Hsin-chi LI1, Yung-heng Hsu1, Yi-Chen Chen2, Chuan-Zhong Deng2, Ya-Wen Hwang1
1 Climate Change Devision, National Science and Technology Center for Disaster Reduction, New Taipei City, Taiwan.
2 Policy and Socio-Economics Division, National Science and Technology Center for Disaster Reduction, New Taipei City, Taiwan.
Taiwan is located in a disaster-prone region, where on average, 3.6 typhoon events occur every year. Crops are highly vulnerable to weather systems and are easily affected by climate disasters. To gain a comprehensive understanding of the mitigation measures for crops, in this study, a system was built to evaluate the possible impact of typhoon-related disasters on crops. First, historical data on agricultural loss caused by natural disasters from 1950 to 2017 were collated. Based on this information, regression models were developed for different crops. Subsequently, a land use map of agriculture was generated to understand the distribution of individual crops. Third, a GIS-based technique was utilized to build an online evaluation model to estimate the possible loss of crops in Taiwan. This disaster assessment platform comprised four parts: 1) a disaster warning function: when the weather conditions reach a given threshold, the system will issue a warning to alert users; 2) disaster frequency inquiry: users can inquire into the frequencies of disasters caused by certain weather conditions in specific areas; 3) historical agriculture loss database: users can inquire into historical agriculture loss caused by each typhoon event; 4) loss assessment model: this provides an online evaluation function for each crop. The purpose of the system is to provide useful information to decision-makers to mitigate the effects of climate disasters on agriculture.
Keywords: typhoon, agricultural loss, regression model, GIS, database
Agricultural development is strongly related to climate conditions. Taiwan is situated across a tropical and subtropical zone. The weather conditions specific to this zone have shaped the development of Taiwan's agricultural industry. The risk of climate-related disasters is greater for Taiwan than for other countries. The National Science and Technology Center for Disaster Reduction (NCDR) was invited to take part in projects exploring agriculture and forest adaptation measures in 2016 to establish an early warning and loss assessment system for crops. Based on the goals of this project, NCDR began to conduct research into a loss analysis and evaluation system (Hsin-chi Li et al., 2014). According to historical data on agricultural loss from 2004 to 2014 (see Table 1), the major types of disaster affecting agriculture are typhoons, heavy rainfall, and cold damage. Furthermore, due to abnormal climate, ascertaining which climate change patterns in real time cause agricultural loss is the greatest challenge in recent years. Therefore, in this study, 55 agriculture zones containing crops of high economic value were selected to establish a real-time monitoring system to reduce the possible risk of agricultural loss.
Table 1. Number of disasters affecting crops from 2004 to 2014
The system had four major functions: disaster warning, disaster frequency inquiry, historical agriculture loss database, and loss assessment models. This system was designed for users who need to monitor instant weather conditions or quickly assess agricultural losses. Detailed descriptions of each individual function will be presented in the following subsections.
Disaster warning function
A series of weather monitoring stations have been built in 55 agricultural zones containing crops of high economic value. The stations are managed by the Central Weather Bureau. The data that are monitored include precipitation, wind speed, temperature, moisture, and duration of sunshine. All the information is collated onto the disaster warning web page of the agricultural loss system (as shown in Figure 1). The left-hand side indicates the total number of monitored locations in the agriculture zone; the right-hand side presents icons for warning signals. The three colors represent different risk levels, where the green light means “safety;” the yellow light means “low risk of damage;” and the red light means “high risk of damage.” Users can monitor the 55 agriculture zones on the webpage to obtain real-time information. Furthermore, as shown in the bottom of Figure 1, the webpage has an in-built timetable that provides weather forecasting information for the next seven days. Users can search for specific dates in this timetable, and if any forecasting information (e.g., precipitation, wind speed, temperature, or moisture) exceeds the set threshold, the signal light will turn yellow or red to indicate the potential risk.
Fig. 1. Interface of the disaster warning function
If any light in the monitored agriculture zone turns yellow or red, users can directly click into this area to read detailed weather information. The click-in webpage is shown in Figure 2. The main screen on this webpage shows hourly climate data, which include rainfall, wind speed, temperature, and moisture data from the Central Weather Bureau. It can tell users when the value exceeds the threshold (color band) and provides forecasting information for the next 24 hours. On the right-hand side of Figure 2, basic information on crops is presented, including planting schedules, the environment within the agriculture zone, and disaster prevention measures. All sources of information are organized as shown in Table 2.
Table 2. Data Sources
Fig. 2. The interface of the disaster warning function
To facilitate accurate management, the background setting of this system is important (as shown in Figure 3). For each crop, a control table and threshold setting for climate conditions are established. Figure 3 presents a setting example for fruit; the warning targets include rainfall, wind speed, highest temperature, lowest temperature, and moisture. The second and third columns show the yellow light and red right threshold settings. For example, when rainfall is less than the threshold, the color of the light is normally green. When the accumulated precipitation reaches 100 mm, the light will turn yellow. This means the agriculture zone is at a low risk of damage. When the accumulated precipitation reaches 300 mm, the light will turn red, which implies a high risk. In addition to the setting function for warning signals, the control table also has month setting functions due to the different planting schedules and weather conditions for each individual crop. To reduce the possibility of a false alarm, the warning information can only be issued in the setting month.
Fig. 3. Interface of the disaster warning function
Disaster frequency inquiry
Based on 5-km ´ 5-km grid climate data from 1960 to 2014, provided by the Taiwan Climate Change Projection and Information Platform Project (TCCIP), an inquiry system was developed to quickly research the frequency of disaster-related weather conditions (Yung-heng Hsu et al., 2017). Inquiry items include temperature inquiry (historical highest temperature per 10-day period, historical average temperature per 10-day period, and historical lowest temperature per 10-day period) and precipitation inquiry (accumulated precipitation per day, average precipitation of N continuous raining days, and accumulated precipitation of N continuous raining days), all of which are listed in Table 3.
Table 3. Inquiry conditions
Users can set the inquiry conditions in the right-hand side of the webpage, as shown in Figure 6. The weather database contains over 4000 grids, and each grid cell (5 km ´ 5 km) contains several types of weather information (listed in Table 3). When the location and weather criteria are chosen, based on the method (Yung-heng Hsu et al., 2017), the frequency ratio for the weather condition will be presented, as shown in Figure 4. To demonstrate this process, an inquiry into historical lowest temperature (X°C) will proceed as follows:
1. Choosing the location. (e.g., Nantou County Xinyi Township, longitude = 120.95, latitude = 23.85)
2. Choosing the time span (e.g., from 2004 to 2014)
3. Choosing the criteria of inquiry temperature ≤X°C (e.g., X = 10°C)
4. Calculating the occurrence of the inquiry temperature X within the indicated time period
5. Calculate the frequency ratio per 10-day period (occurrence of 10-day period divided by total days within the 10-day period)
6. Average frequency ratio over the set number of years.
When all the steps are complete, the results will be presented as a line chat, as shown on the left-hand side of Figure 4. The “blue line chat” represents the average frequency ratio over the period from 1960 to 2014; the “block line chat” represents the average frequency ratio within the 10-year period (from 2004 to 2014). Users can easily compare the results of the long-time scale with results in the latest 10 years. The figure shows that the frequency ratio of long-time scale and the latest 10 years are 0.34 and 0.26, respectively. This means that the frequency ratio under 10°C is gradually decreasing compared with the long-time scale. This is very useful information for farmers. For crops that are vulnerable to cold weather, the frequency of cold damage will be reduced, which is beneficial for their production. For crops that require cold conditions for growth, a useful adaptation strategy would be to plant early to avoid heat damage during this period.
Fig. 4. Interface of the disaster-related weather warning function
Historical agricultural loss inquiry
A database of historical agricultural loss was constructed for each typhoon event provided by the Council of Agriculture, Executive Yuan. Yung-heng Hsu et al. (2017) showed that agricultural loss is strongly related to typhoon level and routes. These two factors were used to build the inquiry database, where users can research agricultural loss at town-level by specific year, typhoon level, and routes (as shown in Figure 5). The steps for an inquiry are follows:
The top of Figure 5 presents the inquiry interface. For the search criteria inputted by users, the results are shown at the bottom of the webpage. The left-hand side of the map depicts the total accumulated precipitation according to the selected typhoon event. From this map, the distribution of rainfall can easily be analyzed to identify the region where it is most severe. The intermediate map presents town-level agricultural loss under the same accumulated precipitation, and the accompanying map shows the typhoon route. It is useful to present the relationship between the typhoon route and agricultural loss. This map is an interactive webpage where users can select a specific town, and the value of agricultural loss will be automatically presented. Finally, on the right-hand side, the most damaged crops and the approximate geographic location of the crops are also shown. The application of the historical agricultural loss inquiry database not only helps the decision-maker predict the possible impact from similar typhoons but it can also provide useful information for high-risk crops according to loss ranking so that the appropriate mitigation measures can be adopted.
Fig. 5. Interface of the historical agricultural loss inquiry
Agricultural loss assessment model
According to many studies (Lan-Fen Chu et al, 2007; Hsin-chi Li et al., 2014, 2015), several important factors are strongly related to agricultural loss, including climate factors (rainfall, temperature, moisture, and wind speed), time factors (season and month), and crop factors (yield, plant area, crop type, crop yield, and loss ratio). In this study, the assessment models were therefore built using the aforementioned factors. Two major models were built into this system: multi-loss model and loss ratio model. The multi-loss model includes an assessment of flood damage and cold damage. The flood damage function utilizes rainfall, wind speed, month, yield, plant area, crop type, crop yield, and loss ratio as predicting factors. In addition to these factors, two additional factors (temperature, moisture) are added in the cold damage function. The loss ratio model utilizes rainfall, wind speed, location, crop type, month, and plant area to assess the possible percentage of crops under specific climate conditions. The framework of the evaluation system is shown in Fig. 6.
Fig. 6. Interface of the historical agricultural loss inquiry
Based on this framework, an interface was developed to access information on individual crop loss. Figure 7 presents the operating interface for loss ratio assessment. The left-hand column is the inquiry table for specific factors, including the year, city, town, season, area, and 24-hour accumulated rainfall. When users select the inquiry information, the results of the loss ratio will be presented as a pie chart on the right-hand side.
Fig. 7. Interface of the loss ratio inquiry
The other operating interface for loss assessment is the multi-loss model shown in Figure 8. There are three input columns on the right-hand side and the inquiry steps are as follows:
1. Choose disaster type (includes flood damage and cold damage)
2. Choose crop type (each crop type has a different loss function)
3. Choose month (different seasons yield different loss functions for the same crop type)
4. Input inquiry table comprising specific factors, including planting area, price, and loss ratio under the chosen conditions.
When users complete their inquiry data, the multi-loss results will be presented as a box chat on the left-hand side. For example, on the left-hand side of Figure 8, the results from two models are shown. Users can objectively understand the extent of agricultural loss based on multi-model analysis. Furthermore, because every model has its own up-bound and low-bound, users can also determine the loss range for each model.
Fig. 8. Interface for multi-loss assessment
An online agricultural loss evaluation system was built to estimate or make inquiries into the loss of crops. This system will provide useful information to reduce possible agricultural loss. Based on historical agriculture data, disaster warning functions, inquiry database, and loss models were established. This system adopted GIS-based technology to create a spatial map that is easily read and understood by users. The system had four main functions: 1) disaster warning function; 2) disaster frequency inquiry function; 3) historical agriculture loss database; and 4) loss assessment function. This system can not only be used to access agricultural loss for past typhoon-related events but it can also estimate the possible impact of future typhoon events based on weather forecasting information. This will help reduce agricultural loss and enable suitable mitigation measures to be implemented.
Central Weather Bureau, historical typhoon database, http://rdc28.cwb.gov.tw/TDB/ntdb/pageControl/basic
Lan-Fen Chu, Chi-Chung Chen, Hsing-Jui Chen, An Estimation of Rice Damage Function by Typhoon and the Premium Calculation of Disaster Insurance in Taiwan, Taiwanese Agricultural Economic Review , 13(1), 37 – 67
Hsin-Chi Li, Lung-Sheng hsieh,Lien-Kuang Chen,Lee-Yaw Lin,Wei-Sen Li, 2014, “Disaster Investigation and Analysis of Typhoon Morakot” Journal of the Chinese Institute of Engineers. Vol.37, No. 5, pp.558-569.
Hsin-Chi Li , Shiao-Ping Wei, Chao-Tzuen Cheng, Jun-Jih Liou, Yong-Ming Chan, Keh-Chia Yeh ,2015, Applying Risk Analysis to Disaster Impact of Extreme Typhoon Events under Climate Change, Journal of disaster research, Vol. 10, No. 3, pp.513-526.
Yung-heng Hsu, Hsin-chi Li, Kwi-chin Huang, Ya-wen Hwang, Yung-ming Chen, 2017, Establishing the historical graph of climatic characteristic for Taiwan’s farmland, National Science and Technology Center for Disaster Reduction, New Taipei city, Taiwan