed.
5. Ni ă, M.D., et al.: Hydrological Mapping of the Vegetation Using Remote Sensing Products 77
Fig. 5. Correlation between Frevert
coefficients and new method coefficients
The flow coefficients of 23 watersheds
were calculated with classic methods and
the new method and their values compared.
Fig. 3. Hydrological classes thematic The results of the comparison of showed
raster for Upper Tărlung Watershed a high correlation between traditional
methods and the new method, as can be
In the Upper Tărlung Watershed area seen in Figure 5.
85% was identified as forested and 31%
fell in category A, 51% fell in category B 5. Conclusions
and 3% in category C (Figure 4).
The rest of the area was identified as By adapting the application method from
being comprised of 14% pastures and 1% FCD model for hydrological mapping of the
bare soil (i.e. rocks, roads and agricultural vegetation a new method was created based
land). on remote sensing products. The results of
the method put it on a quality scale between
Frevert method (which uses few data
input) and Apostol method (which uses
many data taken from forest management).
Regarding the applicability, the proposed
method proves to have a very high
precision when applied to extended areas.
The high correlation between the flow
coefficient values determined from forest
management data and those determined
with the new method, gives the proposed
method a high level of trust.
Fig. 4. Hydrological class distribution of Using the remote sensing products offers
vegetation from Upper Tărlung Watershed advantages, the most important being the
actuality of the data. Using vegetation
Having the land cover category at pixel indexes, the obtained results offer an
level offers easy manipulation of data. From accurate mapping of the real situation in
this point, it was easy to determine the the field, indirectly improving the flow
flow coefficient determination on the small coefficient calculation and implicitly, the
watersheds from Upper Tărlung Watershed. peak discharge prediction.
6. 78 Bulletin of the Transilvania University of Braşov • Vol. 3 (52) - 2010 • Series II
Another advantage is given by the high University Publishing House, 2001.
degree of coverage of the product offered. 2. Pacurar, V.D.: O nouă metodă de
The application of the method has a cartare hidrologică a terenurilor
minimum unity with 0.1 ha area, which forestiere utilizând sistemele de
makes it suitable for the mapping of small informa ii geografice (A New Method
areas. At the same time, the product gives of Hydrological Mapping of Forested
the possibility of mapping larger zones Lands using Geographic Information
with the help of LANDSAT images. Systems). In: Revista Pădurilor 2
The method shows some disadvantages (2005), p. 28-31.
at processing (level of expertise in remote 3. Rikimaru, A.: Landsat TM Data
sensing must be intermediate) and inter- Processing Guide for Forest Canopy
pretation (clouds occurrence); nevertheless, Density Mapping and Monitoring
it offers a better alternative of hydrological Model. In: Proceedings of ITTO
mapping of the vegetation especially in Workshop on Utilization of Remote
zones where is no GIS data is available or Sensing in Site Assessment and
where such data are out of date. Planning for Rehabilitation of
The offered product ensures easy Logged-Over Forest, Bangkok,
manipulation in GIS in raster format. Thailand, July 30-August 1, 1996, p.
Furthermore, the proposed method proves 18-24.
to be useful not only in small watersheds 4. Rikimaru, A., Roy, P.S., Miyatake, S.:
but in larger basins too. Tropical Forest Cover Density Mapping.
In: Tropical Ecology 43 (2002) No. 1,
Acknowledgements p. 39-47.
5. Tamaş, S., Clinciu, I.: Cercetări
This paper is supported by the Sectoral privind posibilită ile de utilizare a
Operational Programme Human Resources sistemelor de informa ii geografice
Development (SOP HRD), financed from (GIS) în fundamentarea hidrologică a
the European Social Fund and by the proiectării lucrărilor de amenajare a
Romanian Government under the contract bazinelor hidrografice toren iale.
number POSDRU/6/1.5/S/6. (Research on Usage of GIS in
Hydrological Substantiation of
References Management Works in Hydrographic
Torrential Watersheds). Scientific
1. Clinciu, I.: Corectarea Toren ilor Report, Braşov. Transilvania University
(Torrent Control). Braşov. Transilvania Publishing House, 2006.