Soil moisture mapping for drought monitoring in urban areas

Soil moisture is an important indicator for drought monitoring. Mapping of soil moisture in this study uses remote sensing, namely Landsat 9 OLI imagery, because it can be relied upon as a cheap source of information, and its good temporal resolution or revisit period. Two parameters indirectly related to soil moisture, namely vegetation were analyzed using the Normalized Difference Vegetation Index (NDVI) approach, and land surface temperature (LST). Drought analysis was verified using the Normalized Difference Drought Index (NDDI). The remote sensing imagery used in this study is Landsat 9 OLI imagery by selecting images with 30% cloud cover from 1 January 2022 to 31 December 2022 with the support of the cloud-based Google Earth Engine computing platform. The results of the analysis indicate high LST values in the southern part of the study area whose dominant land use is built-up areas, namely in the sub-districts of Depok, Gamping, Ngaglik, and Mlati. The effect of vegetation on soil moisture is indicated by the NDVI value, which has a relatively strong positive correlation with SMI (R= 0.46). The SMI value is in contrast to LST, where the spatial distribution of high SMI is spread in the northern part, namely Pakem, Cangkringan, and Turi districts. On the other hand, a low SMI is spread across the central and southern parts of the study area, which have a high drought index (extreme moderate). Overall, it is concluded that the SMI has the potential to map drought and is a reliable index for initial analysis of drought risk management.


Introduction
Soil moisture is an important variable in the climate system that can be used to predict drought, floods and the impact of climate change.In addition, soil moisture can affect the distribution of vegetation.Soil moisture is the main water resource for agriculture and natural vegetation, which is not only influenced by energy and moisture but is also influenced by runoff [1].Soil moisture may increase due to increased rainfall [2].Much research has been done on soil moisture in Indonesia, as well as drought and its impact on agriculture have been widely studied, but studies on soil moisture in urban areas have not been explored much [3].Many problems are caused by drought in urban environments such as land subsidence that can increase the urban heat island effect [4][5][6].Drought impedes nature's ability to provide various social-environmental benefits, climate change adaptation and mitigation.Anthropogenically induced loss of vegetation cover leads to an increase in built-up land and impermeable surfaces in urban areas, resulting in changes in monthly and annual average temperatures [7].
Soil moisture monitoring is carried out using a remote sensing approach because it can simultaneously monitor land surface temperature and vegetation index.Several previous studies used remote sensing in mapping soil moisture with various approaches including land surface temperature 1314 (2024) 012087 IOP Publishing doi:10.1088/1755-1315/1314/1/012087 2 [8,9], linear decomposition algorithm of mixed pixels [10], and transformation of the vegetation index [11,12]].In this study, soil moisture was estimated using the LST approach to derive soil moisture index (SMI) values and the normalized difference vegetation index (NDVI) approach to obtain information on vegetation density.LST can be measured using remote sensing techniques, which are more cost effective than traditional soil moisture measurement methods [13,14].LST is a key variable for studying urban land surface processes and surface-atmosphere interactions, being an important component in studies of surface energy and water availability [15].Although vegetation is the most influential parameter on soil moisture [16], it cannot be used as an indicator for estimating soil moisture because the thermal inertia of the vegetation canopy is relatively lower than the thermal inertia of the soil [17].The minerals that make up the soil have a greater density than the leaves, and the mineral constituents and organic matter in the soil have a higher heat capacity than the leaves [18].In addition, soil moisture has a marked effect on soil thermal properties such that higher soil moisture further increases the thermal inertia of the substrate relative to the vegetation canopy [19].Apart from vegetation, surface temperature is another important indicator often used in estimating soil moisture.The existence of thermal inertia in the soil affects the temperature of the surrounding surface.
The case study was conducted in Sleman district, Yogyakarta, Indonesia, which is the most populous district in Yogyakarta due to the high level of urbanization and the dynamic increase in built-up land [20].On the one hand, the Sleman Regency is a water catchment area in the Special Region of Yogyakarta, including an area with low groundwater potential [21].Water catchment areas are feared to decrease with the increase in built-up areas.Soil moisture mapping in suburban areas helps provide valuable information for urban planning and decision-making regarding land use change [3,22].Estimating drought vulnerability spatially in urban areas is a step towards planning for adaptation to climate change and reducing vulnerability.

Methods
The The stages in this study are described in a flowchart (Figure 2), where the process begins with filtering Landsat 9 2022 imagery on the Google Earth Engine using the following script: Drought analysis was performed using NDDI derived from the Normalized Difference Water Index (NDWI) and NDVI formulas.NDWI is used to see open water features in satellite imagery, which allow the body of water to stand out from the ground and vegetation [23,24].

Figure 2. Methodology flowchart of soil moisture index calculations
Humidity index values range from 0 to 1, with 0 indicating extreme dry conditions and 1 indicating wet conditions [25].Calculating the soil moisture index using a remote sensing approach is based on the relationship between the LST and NDVI.LST is a condition that is controlled by the balance of surface energy, the atmosphere, the thermal properties of the surface and subsurface media [26][27][28][29].If the surface temperature is low, then the surface has low infiltration indicating high soil moisture [30].In Potic [31] classifies the SMI as ranging from 0 to 1 which describes the level of soil moisture.The SMI formula is written as follows [30]: Where, SMI = soil moisture index (3) Where, a1,a2,b1,b2 is obtained from linear regression which defines the warm and cold edges of the data.
The Normalized Difference Vegetation Index is a transformation of the vegetation index, which is widely used to show the proper status of vegetation growth [10,32,33] .NDVI values are classified based on the range of values that refer to previous studies [34] presented in Table 1.In the drought level analysis using the NDDI formula in [35].

Results and Discussion
The condition of the vegetation in Sleman Regency based on the results of the NDVI analysis shows that low non-vegetation classes dominate the southern part because it is an area dominated by built-up land, namely Ngaglik, Mlati, Depok and Gamping sub-districts (Figure 3).Meanwhile, in the northern part, namely the Turi, Pakem, and Cangkringan sub-districts, the moderately high-to-high class dominates this area.In detail, the percentage of vegetation class based on the NDVI value is nonvegetation (1%), very low (5%), low (12%), moderately low (19%), moderately high (25%), and high (34%) (Figure 4).The NDVI value has the opposite value to the LST value, where a low NDVI value has a high LST value.Conversely, the higher the NDVI value, the lower the LST value (Figure 4).This shows that the presence of vegetation affects the land surface temperature [6,26,27].

Land Surface Temperature
The NDVI value is the opposite of the LST value, i.e. the high density of vegetation causes the surface temperature to be lower.The LST statistical values in the Sleman area were recorded to have a minimum temperature of around 10ºC (Table 1), as shown in Figure 4, the areas that have minimum temperatures are concentrated in the northern part of the study area, namely Turi, Pakem, and Cangkringan districts in the peak area of Mount Merapi.While the average temperature from the results of the LST value is 23.15ºC that is the temperature that is spread over the central region and is the most dominating value in Sleman Regency.Most of the areas that are colored red or have the highest temperature of 30.47ºC are scattered in the southern part of Sleman, such as in the sub-districts of Gamping, Mlati, Ngaglik and Depok.These four sub-districts have high LST scores due to a high population.Urbanization creates an evolving inverse relationship between impermeable cover and vegetation, and generates new LST patterns due to the correlation of LSTs with impermeable cover and vegetation [15].Table 3 shows the four sub-districts that have the highest population numbers in 2020 and 2021, namely Gamping, Mlati, Ngaglik, and Depok, with each having a population of more than 100 thousand people, accompanied by a high population density of over 2.7 thousand people/km².So, the higher the population density, the higher the LST value of the area.Human-influenced areas with large settlements will have higher temperatures [36,37].

Soil Moisture Index
To determine the soil moisture index (SMI), requires two variables in the form of surface temperature index (LST) and vegetation density (NDVI) with both variables affecting the moisture level of a soil.High surface temperatures can cause soil moisture and vegetation levels to experience a deficit or decrease [38].SMI analysis results show an average humidity value of 0.37 which is spread over most of the central part which is light blue (Figure 6).The northern part of the Turi, Pakem and Cangkringan sub-districts mostly have high SMI reaching a maximum value of 1 which has extreme wet soil conditions.Meanwhile, the minimum value of the SMI is 0 which indicates the potential for extreme dry conditions are spread across the southern region, namely parts of the Gamping, Mlati, Ngaglik, and Depok sub-districts or those that are colored white in Figure 6.  Figure 7 shows the correlation between the NDVI value and SMI (R=0.46),meaning that the SMI value has a fairly strong correlation with the NDVI value.In the graph, its observable that the higher value of the vegetation density will be followed by an increase in the soil moisture value.Several previous studies have shown that the relationship between NDVI and soil surface moisture has a strong positive correlation [10,22].The results of the drought index classification show that the Sleman region is dominated by the severe drought class (Figure 9).Meanwhile, the extreme drought class is distributed in the central region (Figure 8), which is generally built-up area.The results of testing the SMI values and the drought index of several samples in built-up areas are presented in Table 4   [39].However, the drought prediction results based on NDDI in Figure 8 are largely determined by the image conditions used, as well as the SMI value.Meanwhile, drought is a complex phenomenon and its monitoring depends on the availability of good quality data, so the performance of the drought index can vary from one region to another [40].6) and high NDDI (Figure 8).High NDDI is clustered in the southern part of Sleman Regency, namely in the sub-districts of Depok, parts of Mlati, Godean, and Kalasan (Figure 8).This means that the SMI value is the opposite of the NDDI, where the higher the SMI, the lower the NDDI index and vice versa [41,42].When the soil moisture index falls below a certain threshold, this indicates that the soil is drier than usual, which can lead to drought conditions [43].

Conclusions
This study uses the soil moisture index to monitor potential drought areas in Sleman district, Yogyakarta.It is known that the presence of vegetation greatly influences soil moisture levels with a fairly strong correlation (R=0.46).For several reasons, the SMI map could not be validated, but the results appear logical and show low humidity index values that are spatially distributed in areas with a dominance of built-up land, namely the southern part of the study area such as Depok, Gamping, Mlati and Ngaglik sub-districts.Based on secondary data, the four sub-districts are associated with the highest density compared to other sub-districts.Future studies may consider other parameters to perform temporal monitoring of drought in urban areas using higher resolution imagery.Effective drought management can be achieved by monitoring current drought conditions and predicting future droughts.This research is limited to one year, similar research needs to be carried out in the future to monitor drought temporally and see the effect of development of built-up land on soil moisture and drought potential.

Figure 3 .
Figure 3.The NDVI Map of Sleman Regency in 2022

Figure 4 .
Figure 4. Percentage of NDVI values in Sleman Regency

Figure 5 .
Figure 5.The LST Map of Sleman Regency in 2022

Table 3 .
LST Statistical Values

Table 4 .
SMI Statistical Value

Table 5
shows that the SMI value correlates with the drought index.several samples with extreme drought classes are built-up land.This proves that soil moisture is a quantitative indicator of drought

Table 5 .
NDDI value and drought classes

Table above
shows several sample points that have the "severe drought" class, namely areas that have low SMI (Figure