SAHAJA: Development of a Cloud-Based WebGIS-Multi Criteria Decision Analysis for Agri-aquacultural Site Suitability in Central Java

Central Java Province, Indonesia, is endowed with vast agri-aquacultural potential. However, harnessing the full benefits of spatially enabled technology for supporting integrated sustainable agriculture and covering wide areas remains a challenge. This research aims to bridge this gap by developing a cloud-based WebGIS platform “SAHAJA” which integrates Multi-Criteria Decision Analysis (MCDA) for assessing agricultural site suitability. The study utilizes the Google Earth Engine incorporating agricultural parameters such as SAVI, soil properties, rainfall, topography, and Land Surface Temperature as well as aquacultural parameters such as chlorophyl-a, Total Suspended Soil, and Sea Surface Temperature. The platform provides users with the option to select predefined values for assessing site suitability for various purposes such as sugarcane cultivation, tea plantation, and marine fishing. Additionally, it offers an intuitive chart, allowing users to gain valuable insights from the data. The developed WebGIS serves as a decision support tool for both communities and stakeholders involved in land use planning, agri-aquacultural investment, and sustainable practices, thereby supporting smart city initiatives. The urgency of this research stems from the significant agri-aquacultural potential in Central Java and the need to effectively utilize spatially enabled technology to enhance integrated and sustainable agri-aquaculture. By addressing the challenge of wide area coverage, this study contributes to advancing the adoption of spatial technology in agricultural decision-making processes, fostering the productivity and resilience of Central Java’s agri-aquacultural sector.


Introduction
Central Java is one of the largest provinces in Indonesia in terms of territorial coverage and the potential of its agricultural and fisheries sectors.According to data from the Department of Agriculture and Plantation of Central Java Province [1], Sugarcane and Tea are two prominent commodities possessed by Central Java.Sugarcane (Saccharum officinarum L.) serves as the primary raw material for sugar production, while Tea (Camellia sinensis) stands as one of the most consumed beverages in the world [2].Furthermore, in terms of geospatial distribution, Central Java encompasses both northern coastal areas facing the Java Sea, including districts such as Pemalang, Pati, Demak, Batang, 1264 (2023) 012002 IOP Publishing doi:10.1088/1755-1315/1264/1/012002 2 Pekalongan, Tegal, Brebes, Rembang, Kendal, Pekalongan City, Semarang City, and Tegal City, as well as southern coastal areas facing the Indian Ocean, including regions like Wonogiri, Purworejo, Kebumen, and Cilacap.This geographical configuration bestows abundant marine wealth potential upon Central Java, particularly its maritime domain [3].The Grouper fish (Epinephelus lanceolatus) contributes to 26.5% of the dominant commodities in Indonesia's marine aquaculture sector [4], and approximately 4% of this contribution originated from Central Java Province in 2018 [5].These data underscore the ongoing efforts by the local government, specifically the Provincial Government of Central Java, to continually update statistical data and disseminate it to the public.However, the data often takes the form of numerical values at the smallest administrative units, such as city or district levels, as provided by the province, and sub-district levels, as provided by the city or district authorities.This practice, however, neglects the potential for agricultural and fisheries development at scales smaller than administrative village units [6].Therefore, the application of spatio-temporal geospatial analysis and visualization technology becomes imperative to unearth the full potential of suitable locations for agriculture and fisheries across time for the three aforementioned key commodities in Central Java.
The utilization of Web-accessible computing services and a Web-based Geographic Information System (WebGIS) stands out as the most effective means to analyze and present comprehensive multitemporal geographical information [7].The concept of Web GIS is the ability to provide an overview of spatial information based on spatial data which is useful for providing information in decision making.To support the implementation of WebGIS, you can utilize open-source software, so the costs are cheaper [8].Google Earth Engine (GEE) emerges as a cloud-based geospatial technology that provides access to extensive datasets and open and free analytical capabilities.GEE has gained widespread use in analyzing environmental conditions, weather patterns, land changes, and land suitability assessments [9,10].However, it is important to note that GEE fundamentally isn't designed as a geovisualization platform, resulting in limitations in digital cartography features.Thus, a separate platform is essential for handling data visualization in the form of WebGIS.This can involve leveraging extensions such as the PostGIS spatial database, spatial data interoperability standardization with Geoserver, map display through OpenLayers [11], interactive chart presentation through ChartJS, all integrated within the NGINX web server.
Previous researchers have conducted various land suitability assessments for plantations and fisheries.For instance, tea plantation suitability was mapped in a limited area in Batang Regency, Central Java, using medium-resolution satellite imagery like Sentinel 2B and Landsat 8 [12].Additionally, a single-temporal sugarcane plantation analysis for the year 2015 in Central Java Province was carried out using Geographic Information System-based spatial analysis with several secondary datasets [13].In the context of fisheries, catch fisheries were assessed in the Fisheries Management Zone of Jepara Regency using Sentinel 2A satellite imagery [14].Beyond this, studies have explored the utilization of GEE to analyze land conditions and changes in Kulon Progo Regency [9], as well as general agricultural land suitability assessments visualized through WebGIS in the Abbay Basin, Ethiopia [15].These studies each have their specific research focus, yet none have performed multiepoch analysis using GEE followed by intuitive visualization through WebGIS.
Hence, the primary objective of this study is to perform a spatio-temporal analysis of the suitability of locations for Sugarcane, Tea, and Grouper fisheries in Central Java using cloud computing technology GEE.This analysis will then be translated into a WebGIS platform named SAHAJA "Kesesuaian Lahan Jawa Tengah" to effectively display and provide valuable insights regarding the spatial data to users.The novelty of this study lies in the utilization of open spatial data sources and platforms, employing the cloud computing platform GEE, for conducting Multi Criteria Decision Analysis (MCDA) on multi-temporal data.Furthermore, the study innovatively presents the outcomes through an interactive WebGIS.This approach is applied within the context of site suitability assessment for sugarcane, tea, and grouper fisheries in the region of Central Java.Data & Methods

Conceptual Framework
The conceptual framework of SAHAJA, as depicted in Figure 1, encompasses GEE as the computing back-end and a ReactJS-based web-client as the front-end.This front-end architecture is supported by PostGIS extension serving as the spatial database, Geoserver facilitating the translation of spatial data into Open Geospatial Consortium (OGC) standards, Gunicorn's FastAPI for spatial queries, and OpenLayers for map visualization.Additionally, ChartJS is employed to render interactive graphs.The front-end serves as the Graphical User Interface (GUI), enabling users to display site suitability layer maps, select locations within administrative boundaries down to sub-district levels for sugarcane and tea site suitability maps, decide map years, and generate interactive charts to ascertain the percentage distribution of each site suitability class.

Figure 1. Conceptual Framework and technologies used in SAHAJA
On the back-end side, GEE serves the purpose of accessing open global big datasets, such as satellite imagery and other raster data.This cloud computing platform is further utilized for conducting multi-temporal spatial analysis using the provided Application Programming Interface (API).The spatial analyses conducted involved data aggregation, spatial statistics, weighting & spatial data overlays (MCDA Process), and reclassification, culminating in the creation of site suitability maps.These maps are subsequently exported to Google Drive which will automatically store in a PostgreSQGL database extended with PostGIS.All spatial queries initiated by users in the GUI to confine analysis to specific administrative locations are executed within PostGIS.Following data processing within the database, the information is ready for conversion into Open Geospatial Consortium (OGC) standards via Geoserver using Geoserver Representational State Transfer Architectural (REST) API, resulting in a Web Map Service (WMS) as the output.This WMS is what will eventually be presented on the web-client, seamlessly integrated with its corresponding charts.
Infographic/chart data is derived from calculations conducted on the backend using Python.When the "Analysis" button is clicked, the API triggers backend analysis processes through the POST method.The parameters utilized for analysis incorporate (1) the chosen masking layer, presented in GeoJSON format, representing either sub-district administrative boundaries or marine management areas, (2) the type of land suitability data, (3) the initial year, and (4) the final year.
The raster Geotiff site suitability data stored on the server is selected based on the chosen land suitability type and designated years.Subsequently, the chosen Geotiff file is read and converted into a matrix with values corresponding to digital numbers using the rasterio library.The rasterio's masking function is employed to apply a mask to the land suitability matrix using the polygon masking.
Digital numbers from individual cells within the polygon masking coverage are classified based on the four land suitability classes: 1=Not suitable, 2=Moderately Suitable, 3=Suitable, and 4=Highly Suitable.The percentage area of each class is calculated by comparing the number of cells for each class against the total number of cells within the polygon masking coverage.The calculated area proportions for each class are presented as JSON-formatted data through an API.This data is subsequently consumed on the frontend and presented in the form of bar charts using ChartJS.Additionally, Docker's containers are leveraged to simplify every component installation and deployment processes within this developed system.
To ensure the operational functionality of the web-client, the NGINX web server is used as the routing platform.Simultaneously, SAHAJA offers an APIs that enables third parties to access its data.This comprehensive system design positions SAHAJA as an entirely open platform, emphasizing interoperability and data sharing.This is in line with the core principles of Indonesia's One Map Policy, a fundamental aspect of the national smart city initiatives as outlined in Presidential Regulation of the Republic of Indonesia No. 23 of 2021 [16].

Study Area and Timeframe
The scope of this study is centered on Central Java Province, Republic of Indonesia, extending its coverage from the most granular administrative tier to the sub-district (kecamatan) level.Consequently, this research comprehensively covers 537 sub-districts distributed across 29 districts (kabupaten) and 6 cities (kota).Furthermore, the analysis also includes the District Fisheries Management Zones up to a distance of 4 Nautical Mile (Nm), adhering to Article 14 Paragraph 6 of the Republic of Indonesia Law No. 23 of 2014 [17].This involves 14 districts and 3 cities situated on both the northern and southern shores of Central Java Province, as shown in Figure 2. The entirety of the analysis was conducted in a multi-temporal fashion over a three-year observation period spanning 2020 -2022.

Data
From a range of literature sources, several data requisites have been identified to comprehensively address the analysis of suitability for sugarcane, tea, and grouper fisheries locations (as outlined in Table 1).These necessities entail vector data outlining administrative boundaries as well as the 4 Nautical Mile (Nm) District Fisheries Management Zone.Furthermore, raster data are pivotal, using Landsat 8 Operational Land Imager (OLI) Top-of-Atmosphere (ToA) & Surface Reflectance (SR) products, daily precipitation, soil properties, topography, and bathymetry [2,4,12,13,18].

General Site Suitability Mapping
Site suitability mapping is executed employing the Multi-Criteria Decision Analysis (MCDA) methodology as the foundation for conducting a weighted overlay analysis.MCDA facilitates mapping through the utilization of multiple data layers that interact according to predetermined weights [15].The MCDA process was done using GEE to efficiently manage large multi-temporal spatial datasets, thereby expediting analysis without significantly taxing local computer resources.The output mapping results from GEE materialize as GeoTIFF files stored within Google Drive.These maps consist of four distinct equal interval suitability classes: highly suitable, suitable, moderately suitable, and not suitable.The Site Suitability Mapping is depicted in Figure 3.

Sugarcane and Tea Site Suitability Mapping
In this study, the classification guidelines outlined by the Food and Agriculture Organization (FAO) were employed to categorize the land suitability for both sugarcane and tea crops [22].The parameters utilized to assess the suitability of locations for these commodities were streamlined and treated as equivalent, comprising aspects like Water availability, Soil Properties, Topography, Spectral Indices, and Climate.These five factors were identified as the pivotal determinants influencing the land suitability for both commodities [23].Concerning the water availability parameter, the daily rainfall data from the Climate Hazards Group InfraRed Precipitation with Station Data (CHIRPS) were aggregated into annual precipitation data using the Reducer: Sum function within GEE.Moving to the soil properties parameter, it was divided into three categories: soil texture, soil pH, and soil salinity.The soil texture classification aligned with the categorization prescribed by the United States Department of Agriculture (USDA), while soil pH data at a depth of 10 cm were extracted from OpenLandMap records.Additionally, the third facet of the soil properties parameter, namely soil salinity, was sourced from the Global Soil Salinity Maps.The topographical data utilized for this study comprises the Digital Elevation Model Nasional (DEMNAS), which subsequently underwent an analysis of slope and elevation classification.Moreover, in order to identify the existing plantation sites, Topographically Corrected Images were used.These images served as a basis for calculating vegetation indices, specifically the Soil-Adjusted Vegetation Index (SAVI).The selection of SAVI was predicated on its capability to emphasize the vegetative reflection characteristics while simultaneously mitigating the influence of soil background effects [25] increasing the accuracy of vegetation estimate.SAVI-derived images were employed as a surrogate to cartographically represent the fractional canopy cover (FC) within the sugarcane and tea plantation area [12].SAVI Equation is as follows: Where, NIR and RED denote the spectral reflectance measurements taken from the near-infrared (NIR) and red wavelength bands, respectively, while L stands as a soil background correction parameter.The value of L was established as 0.5, a consistent factor derived from outcomes of field trials within agricultural and pasture investigations, effectively encapsulating the most accurately modeled impact of soil on various vegetation species [25].
The concluding parameter utilized is associated with Land Surface Temperature (LST), which is derived from the Thermal Band of the Landsat 8 Surface Reflectance dataset.The Google Earth Engine (GEE) platform offers access to LST data without the need for intricate computations, achieved by retrieving the ST_B10 band within the Landsat 8 Level 2, Collection 2, Tier 1 dataset [26,27].By default, the LST obtained from this band is in kelvin units, necessitating a preliminary conversion to Celsius.The values for each class and parameter are indicated in Table 2.

Fishing Ground Mapping
In contrast to sugarcane and tea, the mapping of the fishing grounds for grouper fish is naturally centered on marine regions.As depicted in Figure 3, the fishing grounds are mapped using multiple parameters such as seabed depth/bathymetry, chlorophyll-a concentration, Sea Surface Temperature (SST), and Total Suspended Solid (TSS).The ETOPO dataset from NOAA is employed for extracting seabed depth, while the estimation of the other three factors is carried out using Landsat 8 data [4].The estimation of chlorophyll-a concentration can be carried out utilizing Landsat 8 Top of Atmosphere (TOA) data, following the formula proposed by Poddar et al. (2019) [28].Meanwhile, the estimation of Sea Surface Temperature (SST) and Total Suspended Solid (TSS) is accomplished using equations put forth by Kangkan (2006) [29].Subsequently, each parameter is weighted according to four classes as described in the following Table 3: 3.

Result and Analysis
According to research conducted by Sholikhah et al. ( 2021), GEE can perform calculations known in GIS as zonal geometry [10].These calculations are conducted to determine the area in km 2 for each suitability class at the provincial administrative boundary level in Central Java as presented in Table 4. From the obtained data, it can be observed that the classes for each commodity exhibit fluctuations, yet they maintain a relatively stable pattern, particularly within the suitable class category.The results of the suitability mapping for sugarcane, tea, and grouper fish locations can subsequently be accessed through SAHAJA by users.SAHAJA consists of primary features as well as supporting functionalities.These supporting functionalities are common features present in WebGIS, designed to facilitate map control tasks such as managing layer visibility and opacity, geocoding, geolocation, zooming in/out, and activating full-screen mode whereas these features are not available comprehensively and in a user-friendly manner on the AgriSuit platform developed by Yalew et al. (2016) [15].Furthermore, SAHAJA can be executed more interactively and swiftly since the maps are accessed from a structured spatial database in the form of WMS provided by GeoServer, rather than being stored in their raw file format [11].The core features offered by this WebGIS include the ability to perform location suitability analysis, allowing users to select locations down to the sub-district level throughout the entire Central Java Province.Moreover, users can opt for different suitability map layers and generate charts as part of the analytical process.Users of SAHAJA can access the map page through an online web browser.To conduct land suitability analysis, users will be prompted to select the Land Suitability Type, which comprises tea plantation, sugarcane plantation, and fisheries.The chosen Land Suitability Type determines the type of polygon masking employed for the analysis.If the user selects tea or sugarcane plantation as the Land Suitability Type, the polygon masking utilized will be the administrative boundaries of districts/cities and sub-districts as it shown in Figure 4.
Conversely, if the chosen Land Suitability Type is fisheries, the polygon masking will pertain to the marine management areas within the districts (Figure 5).The options for district boundaries, subdistricts, and marine management areas are asynchronously retrieved with 'get' capabilities from an integrated API linked to the database.Once the polygon masking is determined, users will be prompted to specify the initial and final analysis years.Subsequently, they can proceed by pressing the "Analysis" button.In the Figure 6, the outcomes of the analysis will be exhibited in distinct sections as collapsible panels containing bar graphs.These graphs will visually represent the proportionate extent of site suitability categories (unsuitable, moderately suitable, suitable, and highly suitable) relative to the total area covered by the corresponding masking layer for each respective year.As an illustration, in the scenario where a user opts for sugarcane land suitability assessment within the Sidoharjo sub-district spanning from 2020 to 2022, the graphical representation will demonstrate the distribution of each suitability class's area as a percentage of the entire Sidoharjo sub-district area over the mentioned years.The strength of this WebGIS lies in its capacity to visualize data, providing users with more substantial insights compared to mere map displays.Furthermore, since the digital maps shown are essentially Web Map Service (WMS)-based, the map loading process resulting from user interactions is significantly expedited compared to maps displayed without WMS.Moreover, SAHAJA's capability to select administrative boundaries down to the sub-district level brings benefits not only to the provincial level stakeholder but also to district governments and the general public.
Nonetheless, there are certain limitations.For instance, some weighting processes are inconsistent with expert judgements, the diversity of commodity types is not yet fully available, and there are constraints regarding the analysis timeframe.However, given that SAHAJA is a GEE-powered WebGIS, these limitations are likely to be easily overcome in further development.

Conclusion
This study introduces the SAHAJA framework for sugarcane, tea, and grouper fishing ground site-suitability based on MCDA, implemented on the GEE cloud computing platform.The framework empowers users to select commodity layers and visualize them, set administrative boundaries down to sub-district levels, and generate charts, enhancing the valuable insights derived from map data.SAHAJA presents numerous opportunities for site suitability assessment accessible to stakeholders at national, provincial, district, and sub-district levels, as well as the general public.Furthermore, this framework is fully open source and freely accessible.However, there are specific limitations inherent to the study's framework, including challenges associated with the weighting process (discrepancies in expert assessments).As the next step in the SAHAJA development process, we recommend expanding the foundational data and commodity types, extending the timeframe, strengthening the MCDA algorithm, and transitioning SAHAJA to an online platform, given that this current WebGIS is still in the prototyping stage.

Figure 2 .
Figure 2. Districts Boundary and 4 Nm District Fisheries Management Zone in Central Java Province

Figure 3 .
Figure 3. Site Suitability Mapping Process

Figure 4 .
Figure 4. Layer and boundary administration selection

Figure 5 .
Figure 5. Fishing ground layer will visualize district management zone

Figure 6 .
Figure 6.Result of analysis triggered by user

Table 1 .
Data sources

Table 3 .
Parameters weighting of fishing ground.

Table 4 .
Site suitability of Sugarcane, Tea, and Grouper fishing ground area in Central Java Province