Spatial basis risks and weather indexed crop insurance
by Gayathri S*
*Gayathri S, Programme Leader, DHAN Foundation
Agriculture is the livelihood of most households in India. Nearly 50 percent of labour force is engaged in agriculture. Indian farmers are faced with multiple production and market risks. The farm output in India largely depends on weather parameters, of which rainfall is the key factor. About 60 percent of land area in India is under agriculture, of which 61 percent is under dry land farming. In the absence of perennial rivers, these areas depend only on monsoon rains. India gets rain from two monsoons namely, Southwest monsoon and North east monsoon contributing 80 percent and 20 percent of annual rainfall respectively. However, vagaries in the rainfall in terms of distribution and quantity make farming the worst nightmare. Despite the availability of vast stretch of cultivable land and ample human resources, farming does not give the due comparative advantage because of high involvement of weather risks. The farming households often find it very difficult to cope up with these risks by adopting traditional instruments such as community help, saving for contingencies, managing the loss by taking loans etc., apart from appropriate agronomic practices. However these are not adequate and farming households are left with high debts, resorting to migration to other places or other livelihoods. Adequate investments are not being made in the farming sector in general and dry land farming in particular. This poses a great challenge to the economy as a whole by not only affecting the food supply but also by limiting the supply of raw materials to the firms affecting all players in the flow.
Properly designed and administered crop insurance programs could offer effective solution to mitigate the production risk in farming. Weather index based insurance is one such solution with advantageous features over the traditional yield based indemnity products. The yield based crop insurance programs indemnify the farmer based on the loss in yield. Such indemnity products involve high cost of administration besides suffering from issues such as moral hazards and longer time taken for ascertaining Spatial basis risks and weather indexed crop insurance the yield loss. This makes the product unaffordable and does not provide timely compensation to farmers who are forced to deal with the situation by taking loans.
On the other hand, weather indexed insurance program has the advantage of ascertaining the likely loss in the yield by objectively measuring the weather parameter that determine the yield. Further, instrumentation and automisation of measurement of the chosen weather parameter helps in getting real time weather data which minimizes the errors and also results in quick payout to the farmers in the event of adverse outcome. However, weather indexed insurance suffers from the existence of basis risks. Basis risk is the situation that arises when the index does not correlate with the actual loss. This flaw can be minimized by computing a weather index that perfectly correlates with the loss situation. There are three types of basis risks that concern those
dealing with weather indexed insurances.
Product basis risk arises when there is no clear-cut relationship noticed between loss and the indexed weather peril (e.g. rainfall). For example, yield loss in mango may be more due to wind speed during flowering rather than quantity of rainfall or relative humidity. Temporal basis risk arises due to interannual variations in seasonal crop phases which mean that the insurance phases are not temporally aligned with the intended crop growth stage. This may happen due to changes in the sowing dates. Sowing decisions are taken based on the onset of monsoons. However, the weather indexed crop insurance products are designed with assumptions on sowing period. When the actual sowing period changes, the critical crop growth stages would not coincide with the critical periods of risks, assumed in the crop insurance product design. Spatial basis risk arises due to local variations in the occurrence of the peril (e.g. rainfall) within the area surrounding a weather station. The research study highlighted in this article deals with the spatial basis risk and the need for having optimum reference rain gauge
stations for measurement of more accurate rainfall data.
Spatial variation in rainfall
The spatial variation of rainfall is largely determined by the spatial and temporal variations of the vertical motion of air needed to cool the air and condense the water vapor contained in it. There are three types of vertical motion of air namely, convective vertical motion, bounded vertical motion and general vertical motion.
Convective vertical motion refers to the elevation of air in an area of one to 5 miles diameter and to a height of few thousand to more than 50,000 feet. This is the cause of rainfall in tropical regions of India especially during summer season. This is the type of rainfall that shows very high variation within a very short horizontal distance. Hence, it calls for a dense network o f rain gauges.
Bounded vertical motion of air occurs over a band of 5 to 50 miles. Hence, within this band the rainfall is fairly uniform without much variation. This is the phenomenon causing rainfall during monsoon seasons and rainfall in humid regions. However, variations are significant in shorter periods among the shorter distances.
General vertical motion is caused by lifting up of air due to depression resulting in highly erratic rainfall.
In general, the spatial variability of rainfall is greater in arid and mountainous regions over small areas, than in humid and flat regions over large areas. Hence, the locations in the peninsular regions of India need denser network of rain gauges.
Measurement of rainfall data
Rain gauges are the instruments used to measure the rain fall. The number of rain gauges to measure the rainfall over an area depends on the purpose of measurement, size of the area, frequency of measurement, relative economic importance of measurement/ use of data. In India, rain gauge stations are commissioned by various organizations.
Indian meteorological organization is the premier organization which has established the weather stations in each district. Agriculture universities and research stations, state departments also have established rain gauge stations for their own purposes being crop modeling, forecasting calamities and disaster management, etc; crop insurance is rarely a purpose for establishing rain gauge stations. However, the insurance companies have an arrangement with these authorities for data sharing based on which the rainfall data are obtained and used for the administration of weather indexed crop insurance.
In crop insurance, the index of rain fall is taken as a proxy for the success or failure of crop. Unless the rainfall measured in the reference rain gauge is close representative of the actual rainfall of the farmers’ field, the basis risk will make the farmer lose his confidence on the mechanism of rainfall indexed crop insurance. This calls for having optimum number of rain gauges reasonably distributed in the area covering the farmers’ field. This in turn decides the economic feasibility and sustainability of the initiative. The World Meteorological Organization and the Bureau of Indian Standards have prescribed the following densities with regard
to the measurement instrument being rain gauges.
Standards of World Meteorological Organisation (WMO)
Standards of Bureau of Indian Standards
However, these standards are blanket recommendations and may not be suitable for the purpose of rainfall indexed insurance program. In order to reduce spatial basis risk, insurers often stipulate the maximum distance of 20
kilometers between the measurement instrument and the insured location. The stipulated distance of 20 kilometers between the weather station and the insured farm is the result of cost consciousness by the insurance providers and consequently the farmers not getting their legitimate claim.
DHAN’s pilot on rainfall indexed insurance program
In the absence of availability of weather indexed insurance products without basis risks, DHAN Foundation initiated a pilot project on rainfall indexed crop insurance with the support of International Labour Organization (ILO). About 159 automated rain gauges in fifteen locations in two states of India were installed for recording the
rainfall data. Each project location was given with 7 to 13 rain gauges. Placements of rain gauges were not decided on geometric distances but on potentials. The rain gauges were erected at strategic villages with mobile connectivity and where there are adequate members of farmer associations and there is existence of buildings without obstacles and there is willingness of building owners for installation of rain gauges. Thus, in practice the distance between rain gauges happens to be about 2 kilometers to 10 kilometers.
Product design and index decision are the key factors in minimizing product and temporal basis risks. Designing the product and product pricing involve identifying and quantifying the risk to be insured, computing the index based on the correlation between the weather parameter and the risk as well as deciding on how the index will be measured and monitored. The pilot envisages the involvement of farmer communities in evolving the design inputs such as critical periods of crop growth, rainfall requirement, amount of loss in the event of different level of peril, etc. Based on these inputs, indexed insurance product and premium are worked out actuarially.
The pilot confirmed the existence of micro climate zones which could be observed from the variation in rainfall among the rain gauges installed in the same block. Hence, a study was commissioned to estimate the optimum number of rain gauges required for each block with in a geographical area of about 400 to 500 sqkms and to identify the appropriate location for installing those rain gauges.
The study used the tools of coefficient of variation method to determine the number of rain gauges and Theissen Polygon method to identify the location for installing rain gauges prescribed by Bureau of Indian Standards (IS 4986: 1983 and IS 5542: 1969 respectively).
The Bureau of Indian Standards (BIS) gives the following formula for estimation of optimum number of rain gauges in an area by correlating the rainfall data of existing rain gauges in the area.
N = (CV/P)2 Where,
N = Optimum no. of rain gauges to be installed in the region
CV= Coefficient of variation of rainfall for the existing rain gauge network
P = Permissible degree of percentage error for the estimation of the average areal rainfall
(Lower the degree of error, higher will be the number of rain gauges required)
Whereas, Thiessen polygon method proposed by BIS, suggests a way of creating polygons of uniform weight, in terms of depth of precipitation. These polygons indicate the locality for setting up of rain gauges.
This article is written with reference to T.Kallupatti block which is one of the locations taken up for the study. T.Kallupatti is located in Madurai district of Tamilnadu state, India. Crops like sorghum, pearl millet, finger millet, maize and cotton are
grown through rain fed farming. Most of the farmers are small and marginal farmers with very small land holdings of less than a hectare. There are 7 village level rain gauges functioning in the block established by DHAN. There is a rain gauge station at Airport, Madurai, established by India Meteorological Department (IMD). It is about 40km from T.Kallupatti block. The rainfall data during the cropping season helps to understand the spatial variation in the rainfall between the rain gauges and the IMD rain gauge at the airport.
The above table emphasizes the need for having a dense network of rain gauges in the block. However, in order to understand the number of rain gauges required to minimize risk of spatial variation in rain fall, the coefficient of variation method was used as follow:
After arriving at the optimum number of rain gauges required for T.Kallupatti block, the next steps is to decide on the additional rain gauges and exact locations to install these additional rain gauges. This was done by adopting the Theissen Polygon method. Polygons were created to measure the area covered by each of the existing rain gauge.
Construction of the polygons occurs in four steps, illustrated below:
Step 1: Plot the locations of the stations and the boundary of the region on the map.
Step 2: Connect adjacent stations with straight lines.
Step 3: Construct perpendicular bisectors across the lines connecting stations.
Step 4: Connect the bisectors to outline the polygon belonging to each station.
Step 5: Count squares on the graph paper to determine the size of each area.
To determine the area covered by the rain gauge in the open polygon, extrapolation method was
Step 1: Equal length should be taken on the opposite side connecting the nearby rain gauge. This should be done taking into account all the nearby rain gauge stations.
Step 2: Construct perpendicular bisectors taking into account the mid-point of the line length taken on the opposite side.
Step 3: Connect the bisectors to outline the polygon belonging to each station.
Step 4: Areas of polygons are calculated and expressed as fractions of the total area.(P)
Step 5: Each station is weighted by multiplying the optimum number of raingauges (arrived earlier from CV method, in this case 12) and the ratio of the rain gauge area to total area for the respective rain gauge region. This gives the number of rain gauges to be installed in the respective region of the rain gauge. The difference between this and the existing number of rain gauges in the rain gauge region gives the number of additional rain gauges required for the respective rain gauge region.
The process of arriving at the additional rain gauges for T.Kallupatti block is given in the table below:
The exact location of the additional rain gauges is decided based on the availability of buildings suitable for installing rain gauges and the willingness of owners of the buildings.
The study is not without limitations. The co efficient of variation method does not have much limitation except for considering the appropriate level of error to be accounted in the formula. Higher the percentage of error, less will be the optimum number of rain gauges required. Moreover in the Thiessen polygon method, finding the area through extrapolation method may not resemble the exact ground reality. This would require correlation of data manually for a reasonably longer period.
Despite these limitations, the tools followed in the study are simple and it fairly reflected the rainfall experienced in the T.Kallupatti block during the study period. However, this study has to be continued for a reasonably longer period and further it has to be related to the other parameters like topography to arrive at some theories that could be applied in similar situations.