Animal husbandry drought loss assessment method

Abstract

The invention discloses an animal husbandry drought loss assessment method, and relates to the technical field of animal husbandry loss assessment, and the method comprises the following steps: S1, collecting multi-source dynamic data which comprises meteorological data, animal husbandry data, management data and economic data; s2, preprocessing the multi-source dynamic data to obtain standardized data; s3, constructing a dynamic coupling evaluation model, inputting the standardized data into the dynamic coupling evaluation model, and outputting classified and staged animal husbandry drought loss values; wherein the dynamic coupling evaluation model is associated with pasture growth parameters and livestock metabolism demand parameters, and a loss offset coefficient corresponding to a management intervention measure is introduced; and S4, generating an animal husbandry drought loss assessment report based on the loss value. According to the animal husbandry drought loss assessment method, four types of data of weather, animal husbandry, management and economy are fused, a dynamic coupling model is constructed, assessment errors are reduced, and the problem that traditional single data assessment is large in deviation is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of livestock loss assessment, in particular to a livestock drought loss assessment method. BACKGROUND

[0002] Livestock is an important part of agricultural economy, and its production stability is significantly affected by meteorological conditions, especially drought disasters, which can lead to reduced pasture yield, difficult drinking water for livestock, and further losses such as reduced stock levels and slow growth. Accurate assessment of livestock drought losses is of great significance for developing emergency measures and optimizing industry layout.

[0003] The existing livestock drought loss assessment methods have the following defects:

[0004] They rely heavily on drought indices or historical statistical data released by meteorological departments, ignoring the impact of pasture growth status, differences in livestock breeds, and management measures on the pasture, resulting in a large deviation between the assessment results and the actual losses. The loss differences between artificially supplemented pasture and natural pasture cannot be quantified, making it difficult to reflect the offsetting effect of management intervention on losses. A unified standard is used to assess the losses of different regions and different breeds of livestock, without considering the differences in pasture recovery capacity between arid and humid regions, and the loss rate differences between drought-tolerant breeds (such as Mongolian sheep) and moisture-loving breeds (such as Holstein dairy cows), resulting in poor applicability of the assessment results. Most of the existing methods are retrospective assessments after the drought ends, which cannot predict the loss trend in advance and cannot clearly identify the causes of the losses (such as whether the drought directly caused the losses or poor management exacerbated the losses), making it difficult to guide pasture to take timely emergency measures. Therefore, we propose a livestock drought loss assessment method. SUMMARY

[0005] The purpose of the present application is to provide a livestock drought loss assessment method to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides the following technical solution: a livestock drought loss assessment method, comprising the following steps: S1: collecting multi-source dynamic data, the multi-source dynamic data including meteorological data, livestock data, management data and economic data; S2: preprocessing the multi-source dynamic data to obtain standardized data; S3: constructing a dynamic coupling assessment model, inputting the standardized data into the dynamic coupling assessment model, and outputting livestock drought loss values by product category and stage; wherein the dynamic coupling assessment model is associated with pasture growth parameters and livestock metabolic demand parameters, and introduces loss offset coefficients corresponding to management interventions; S4: generating a livestock drought loss assessment report based on the loss values.

[0007] Further, in step S1: the meteorological data includes soil moisture obtained by satellite remote sensing, pasture coverage by unmanned aerial vehicle, and time series rainfall; the livestock data includes livestock intake, body weight change rate, and reproduction rate collected by Internet of Things devices in the pasture; the management data includes artificial supplementary feeding amount, drinking water facility coverage, and emergency transfer record; and the economic data includes time series fluctuation data of regional feed market price and livestock market price.

[0008] Further, in step S3, the construction process of the dynamic coupling evaluation model includes:

[0009] establishing a pasture growth rate model to calculate the pasture supply amount based on soil moisture and pasture coverage;

[0010] establishing a livestock metabolic demand model to calculate the feed demand amount per unit time based on livestock breed, body weight, and growth stage;

[0011] calculating the basic loss rate based on the difference between the pasture supply amount and the feed demand amount, combined with the loss offset coefficient, which is dynamically adjusted according to the supplementary feeding amount and the transfer frequency, and the loss offset coefficient is increased by 5%-8% for every 10% increase in supplementary feeding amount;

[0012] converting the basic loss rate to economic loss value combined with the price fluctuation in the economic data.

[0013] Further, in step S3, the sub-categories include high drought-tolerant livestock, medium drought-tolerant livestock, and low drought-tolerant livestock; the high drought-tolerant livestock includes camels and Mongolian sheep, the medium drought-tolerant livestock includes Simmental cattle and small-tailed Han sheep, and the low drought-tolerant livestock includes Holstein dairy cows and meat goats.

[0014] Further, in step S3, the sub-stages include the early stage of drought, the middle stage of drought, and the late stage of drought; the relative humidity of soil in the early stage of drought is 50%-60%, in the middle stage of drought is 40%-50%, and in the late stage of drought is <40%; different stages correspond to different loss amplification coefficients, the loss amplification coefficient in the middle stage of drought is 1.5-2 times that in the early stage, and in the late stage is 2.5-3 times that in the early stage.

[0015] Further, it further includes step S5: constructing a loss prediction module based on the LSTM long short-term memory network, inputting the meteorological time series data and livestock time series data in the past 5 years, predicting the potential loss rate under different drought development trends in the next 1-3 months, and outputting the warning level.

[0016] Further, in step S5, the warning level includes a blue warning (potential loss rate < 5%), a yellow warning (5%-15%), and a red warning (> 15%); the loss prediction module also performs attribution analysis on the potential loss rate by SHAP interpretable algorithm, and outputs the direct loss proportion of drought, the indirect loss proportion, and the management bias loss proportion.

[0017] Further, the direct loss of drought is the loss of livestock death and weight loss caused by pasture shortage, the indirect loss is the loss caused by the increase in feed prices and water cost, and the management bias loss is the additional loss caused by the lag of emergency measures.

[0018] Further, in step S2, the preprocessing includes outlier rejection, data normalization, and spatio-temporal alignment; the spatio-temporal alignment unifies the time granularity of different source data to daily level and the spatial granularity to 1km*1km grid unit.

[0019] The present application provides a livestock drought loss assessment method, which has the following beneficial effects:

[0020] The present application integrates meteorological, livestock, management, and economic data, constructs a dynamic coupling model, reduces the evaluation error, and solves the problem of large evaluation deviation of traditional single data; through the fine design of sub-category (drought tolerance level) and sub-stage (drought degree), combined with the regional correction coefficient, the result is more in line with the actual loss of different pastures. The introduction of the AI prediction module realizes the loss trend prediction, and the SHAP attribution makes the loss causes clear, which provides quantitative basis for the livestock to reserve feed and optimize management measures in advance, and breaks through the limitations of traditional post-evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 The present application provides a livestock drought loss assessment method, which has the following beneficial effects: DETAILED DESCRIPTION

[0022] The embodiments of the present application will be further described in detail below in conjunction with the drawings and examples. The following examples are used to illustrate the present application, but cannot be used to limit the scope of the present application.

[0023] The following is a detailed description of the livestock drought loss assessment method of the present application based on a certain large-scale grassland pasture located in Xilingol League, Inner Mongolia Autonomous Region as an actual application scenario. The pasture mainly breeds Mongolian sheep (high drought-tolerant livestock) and Holstein dairy cows (low drought-tolerant livestock), with a breeding scale of 50 Mongolian sheep and 5 Holstein dairy cows. The pasture is mainly dominated by grassland of Gramineae, and adopts a management mode of "natural grazing and artificial supplementary feeding". The pasture is affected by drought weather from spring to summer every year, and the present application method is used to accurately evaluate the drought loss and guide emergency management.

[0024] Step S1: Multi-source dynamic data collection

[0025] In this embodiment, the collection of multi-source dynamic data is achieved through the combination of "satellite, unmanned aerial vehicle, Internet of Things and manual recording", and the specific collection content and method are as follows:

[0026] The soil moisture data of the pasture in the 1km x 1km grid unit is obtained by the MODIS satellite remote sensing system, and the monitoring frequency is once a day. The average soil moisture in the evaluation period is 45%; the pasture coverage rate is obtained by the DJI agricultural unmanned aerial vehicle (model T40) for 2 times a week, and the average pasture coverage rate is 60% after image recognition processing; the regional automatic weather station data in the pasture is accessed synchronously to obtain the time series rainfall, and the highest daily rainfall in the evaluation period is 5mm, the lowest is 0mm, and the average daily rainfall is 2.3mm.

[0027] Livestock data: Mongolian sheep are equipped with intelligent collars to collect real-time feeding data of each sheep, and the average daily feeding amount of Mongolian sheep in the evaluation period is 1.2kg per head; the intelligent weighing platform is set up in the dairy cow breeding area, and the body weight change is recorded 3 times a week, and the body weight change rate is calculated as-1.2% per day (i.e. the average daily weight decreases by 1.2%); the breeding rate data is recorded through the pasture breeding management account, and there is no new born calf in the evaluation period, and the breeding rate is temporarily recorded as 0%.

[0028] Management data: The artificial supplementary feeding amount is recorded through the pasture feed management system, and the average daily supplementary feeding amount for Mongolian sheep in the evaluation period is 100kg (mainly silage feed), and the average daily supplementary feeding amount for dairy cows is 500kg (mainly mixed concentrate feed); the pasture drinking water facilities are checked on site, and there are 5 drinking water points covering all breeding areas, and the drinking water facility coverage rate is 100%; there is no extreme drought weather in the evaluation period, and no emergency transfer is implemented, and the emergency transfer record is "none".

[0029] The regional feed market price and time series fluctuation data are obtained from the "Inner Mongolia Livestock Market Monitoring Platform", and the average price of silage feed in the evaluation period is 2.8 yuan / kg, and the average price of mixed concentrate feed is 3.5 yuan / kg; the market price data is obtained from the local livestock market, and the market price of Mongolian sheep is 800 yuan per head, and the market price of Holstein dairy cows is 15000 yuan per head, and the price fluctuation range is less than 5% (no significant price anomaly).

[0030] Step S2: Multi-source dynamic data preprocessing

[0031] The multi-source data collected above is preprocessed to ensure that the data is standardized and can be used for subsequent model calculation, and the specific processing process is as follows:

[0032] Outlier rejection: adopt 3σ principle to filter abnormal data, for example, in the data of Mongolian sheep intake, it is found that there is a record of "single-day intake of 5 kg per sheep", which is more than 3 times the standard deviation of the mean value (0.3 kg, 3 times the standard deviation is 0.9 kg, mean value + 3 times the standard deviation = 2.1 kg), which is determined as an abnormal value caused by temporary sensor failure, and is deleted, and the average of the adjacent two days (1.1 kg and 1.3 kg, the average is 1.2 kg) is used to complete the missing data.

[0033] Data normalization: convert data of different dimensions to the interval [0, 1], for example, soil moisture 45% corresponds to normalized value 0.45 (calculation formula: normalized value = actual value / 100, because the maximum value of soil moisture is 100%), grass coverage rate 60% corresponds to normalized value 0.6, and Mongolian sheep intake 1.2 kg per sheep corresponds to normalized value 0.3 (based on the maximum intake of 4 kg per sheep, the calculation formula is: normalized value = actual intake / maximum intake).

[0034] Space-time alignment: unify the time granularity of all data to daily scale, for example, daily soil moisture data obtained by satellite remote sensing (if weekly, it is decomposed into 7 daily scale data by linear interpolation method); weekly grass coverage rate data obtained by unmanned aerial vehicle, combined with daily weather data trend (such as rainfall change) for daily scale fitting and completion; the spatial granularity is unified to 1 km x 1 km grid unit, the grass coverage rate data of the whole pasture obtained by unmanned aerial vehicle is divided and distributed to the corresponding grid according to the grid unit, to ensure that the weather data and livestock data are matched in space (i.e. daily scale weather data in each grid unit is matched with livestock data in the region.

[0035] Step S3: dynamic coupling evaluation model construction and loss value calculation

[0036] Based on the pre-processed standardized data, a dynamic coupling evaluation model is constructed, and the loss value is output in categories and stages according to the logic of "grass supply, livestock demand, loss calculation and economic conversion", the specific process is as follows:

[0037] Grass growth rate model and grass supply calculation: since the grass type of the pasture is Gramineae grassland, the grass type correction coefficient k1 is set to 1.0; by substituting the soil moisture normalized value 0.45 (corresponding to the actual value 45%) and the grass coverage rate normalized value 0.6 (corresponding to the actual value 60%), the formula "grass supply M = 0.3 x soil moisture actual value + 0.7 x grass coverage rate actual value x k1" is used to calculate, and M = 0.3 x 45 + 0.7 x 60 x 1.0 = 13.5 + 42 = 55.5 kg / day·km 2 (i.e. 55.5 kg of grass per day per 1 km 2Within each grid unit, the daily supply of forage is 55.5 kg.

[0038] Livestock Metabolic Requirement Model and Feed Requirement Calculation: For Mongolian sheep (highly drought-resistant, adult, average weight 50 kg), the following coefficients are set: weight coefficient a = 0.02 (adult livestock coefficient), growth stage coefficient b = 0.8 (adult livestock coefficient), and breed drought resistance coefficient k2 = 0.8 (highly drought-resistant breed coefficient). Substituting these values ​​into the formula "Feed Requirement D = (a × weight + b × growth stage coefficient) × k2" (where the growth stage coefficient is 0.8 and has no unit), the daily feed requirement per Mongolian sheep is calculated as D = (0.02 × 50 kg). +0.8×0.8)×0.8=(1+0.64)×0.8≈1.31kg / head·day; For Holstein dairy cows (low drought tolerance, adult, average weight 500kg), set the weight coefficient a=0.02, growth stage coefficient b=0.8, and breed drought tolerance coefficient k2=1.2 (low drought tolerance breed coefficient), and similarly calculate the average daily feed requirement per cow D=(0.02×500+0.8×0.8)×1.2=(10+0.64)×1.2≈12.77kg / head·day.

[0039] Basic loss rate calculation: First, calculate the total feed requirement. The total feed requirement for Mongolian sheep = 1.31 kg / head / day × 50 heads = 65.5 kg / day. The total feed requirement for dairy cows = 12.77 kg / head / day × 5 heads ≈ 63.85 kg / day. The total feed requirement for the pasture ≈ 65.5 + 63.85 = 129.35 kg / day. Since the available pasture supply M = 55.5 kg / day < the total feed requirement of 129.35 kg / day, the basic loss rate needs to be calculated. Based on soil moisture of 45% (corresponding to mid-drought period, with an amplification factor k3 = 1.8 for the drought stage), and supplementary feed amounts (100 kg / day for Mongolian sheep and 500 kg / day for dairy cows, with the supplementary feed amount accounting for approximately (100 + 500) / 129.35 ≈ 4.64 of the total feed requirement, and each 10% increase in supplementary feed amount corresponding to a 5%-8% increase in the loss offset coefficient, the calculated total loss offset coefficient c = 0.25), the formula "basic loss rate L0 = (1 - M / total feed requirement) × k3" is substituted into the formula. The basic loss rate of Mongolian sheep is calculated as L0 = (1-55.5 / 129.35)×1.8×(1-0.25)≈(1-0.43)×1.8×0.75≈0.57×1.35≈0.77 (i.e., 77%), and the basic loss rate of dairy cows is L0 = (1-55.5 / 129.35)×1.8×(1-0.25)≈0.77 (i.e., 77%). Since the total feed demand is calculated uniformly, the same basic loss rate is temporarily taken. In reality, it can be subdivided according to breed.

[0040] Economic loss value conversion: Combining the prices in the economic data, and substituting them into the formula "Economic loss value L = (L0 × number of sheep × slaughter price) + (L0 × supplementary feed amount × feed market price)", the economic loss value for Mongolian sheep is calculated as follows: L = (77% × 50 × 800) + (77% × 100 × 2.8) ≈ 30800 + 215.6 = 31015.6 yuan; the economic loss value for dairy cows is calculated as follows: L = (77% × 5 × 15000) + (77% × 500 × 3.5) ≈ 57750 + 1347.5 = 59097.5 yuan; the total economic loss value for the ranch is approximately 31015.6 + 59097.5 = 90113.1 yuan.

[0041] Step S4: Generation of Livestock Drought Loss Assessment Report

[0042] Based on the loss values ​​calculated above, a livestock drought loss assessment report for this ranch is generated. The core contents of the report include:

[0043] Loss distribution by breed: Mongolian sheep accounted for approximately 34.4% of the losses (31,015.6 yuan / 90,113.1 yuan), while Holstein dairy cows accounted for approximately 65.6% of the losses (59,097.5 yuan / 90,113.1 yuan). It is clear that the losses of low drought-resistant breeds (dairy cows) were significantly higher than those of high drought-resistant breeds (Mongolian sheep) during the middle of the drought.

[0044] Explanation of phased losses: During the assessment period, soil moisture remained stable at 45%, corresponding to the middle stage of drought, at which the loss rate was 100%. Comparing the loss amplification factor between the early stage of drought (soil moisture 50%-60%) and the later stage (<40%), if the drought continues to develop into the later stage (soil moisture drops to 35%), the total loss rate is expected to increase to 192.5% (77% × 2.5, with the later stage loss amplification factor taken as 2.5), and the total economic loss will increase to approximately 213,497.6 yuan.

[0045] Analysis of the effectiveness of management measures: The supplementary feeding measures (600 kg of supplementary feed per day) during this assessment period contributed 25% of the loss offset coefficient. It was calculated that the economic loss was reduced by approximately 30,037.7 yuan (90,113.1 yuan × (0.25 / (1-0.25))), which clearly shows that supplementary feeding is an effective means to reduce drought losses at the current stage; the drinking water facility coverage rate was 100%, and no additional losses were incurred due to drinking water problems.

[0046] Step S5: Loss Prediction and Attribution Analysis

[0047] Based on nearly 5 years of historical data from the ranch, a loss prediction module was constructed using an LSTM (Long Short-Term Memory) network, and attribution analysis was performed using the SHAP (Shape-Based Interpretive Algorithm). The specific implementation process is as follows:

[0048] Loss prediction module training and prediction: Using the past 5 years of daily meteorological data (rainfall, soil moisture) and monthly livestock data (livestock size, growth curve, feed intake) as the training set, an LSTM prediction model is constructed (the model is trained for 1000 iterations, and the loss function converges to below 0.05); inputting the meteorological trend data at the end of the evaluation period (no effective rainfall is expected in the next 15 days, and soil moisture will drop to 38%), the potential loss rate in the next 15 days under the "continuous drought" scenario is predicted to be 25% (that is, the total loss rate will increase from the current 77% to 102%), and the corresponding warning level is red warning (potential loss rate > 15%).

[0049] SHAP Attribution Analysis: The SHAP interpretable algorithm was used to break down the composition of potential loss rates. The results showed that direct losses due to drought accounted for 55% (mainly due to the accelerated weight loss of dairy cows caused by pasture shortage and insufficient feed intake of Mongolian sheep), indirect losses accounted for 30% (mainly due to the slight increase in feed market prices caused by regional drought, with the price of mixed concentrate feed expected to increase by 8%, increasing supplementary feeding costs), and management deviation losses accounted for 15% (mainly due to the failure to stockpile emergency feed in advance; if the drought continues, there is a risk of feed interruption in the next 5 days, resulting in additional losses). Based on the attribution results, an emergency recommendation was made to "urgently purchase 3000 kg of mixed concentrate feed to cope with the risk of feed interruption".

[0050] The embodiments of the present invention are given for illustrative and descriptive purposes only, and are not intended to be exhaustive or to limit the invention to the forms disclosed. Many modifications and variations will be apparent to those skilled in the art. The embodiments were chosen and described in order to better illustrate the principles and practical application of the invention, and to enable those skilled in the art to understand the invention and to design various embodiments with various modifications suitable for a particular purpose.

Claims

1. A method for assessing drought losses in livestock farming, characterized in that, Includes the following steps: S1: Collect multi-source dynamic data, including meteorological data, livestock data, management data, and economic data; S2: Preprocess the multi-source dynamic data to obtain standardized data; S3: Construct a dynamic coupling assessment model, input the standardized data into the dynamic coupling assessment model, and output livestock drought loss values ​​by category and stage; wherein, the dynamic coupling assessment model associates pasture growth parameters with livestock metabolic demand parameters, and introduces loss offset coefficients corresponding to management intervention measures; S4: Generate a livestock drought loss assessment report based on the loss values.

2. The method for assessing livestock drought losses according to claim 1, characterized in that, In step S1: the meteorological data includes soil moisture obtained by satellite remote sensing, pasture coverage rate and time-series rainfall obtained by drone aerial photography; the livestock data includes livestock feed intake, weight change rate and reproduction rate collected by pasture IoT devices; the management data includes artificial feeding amount, drinking water facility coverage rate and emergency relocation records; the economic data includes time-series fluctuation data of regional feed market price and livestock slaughter price.

3. The method for assessing livestock drought losses according to claim 1, characterized in that, In step S3, the construction process of the dynamic coupling evaluation model includes: Establish a forage growth rate model and calculate the available forage supply based on soil moisture and forage coverage. Establish a livestock metabolic demand model to calculate feed requirements per unit time based on livestock breed, weight and growth stage; The basic loss rate is calculated based on the difference between the available forage and the required feed, combined with the loss offset coefficient. The loss offset coefficient is dynamically adjusted according to the amount of supplementary feed and the frequency of relocation. For every 10% increase in the amount of supplementary feed, the loss offset coefficient increases by 5%-8%. By combining price fluctuations in economic data, the base loss rate is converted into an economic loss value.

4. The method for assessing livestock drought losses according to claim 1, characterized in that, In step S3, the classification includes highly drought-resistant livestock, moderately drought-resistant livestock, and lowly drought-resistant livestock; highly drought-resistant livestock include camels and Mongolian sheep, moderately drought-resistant livestock include Simmental cattle and Small-tailed Han sheep, and lowly drought-resistant livestock include Holstein dairy cattle and meat goats.

5. The method for assessing livestock drought losses according to claim 1, characterized in that, In step S3, the phases include the initial stage of drought, the middle stage of drought, and the later stage of drought; the initial stage of drought is characterized by a soil relative humidity of 50%-60%, the middle stage by 40%-50%, and the later stage by <40%; different stages correspond to different loss amplification factors, with the loss amplification factor in the middle stage being 1.5-2 times that in the initial stage and in the later stage by 2.5-3 times that in the initial stage.

6. The method for assessing livestock drought losses according to claim 1, characterized in that, It also includes step S5: constructing a loss prediction module based on LSTM long short-term memory network, inputting meteorological time series data and livestock time series data of the past 5 years, predicting the potential loss rate under different drought development trends in the next 1-3 months, and outputting the warning level.

7. The method for assessing livestock drought losses according to claim 6, characterized in that, In step S5, the warning levels include blue warning, yellow warning and red warning; the loss prediction module also performs attribution analysis on the potential loss rate through the SHAP interpretable algorithm, and outputs the proportion of direct drought loss, the proportion of indirect loss and the proportion of management deviation loss.

8. The method for assessing livestock drought losses according to claim 7, characterized in that, The direct losses from drought are livestock deaths and weight loss due to pasture shortages, while the indirect losses are losses due to increased feed prices and water costs. Management deviation losses are additional losses caused by delayed emergency measures.

9. The method for assessing drought losses in livestock farming according to claim 1, characterized in that, In step S2, the preprocessing includes outlier removal, data normalization, and spatiotemporal alignment; the spatiotemporal alignment unifies the time granularity of data from different sources to the daily level and the spatial granularity to a 1km×1km grid cell.

Images